When will autonomous vehicles be common in the US and elsewhere? How many positions for cabdrivers and large-truck drivers will there be in the next several decades? See the new projections, with reasons behind them, at http://worksnewage.blogspot.com/2019/06/driverless-vehicles-and-driving-jobs.html.
Any and every time I sign that I am about to change lanes, a car in the next lane over speeds up to block me in. It's an insane state of affairs. That driver was content where he was until he felt like I was "getting in front of" him. These drivers won't even continue until they are ahead of you...they just want to obstruct and deny you.
People are incredibly angry.
Someone said to me that the solution was to not signal, just go.
I disagree.
I signal, watch the guy next to me speed up and then I go. Then I ignore the shouting, high beams and one-finger salutes.
I don't know what an autonomous car would do in that situation, but I think it would never be able to get off the highway.
14
I’ve long been a self-driving car opponent until I test drove a Tesla that included autopilot driving. So try the tech first and your views may change as mine did.
1. Self-driving can be rather safe now during highway driving as long as the car has a driver.
2. Preventive autopilot (driver is in control but the car can take over if a danger arises) is such a great feature.
3. It is absurd to expect zero fatality from self-driving vehicles when driver mistakes harm masses. Do we not all wish to make traffic safer?
4. Yes, we’re not there yet but I believe we can change our roads and legislation such that 75% of the roads can be suitable in 5 years.
6
I enjoy driving a vehicle. I enjoy navigating, predicting hw to get to my destination more quickly by jockeying around the slow-pokes who hog the left lane and never look in their rear-view mirror.
Sorry, but there is no pleasure or challenge in a comptuer-driven vehicle.
Where SOCIETY should insist the developmental money go is into efficient roads and traffic lights. Every new vehicle now has GIS/GPS, but cities and municipalities have not translated those systems into the traffic lights sequencing and lane speed controls.
Many cities and municipalities let the pedestrians decide on traffic flow by giving control to the cross-walk buttons, creating start-stop cycles, consequent congestion, and worse, extra engine exhaust.
It seems that traffic engineers (or their budget-minded masters) are the least imaginative lot around.
6
Surely we could have done better than a photo from 2016 for this article on the state-of-the-art?
3
I like how the dinosaur companies like Ford, who also said Elon Musk could never build a car, and that electric cars are far down the road, are quoted here as if they have any idea what Google and Tesla are doing. Also given how bad humans are at driving it’s not going to take a lot of brilliant AI to do better.
6
No. Stop. Every Tesla performs the equivalent of 2 million hours of on road testing. 400,000 teslas are operating their autonomous driving functionality continuously in shadow mode and comparing the responses to those of the vehicle drivers. Any significant divergence is reported to tesla an issue for evaluation. Teslas operating in autonomous mode are already 5 times safer than the average driver. I urge every reader to review the latest tesla autonomous drive video on their youtube channel, watch Elon Musks talk at autonomy day and read the safety reports published on their website. This false narrative by the legacy automakers and their lobbyists is a last ditch attempt to sway public opinion before Tesla submits their self certification application to DoT NHTSA later this year
9
Sorry - Every week Tesla performs 2 million hours. I.e. 400,000 cars x 5 days x 1 hour commute = 2,000,000 hours of real world verification
4
For passenger vehicles that might go anywhere, yes. But for buses and long-haul trucks, there's nothing standing in the way.
3
Perhaps more realistic is to use the knowledge from developing driverless cars to incorporate significantly more advanced safety features into cars driven by people.
I will make one comment about a bias the almost everyone seems to have on this issue. If a driverless car is in a fatal accident, it is headline news.
But people are actually terrible drivers who make horrible decisions.
30,000 people per year die in car accidents. The lead story on every news broadcast almost every day should be that 100 people died today in.car accidents. We have just come to accept this as inevitable, but almost all of those deaths were preventable and happened because a person made a mistake.
6
They should develop self-driving cars that can only be used on freeways and highways, where there are no bicyclists or pedestrians.
It would be a great way to travel long distances between cities.
4
What I wonder about is how humans,
especially older humans whose reflexes are not so fast,
are expected to react to driverless cars that may know
the optimum braking/acceleration/changing of lanes
to make, but the humans are too slow to react in kind.
Furthermore, what happens when a Used Driverless Car
is for sale. Who is going to certify that all the sensors/computers
are still working in harmony ?
How much will these cars depreciate when new models of
far greater sophistication are sold.
Why not use driverless cars in cities where they stay in the
second lane until the passenger needs to exit an can only
change to the inner lane when no one is behind them and
they can never go one the freeway or over 30 mph.
Meanwhile, let them make deliveries after midnight to dawn.
See how the driverless cars do then, improve them and then
give them a new task learn.
4
AI has been oversold since its origins in the 1960s. For example, fifty years later we are still waiting for ”natural language translation”. “Siri pass the butter” is merely task competence, a much lower bar.
7
The fast track would be to create an environment where all cars on the road are self-driving. To operate these cars in places where pedestrians, bicycles, and other driver operated vehicles are permitted.
1
What about the future of self-driving flying cars?
As a sage once noted the future is always ahead of us because tomorrow ' is the undiscovered country'.
2
If we could have semiautonomous cars and trucks that will cut traffic fatalities and injuries by a factor of ten, the effort will be well worth it. And it should be made mandatory.
2
Frankly I am amazed and disappointed by the NYT for greenlighting an article on its front page that is so transparently written from the point of view of the prevailing techological approach favored by legacy automakers and their suppliers. Even worse, the article's headline starts with the preordained conclusion: it is a long way off and maybe unlikely: "Way in the future...". The author does mention Tesla late in the piece, stating that Elon Musk uses a different approach ("a chip it designed...") and predicts that it will be possible to reach level 5 (fully self driving vehicles) by the end of next year. But of course "many 'experts' are very skeptical..." without citing who those "experts" might be or on what they base their conclusions. The legacy manufacturers have no choice but to use a lidar laser based system as they do not possess or have access to the enormous data required for the AI/machine learning camera based system gathered by Tesla's existing fleet of over 500,000 vehicles equipped with the same camera system has collected over 2.5 billion miles of data and counting foundational to support a machine learning solution. The existing legacy folks are 'slow-walking' because they have no choice. Their very existence is at stake. As one of your readers commented, this is a FUD piece. I had to look that up. It stands for Fear, Uncertainty and Doubt.
6
Not really a direct response to the article, but we ought to change the parlance surrounding autonomous cars. Well, not really autonomous - because that's the problem. Self-driving cars aren't "autonomous," because there's nothing to have autonomy over. They don't truly make any decisions. Every action taken by such a car is pre-determined by the programmers and engineers that made it. These cars aren't autonomous, they're unsupervised.
4
Duh. Imagine someone having an accident because they were gawking at the car next to them with no driver in it.
BTW- what do autonomous cars do when confronted with an accident?
4
Nothing says High-Tech than the fluorescent lights from the 90’s in the lead picture.
While some of the problems are indeed difficult to solve, it should be pretty clear that Tesla’s approach is fundamentally different than the other players. A year might be fanciful, but a few years in places with more rigid rules of the road will be plenty.
I would love to see any autonomous car try to take on driving in Southeast Asia though.
2
As a long-time driver, I have witnessed a number of truly "once in a lifetime" stupid mistakes made by drivers on nearby stretches of road. Most of these involve errors made by two to three drivers simultaneously. Hard for me to believe that the level of "artificial intelligence" – actually, rule-based algorithms – currently available could possible cope with such problems.
6
Drivers, pedestrians, bicyclists and other users of streets and highways do stupid and unexpected things? And you have to expect this and be ready to compensate? Really? If this was a revelation to the geniuses developing autonomous vehicles, then I don’t want their cars anywhere near the roads I travel.
8
I swear Americans are optimistic fools.
7
Will only work with dedicated highways.
6
Thank God!
1
Gen X, Gen Y, Millenials. Should the first generation of kids who don't have to face a driving test due to autonomous vehicles be called the Autonomies ?
2
Good.
3
Another item being shoved down our throats. I like driving...
1
Thank heavens! "Autonomous vehicles" are the worst idea currently on the public's front burner, although "Medicare for All" ranks a close second. Why not "Autonomous Unicorns for All"?
1
As someone who loves to drive, I can't conceive of any reason as to why I'd want to use a "driverless" car. Put aside the fact that I'd assume all liability for any horrendous accident caused by the car not noticing a pedestrian or choosing to hit a bicyclist vs a bus full of school children. Sorry, but, I'm much happier in assuming the responsibility of driving and in embracing the joy of the open road.
4
Maybe 2 or 3 years ago I saw an interview with the president of Honda. As to the question of autonomous cars, he thought that they were at least 5 years away. Moments later I read in the paper about autonomous cars being tested in Philadelphia! What a contrast, the brick and mortar guys, 100+ year old industry, slow to innovate; let me restate that, forced to innovate. Then you have the tech generation, many of whom (Google) want to make the world a better place to live, though technology. For them, it can't happen fast enough. And it does. And it will happen with transport, because those techies are behind it, and impatient.
Self driving cars will, like all AI, change a lot in society. AI won't be able to make the kind of judgements we make, but will have its own set of values when it comes to making decisions.
The other thing is the likelihood that people simply won't own cars. And why should they? Cars sit parked most of their lives, and depreciate in value while they do. Since technological advances are happening at lightning speed, why not just take advantage of the fact and maximize the car's use?
A car could be working 24/7 picking up and dropping off passengers, getting its own fuel, having its own tires changed and engine and brakes serviced. To have a share in a car or cars would be far better value than just owning one that sits in the street or garage 23 hours of the day. It'd run 100,000 miles in a year, last three years, and maximize its use as a new, better, safer and more economical model takes its place.
I forsee ownership in a "level" of car like a rental. Uber, but the owners don't drive it unless they need it. And then, it will come from a pool that's nearby. Cars could shuttle anywhere like this: in a closely packed convoy on long distances, completely removing the need for trains. And they'll take you door to door, and you won't have to park them. Won't even need to pick up a car at the airport rental.
They'll be driven, cleaned, serviced, repaired, parked and refueled automatically.
We simply won't need to own cars at all.
2
@Jack Lee The other thing to bear in mind, of course, is that since people won't own cars, there will be far fewer on the roads, since they'll be shared. A single car could be used by ten people in a day, eliminating nine cars, and nine parking spaces.
Chances are, work will be staggered as people take advantage of working from home and flexible hours as machines do more and more manual and repetitive work.
The need for travel will diminish as AI helps with work, 5G makes massive improvements in speed of computer programming and communications. And, of course, we'll have very good simulations of live meetings in 3D and VR.
This is another example of where the valley bubble precludes useful innovations. Instead of leaping ahead to driverless cars, these companies can save drivers time money and lives by building AI assistance for drivers. Things like speed and grade alerts, braking help, relaying information about sudden stops to drivers behind through a mesh network.
There’s plenty of gains to be had via intelligent systems if they stop long enough to ask drivers what they would want.
3
Data. That's the goal, the gold, the glory. Few articles talk about the data of your life needed to drive these multi-tech vehicles. A driver, like the car, is just a data representation captured for police, insurance, local advertisers, health firms, registries of sex offenders and animal abusers - and more. Tech firms exist only with our data, or was it for the "betterment of society" that Waymo began their project?
1
I’ve been in the software industry for 30 years. Based on my experience on software reliability I wouldn't go near a self driving car and I wouldn’t allow my family on one either,
9
@BM: Back in the 1950s, Robert Heinlein wrote a book called "Door into Summer" - it was supposedly written in the then-near future, and featured a robotic drafting tool ("Drafting Dan") and a robotic maid. He was widely panned for the computerized drafting tool - people thought it would be hundreds of years away. However, they thought the robotic maid would exist by, say, 1970.
CAD/CAM software was available by the early 1960s - but we still don't have robot maids - just robot vacuum cleaners.
So, too, with driverless cars - it's more complicated than it looks. Much though I'd like a self-driving car, I just don't think we've got the sensors to make it cost-effective.
When we get robot cars on the Moon or Mars, capable of navigating without human input, I'll be more impressed
4
There are significant problems to address, and some that might never be satisfactorily addressed, that stand in the way of having mass driverless cars.
1. Liability. Who's at fault when such a car has an accident?
2. Security. Computer-drive cars that are connected to information networks can be hacked.
3. Maintenance/vandalism. Driverless cars will rely on numerous sensors and cameras to operate. It will be easy to put a car out of action, or compromise its safe operation, by damaging a sensor or sensors.
4. Ethics. Programs will be written that include ethics decisions coded into the car's computer. People need to understand that equations will be programmed into a driverless car that decides whether to do A or B in crisis situations when both actions might result in the loss of life. Major ethical debates need to occur; this is part of the liability issue.
5. Jobs and the economy. Driverless vehicles will put many who drive for their jobs out of work. What do they do for work? This will bring a change to the economy.
The technology might be almost ready, but society hasn't dealt with these issues yet. There's been almost no general public discussion and debate about these matters. Driverless cars are already on the road without the public having considered and consented to the policies, rules, and regulations needed to address these issues.
The necessary social debates and decision-making processes have yet to truly begin.
We need to look more before we leap.
7
Self driving cars in rural Maine in the winter will be difficult if not impossible. It makes sense in areas with well defined streets and roads in warm climates with high population densities. One problem is 4G broadband is not available everywhere in rural areas. The density of antennas for 5G needed for self driving cars will be expensive. The population density does not have the critical mass to justify the expense. The snow and dirt roads entering winding roads with rock walls and lots of trees will make for some challenging artificial intelligence programs. Where is the road? Who has the right of way? What? No internet signal? Was that a bush moving in the wind? A person? A deer? A moose? Snowmobile crossing the road? Maine will not have self driving cars for long time.
6
I posted the information about Ford and VW months ago here on the NY Times online. As a self driving skeptic, it was easy to spot this news buried in other stories that were circulating in various publications. It stuck me as big news back in the spring and strange that major news outlets didn't immediately pick up on the story. People, including smart editors and reporters, are so busy covering the news that a lot of news gets missed. Plus, the tech news community REALLY WANTED TO BELIEVE, really wanted, begged, self driving to come along.
When the push for self driving began to hit years ago, I wondered what it was really about. Back then, six or seven or more years ago, I figured it was not truly about self driving cars but driver assisted technologies and that the rush was because no one wanted to be left out of the billions that would be spent on new technologies. If cars are going to be jammed with sophisticated microchips (even more than now), who would want to be left standing by the roadside without a ticket to the party?
Self driving, if it were to become widely available, would involve spending trillions of dollars in the U.S. alone. There has to be some really big benefit and safety has not yet been demonstrated, nor can it be until billions of miles are put on the odometers.
One final question: Why is this story from July 17 being re-run now? Is this a test to see if anyone is paying attention? Well, if it is, I get an "A", right?
4
Based on the studies I’ve seen from industry experts, it won’t be until 2030 to 2032 when autonomous vehicles start seriously disrupting the marketplace.
At last, reality disillusions people dreaming about self driving cars. Good that they are disappointed before a lot of lives are destroyed. Automated systems are not self aware. They are tools not cyber life forms. They have no stake in reality.
The challenge now is how to integrate automated systems with human operators so that they work effectively and reliably, together. That will take a long time to work through but the potential for better driving and fewer accidents is there.
5
My reaction to this article is that there might be an unanticipated benefit in this even if autonomous vehicles never materialize - that trying to teach a computer to drive safely identifies the difficulties and subtleties in safe driving (some of which are described in this article) that could make human drivers more aware and therefore perhaps safer.
1
@Paul
Well considered questions and sound testing most definitely can help. But people who drive a lot under all conditions have seen very strange, unexpected events. Mostly it's people but sometimes its the machines which produce it. Humans who drive a lot tend to accumulate a sense of what is normal and what is not that alerts them to be careful and avoid problems themselves. It's not something that an algorithm can replicate, it's more of an executive function of a conscious intelligence, a mind.
3
@Casual Observer
I agree with you. I realize that experienced drivers have the sense you describe. I was just thinking that the subtleties are probably nowhere explicitly taught to new drivers - and that their identification and articulation by the automation effort might be of benefit.
2
I’m really not looking foreword to the day when autonomous cars are common. What will it be like when five or six of them are following each other? Will you be able to safely pass? Will they be allowed in the fast lane and if so, will they drive at the speed limit no matter what? The idea of people sitting in a car but perhaps not paying attention to the road also creeps me out.
6
I was amused watching a Tesla trying to park in a simple spot, no car in front or behind, no moving traffic. It took six back-and-forth before the driver called it quits - and the final parking was over 30” from the curb. Fogetaboutit for this an other reasons; only when EVERY vehicle on the road is autonomous, with no motorbikes or bicycles, will ‘mostly-autonomous-driving’ be feasible.
3
It sounds like I won't be able to count on self-driving cars to let me stay in my home if I can no longer drive. Get going developers! Baby boomers are plentiful and would love to know this technology will be there.
Once again, Ford has blundered.Has the current CEO done ANYTHING right? Bill Ford Jr. should go as well. How much has Ford invested and lost on this scheme? For the investor's sake, focus on products that work. The Ford Flex is an excellent vehicle but Ford is off chasing stuff that will never work! The company has had an inferiority comp;ex since the Pinto disaster. It needs to move on.
My reaction to this article is that there might be an unanticipated benefit in this even if autonomous vehicles never materialize - that trying to teach a computer to drive safely identifies the difficulties and subtleties in safe driving (some of which are described in this article) that could make human drivers more aware and therefore perhaps safer.
"driverless cars' should be the LAST implementation, following 1) driverless trains (no more engineer human errors going too fast) and 2) metro trolly lines -operating on a track removes many uncertainties. And finally, 3) City bus service -the routes are fixed.
Only THEN should regulators enable tests of driverless cars that can go anywhere.
The way it's being handled and presented by Tesla, Waymo, and all the rest is topsy-turvy upside down with the downside being more people will be killed. All for something we don't really need.
4
My computer on my desk, which I control, still crashes on its own. I don't trust riding in a computer on wheels without rails that I can't control with random humans doing random things around me. Sorry, scientists; end of story for me on this concept.
7
Very bad news for Uber, as the success of their business plan appears to depend on getting rid of human drivers.
8
The best brains in NASA cannot prevent tragic accidents on the launching pad and in space happen. Who would trust the car industry of "surfeited sponges of speculators" (Trollope) in driver's safety with a self-driving car?
6
What’s so bad about driving a car? I like driving!
7
Wait, what? So the things that tech gurus tell us will be so easy are actually more difficult? Unbelievable.
7
As a habitual pedestrian (at my own risk), this sounds like the classic, multibillion dollar solution searching for a problem. All those brainy R&D folks should be putting their money and time into getting people OUT of cars not in to them. Sure, lives will be saved by robo-cars but how many more might be saved by building a comprehensive urban and suburban mass transit systems that get cars off the road into the relative safety of trains and dedicated busy lanes. The big reason why the various auto companies are speeding ahead with automated vehicles is because they want to sell MORE cars to the gullible public not less.
4
There's a difference between bringing a television into someone's home and a self-driving car; one's innocuous, the other isn't. These "disrupters" receive such undeserved acclaim in this society, which only worships money and never asks how it was accrued. They are not visionaries, just extremely selfish humans being cheered for disrupting people's lives and livelihoods while they cash in and move on. Everyone, three cheers for the disrupters!!!
4
I said it before with the last article about driverless cars, unless someone is in the car paying attention they will not work. Driving esp. nowadays is all judgement calls. So for a handicapped person that can see, great. But otherwise unless they make special lanes everywhere .. forget it.
2
@Doctor Woo*** I would like to add ... what if someone hacks the system. Or it fails in someway. You will have major multiple accidents. The Times did an article recently about cities paying off people hacking into the town computer systems that control just about everything. They had no choice. So it could happen .. Also another comment I read said this as well. The car manufacturers are just now figuring out driving is judgement calls. That is pretty amazing & depressing.
1
No thanks to driverless cars. Elon Musk can go first with brain contraption as well. Count me out.
2
Self driving cars remind me of another "wave of the future" broken promise...affordable high speed internet throughout the nation. Millions of people would be working remotely. This would increase productivity and reduce pollution by people spending less time traveling.
We just forgot to factor in middle management trying to justify their jobs because "someone needs to monitor the workforce."
I will start believing that radical change is just around the corner when I start seeing some radical change.
5
There’s an issue here not being addressed.
Whether self driving, taxis, uber, privately owned, cars create congestion & pollution; eventually they are destroyed. Autos require large amounts of metals & plastics. Factor in the longer distances people travel to work, a growing population, construction of roads, bridges, and tunnels.
What's the point? We can tinker with the 'parts' but the entire system does not work. Sorry.
Efficient local mass transit and long distance train travel on a necessary scale are given no priority. Nor is quality, efficient multiple dwelling. These ideas are periodically discussed, examined, proposed, rejected.
Somewhere during the 70’s or 80’s America became complacent. The oil & construction industries, thru lobbying, made sure members of congress did not encourage mass transit or multiple housing at the scale required.
When a green solution is proposed, Congress members from oil states shriek ‘subsidy’ as tho the oil industry has not been subsidized for over a century thru every means including war.
The average resident, by age 60, does not want ‘change’. They have the house and the yard and the car. Understood. But we need to help young Americans begin a life which does not destroy the environment.
Perhaps it will require term limits for legislators, a vastly reduced role of lobbys and the sense of urgency which built the interstate hi way system.
There are many ideas worthy of discussion. Now
4
The car manufacturers are barking up the wrong tree in trying to make cars able to drive themselves. There are simply too many inputs (speed, location, and size of surrounding vehicles/structures/people), each changing at every increment in time.
The thing about automated cars is that EVERY car on the road will have to be connected to a centralized system so that ever car can know where every other car is in space. Like airplanes or trains. How scary would it be to have every airplane driving itself without connected to air traffic control? So all cars will have to be automated or it will not work.
6
Some have said driverless cars would cause more people to make more single-rider trips. Wow. I guess people still think that if somebody else, or a computer, drives the car, it doesn't burn up just as much carbon or take up just as many road-miles as if the individual drove the car.
I see the future of autonomous vehicles as lying in the field of public transportation. Bus and train drivers are very expensive. Even if we don't lay off current drivers, driverless vehicles in dedicated, separated public transit lanes could help with route expansion and frequency of service thereby making mass transit more appealing.
1
Since we won't have self-driving cars, can we please stop loading our vehicles with technology that distracts us humans from keeping ourselves, our passengers, and other folks alive and safe and healthy?
It's proven that humans need our brains to drive, and that devoting half our brains to hands-free phone conversations is just as dangerous as talking with the phone in our hands.
Talking with others in the car, while it can at times be problematic, is much different. The most obvious example is that most passengers automatically shut up (or verbally warn the driver) when the situation becomes hazardous.
Every bit of equipment in our vehicles should prioritize safety, and I'm consistently amazed how little this seems to be taken into account with switches, buttons, dashboard lighting, etc. Wouldn't it be much safer if all vehicles put controls in more or less the same locations, and if these controls worked more or less the same? There are plenty of other ways to stand out from competitors than placing the hazard switch, for example, in some new and innovative and consequently unfindable site.
5
I can think of a couple of other benefits not mentioned; the reduced cost of insurance and not having to take the keys away from Grandpa and Grandma.
1
If cars emitted beacons, as aircraft do, self driving cars would work. This is a case where government has to intervene for the safety of people and the growth of an industry.
It’s a very interesting technological development but a number of factors need to be looked into. Introducing driverless cars in America and other parts of the developed world is entirely different from introducing them in all parts of the world. The factors are
1) Driving in towns is completely different from driving in big cities like New York when it comes to America. Everyone follows traffic rules strictly in towns whereas driving in big cities is entirely chaotic including honking. Even pedestrians most of the time don’t follow the traffic signal while crossing the roads.
2) Condition of the roads in cities and in towns should be taken into consideration.
3) In the event of forced shutdown or what you call forced outage on account of natural calamities or severe power failure traffic signals won’t work.
4) There is plenty of greenery especially in towns. As such there is every possibility of encountering animals such as deer etc even on highways.
5) Traffic Police might suddenly resort to one way or might even close the street on account of unexpected eventualities.
6) Drunk driving or reckless driving.
7) Covering blind spots.
Many more factors such as violation of traffic rules completely, worst roads, narrow roads, encountering animals on all sorts of roads all the time etc should be considered if the driverless vehicles are to be marketed throughout the world.
1
I just don't see how this is going to be possible unless and until all cars on the road are automated. You will never be able to program a machine to anticipate the full range of human behaviors. Indeed, many of us revel in the thought of being unpredictable... especially when driving a car.
I'm putting money on this never happening unless the government makes it happen.
1
Forget autonomous cars. How about autonomous shopping carts that could be used in shopping centers? At a typical shopping center, customers wheel store carts out to their cars, unload them, and usually leave them anywhere they can because there is no "cart corral" within easy walking distance. The carts could be equipped with a simple programmable sensor that would take them back to the store they came from. There would be no need for people from the store to periodically go out to retrieve them. And some of the issues of autonomy could be resolved on a small scale within the confines of the center. The carts would have to deal with "unpredictable" human behavior in a semi-controlled environment. And I, for one, would find such carts incredibly convenient. And I'm sure lots of other people would find them convenient as well. Why don't we start small and convenient and work our way up from there?
@Ruth Ann - Go right ahead and make it happen. Afterwards, your next challenge will be how to get money from those who benefit. I'm guessing they will display advertising... that still doesn't sell.
I hope the young engineers and software people doing this work actually have drivers' licenses and have driven before, many do not.
And the unpredictability of drivers has become worse not better--I just drove cross country (bad enough behavior) only to be almost taken out by a driver on cell phone in my rural neighborhood.
People on highways are horrible--cruise-controlling in left lane making it impossible to pass trucks, when they aren't glued to back of truck out of driver's sight.
The stupidity of people out there in incredible, and non-drivers programming software will make that worse.
At least TX puts up signs "Left lane for passing only".
1
My guess is someone doesn't want this to happen. Cui bono?
1
10 years before autonomous driving ubiquity, ten more before cowardly politicians will allow passengers to drink,smoke wee,etc while riding.
There goes my portfolio strategy!
1
Humans themselves can't always anticipate what other humans will do. People are unpredictable - you know, free will and all that - many of them are little crazy, and there's a sizable minority who are really really crazy. So why should humans expect computers to accurately anticipate what humans will do?
After all, who writes the code for those computers?
2
The paradox of autonomous driving vehicles is that the technology would work if all vehicle driving were automated and would far more efficient and safe. If only we could remove the humans!
I could have told your reporter and those company execs this very thing 5 years ago.
The best opportunity to use this technology is to drive semis from a point 50 miles outside a city to another point 50 miles outside another city. Harbor pilot takes it the last 50 miles.
2
Can't tell you how disappointed we all are to hear that.
Hopefully the self-tying shoes and self-playing guitars are still on track.
7
Has anyone asked the question why do we need self-driving cars? The expense and liability would seem to be more expensive than the current system. Not to mention the vulnerability to hacking, etc.
4
The REAL money will be in cleaning services for shared self driving cars.
Imagine all the sick things people are gonna do in these things!
There once was a man from Racine
Who invented a self driving machine...
4
So many scams and yet we keep falling for them. I hear Lyle Lanley has a great idea. I think it's called a monorail. I'm pretty sure he sold them to Brockway, Ogdenville, and North Haverbrook, and, by gum, it put them on the map!
2
Here is why self driving is coming and coming soon. My guess is that within 10 years drivers will not be able to take their car onto an interstate highway unless it is driving itself. The technology is already there for cars merging on/off and staying in line. With all the cars communicating with one another, pacing and following at safe distances it will improve driving times and greatly increase capacity.
What is more practical, mandating the technology in all cars or building more lanes?
2
SNOW. Sleet. Ice storm. I have never seen how these technologies can handle these natural occurrences.
5
Perhaps we might put in the hard work required to arrange our lives such that the AUTO is not such a damning cynosure of our existence. You may say I'm a dreamer...
6
How to solve the micro-maneuver issue is by banning human driver on the street. By any meausre human drive far worse than robocar so there is no reason why human should remain on the road endangering safe robo drivers.
1
@AmateurHistorian
Not me, I seem to be the only competent driver left, cross country.
But all the idiots I encountered would say the same thing despite being wrong.
1
All these geeks and companies are gonna SO “disrupted” when I finish my teleportation machine this summer!
5
@trblmkr, do you remember the movie, The Fly, with Vincent Price? The scientist in that dematerialized, transported, rematerialized himself from a chamber that unfortunately he shared with a fly. It seems the process switched their heads on their bodies. Better test it first real good before going public IPO.
1
Elona Musk is on the case, not to worry.
As are the Chinese.
1
Thank you NYT for a lead article that is NOT Trump! Aaah relief! Let’s follow Frank Bruni’s lead.
4
Even the Enterprise needs a pilot. When the Enterprise was piloted by a computer, that computer was deemed a person (AKA Data).
On my side street there is a house with a driveway.
The exit from the driveway is perpendicular to the road.
From the street the driver can not see into the driveway.
Children sometimes unexpectedly drive their bikes onto the road.
I go by the house very slowly and often cross a double line to be as far as possible from the driveway.
A computer driven car would not allow crossing that double line.
@david
There is the assumption that AI is "programmed" with "fixed" rules. Not so! AI controlled cars will make an informed decision whether to cross those double lines based on a variety of data and dependent on that data. More importantly, AI may be able to "see" around those corners and either warn the entering traffic or avoid it or both.
In my view it will take autonomous vehicles (AVs) decades to develop effectively and be accepted by all drivers- a mix on the same road will create havoc. In the northeast and many other areas I have driven, traffic often moves at speeds greatly above the legal limit, which AV's will have to obey, unless the driver takes control of his car out of frustration of being passed on all sides. Since typical car lifespan is roughly 15 years, it will take that time to reduce non-AVs to half their present number, provided their replacement with AVs is strictly enforced. It will take another 15 years to reduce their number to a quarter.... and so on. However a quicker acceptance may be possible if states have the equivalent of HOV lanes for AVs with higher speed limits than non-AVs in the remaining lanes (preferably separated by barriers). As technology improves and AVs become ubiquitous their usage can be expanded to a greater varieties of roads. To be really effective all AVs from multiple manufacturers must be able to constantly feed their locations and destinations into a general traffic control center, which in turn adjusts the speeds of all cars on a given road to maximize traffic flow and anticipated possible bottlenecks. I don't think this level of standardization and control is not even on the drawing boards. In the meantime, the fastest way from A to B is in a manually driven car.
2
@Rudy Ludeke
Agree, and moreso--a manual shift car ensures attention and I had always driven them and are critical in mountains.
People driving themselves are dumb enough, I also foresee difficulty melding them with rigidedly controlled cars.
So... why exactly do self-driving cars need to operate flawlessly? Man-bites-dog headlines make it sound like we're in for a future of Russian Roulette on the streets, but self-driving cars are already far (far) safer than any driver in the long run. They don't have to be perfect to improve on what's out there on the highways today.
Insurance rates will eventually reflect this. Maybe not a big deal to geezers like me who have logged a half a million miles, but imagine a 21-year-old male looking at his insurance bill -- likely nearly as much as his monthly car payment. And his bill is that high for a good reason.
I love to drive, but I these things can't get here soon enough. If driving were outlawed tomorrow and everyone issued a self-driving car, we'd have maybe a hundred fatalities a year instead of EVERY DAY.
3
@Skip
So you are will to buy me a self-driving car?
The transition will be difficult.
Driving an automobile safely is part the modern social contract, frail as the skill might be. We will need the improvisational freedom of movement for as long as we have an economy that runs on the efforts of independent individuals.
I think this AI technology would be more useful if it was put into trains and expanded rail/thoroughfare systems, with the capacity to take care of individual vehicles entering and exiting the system.
One person's logical decision can appear incredibly illogical to another person, especially when it's hardwired into a potentially lethal machine.
Hey, I've got a computer that can do ALL of those things they're having trouble with, all in a split second. It can easily tell if there's a bicyclist riding the wrong way between cars. It has no problem dealing with street sweepers. And in a million years, it will NEVER mistake a white truck for the sky.
It's the computer that sits atop my shoulders, the human brain. It's the product of millions of years of evolution. In concert with its network of senses, it's a positive genius in seeing immediately what's going on, and in figuring out the best way to respond. The brains that weren't good at that never survived to the present day.
Why are the tech giants and the car companies trying to, well, reinvent the wheel?
Yes, the brains get tired and distracted and are the servants, to an alarming degree, of an emotional host. But why can't the companies focus on developing things that AID the human user, rather than replacing him or her? Things such as what they are already doing: braking and lane drift assistance?
All the people whose livelihoods depend on driving would greatly appreciate that, but they wouldn't appreciate being replaced. Please -- let's focus on technology that improves our lives, rather than makes us irrelevant.
6
@Cynthia
I agree with your overall premise and will not buy a self driving car. I even hate the lane containment alert on my wife's new car (although I appreciate the blind spot warning feature). I will happily keep my responsive 17 year old BMW and savor the joy and versatility of its quick shifting manual transmission.
The human brain is indeed a marvelous computer, but unfortunately it can easily be rendered ineffective by alcohol and substance abuse- a major contributor to traffic accidents and fatalities. Technology could reduce their frequency, but complacency by government, manufacturers and consumers has thus far proven to be ineffective in implementing such methods..
1
Who needs driverless cars ? Let's have more mass transportation !
4
Mass transit cannot work in the suburbs or rural areas. Walking distances would be too far.
Self-driving cars is a technology that will benefit many people, particularly seniors. These problems will be solved.
1
Kinda like similar utopians over the past few centuries: "It woulda worked except for those darn humans!"
2
We live in a society where some idiots think it’s a laugh riot to lick a tub of ice cream at the store and then put it back in the freezer.
Watch what happens when these people start “messing with” autonomous vehicles.
3
australia. dusk. 180lbs of suicidal red kangaroo coming out of the bush at your headlights at 30mph.
yeah, good luck with that.
6
@Thaddeus
And road trains!
1
An autonomous vehicle would be able to react much quicker than even an above average human. And using more than the visible light part of the spectrum could probably see the roo coming before 20/20 vision ever could.
it would be best, at least a near-term goal, on the highway for long trips. Highways are already far more regulated than secondary and tertiary roads. No pedestrians, formulaic junctions, limited grades, regulated curves, no stop lights. We already feel comfortable with cruise control on highways. I’d love to be able to relax, close my eyes, look off to the side, or get distracted on a 5+ hour drive. Its nearly mindless just requires constant attention. It puts people to sleep! Automation would probably reduce fatalities on highways.
1
@Aaron
Where do you live???
Try driving out west and maneuvering through truck traffic and mountains and the idiots on cruise control in the left lane, preventing one from passing trucks.
In my experience the software will be the real problem. All new software has at least one bug and when the bug is found and corrected you have a piece of new software and it continues.
Let’s focus on getting the electric car normalized then tackle driverless cars.
2
AI is already here and deeply invested by Tesla and BMW, despite what this commercial from Ford/VW would have us believe. The one who wins top of the AI mountain is the one that learns when to alert the driver to make a life or death illegal move, so tech support can come on line and solve live problems.
Man made sensors can never replace the one we were conceived with. The legal and functional success of AI lies with lies, alert sensors, waivers of liability, live tech support and the realization that we are surrendering our common sense to deliberately finite batteries for profit.
Seems like such a waste of energy and money. Fix poverty and the environment and I will get excited. That would be real progress.
3
Has anyone even asked consumers if they want this? I actually like driving my car, and while I might get annoyed being stuck in a traffic jam, I don't want to turn over my life to software code written by a millennial day laborer.
6
When I lived up north there were occasions of blinding snow which stuck to everything. Will the cameras see through the lens which is covered with snow or the tropical rain storms with limited visibility we have in Florida? The testing to my knowledge has not been conducted in adverse conditions. How about a snow storm in downtown Minneapolis with pedestrians slogging through the snow?
1
Self-driving cars were a way for the automakers to continue promoting privately-owned vehicles.
God willing, we’ll actually invest in public transportation now. It’s so much more reasonable and cost efficient. Not to mention global warming...
4
A decade off? Someone better tell Elon Musk who for years has been saying it is only 18 months away.
2
The real answer is that resource depletion means individual passenger vehicles are not in our future, with drivers or without. But people want to point at the emperors's shiny new paint job instead of the elephant in the showroom.
2
"Despite High Hopes, Self-Driving Cars Are ‘Way in the Future’"
I would go even further: "Despite High Hopes, Modern Automobiles Have Over-Promised and Under-Delivered"
The modern automobile is now an electronic device on wheels. In an era where texting and driving is illegal, somehow car manufacturers are allowed to make cars that require interacting with the vehicle in an inherently dangerous way. How and why does the government even allow these vehicles to be outfitted with interactive touch-screen displays? It is beyond absurdity. The majority of "modern technology" needs to be eliminated from vehicles. Need evidence?
"Backup cameras haven’t stopped drivers from backing into stuff"
https://www.latimes.com/business/la-fi-hy-back-up-cameras-20160617-snap-story.html
"Distracted At The Wheel: Car Infotainment Systems"
https://powerlegalgroup.com/car-infotainment-systems-distracted-driving/
Forget "self-driving" cars...go back to just making an Automobile: A car (or Automobile) is a wheeled motor vehicle used for transportation.
4
@John
I hate all that garbage too, and all the noise alerts the cars I rent for cross country trips.
And last was a GMC (I kept calling it an AMC!) that in lieu of real transmission or paddle shifters on steering wheel (good) had NINE "low" gears on a push button on far right dash down below--I had to take eyes off twisting mountain roads and reach over and down to down shift.
Insanely stupid, but so is idea of driving an automatic when you are dropping 7,000 ft off mountain.
Guess I should burn out their brakes?
2
Self-driving cars will work great. When all cars are self-driving.
It really is an all or nothing proposition.
I could see a city going 100% driverless in the future, but not a whole nation. Especially a car-loving nation like the US.
Semi trucks could go driverless, but only if special lanes for them were added to every highway. They'd still need a human driver when something unexpected happened, or when they get close to their destination, and have to navigate in a city.
1
It’s this sort of pessimistic babble that’s going to leave us receding in the road-dust of the Chinese...
10 years ago, incumbent US car-makers saw autonomous driving the way their predecessors saw auto safety and quality, a half-century before that...
They refused to internalize it, tried to talk it off the radar-screen of consumers, and hoped it would just go away...
Having devolved from being founded by visionary innovators to being operated by myopic tactical capitalists, they now only respond in the face of two existential threats:
1. Nation-scale litigation
2. Global-scale competition
As with the earlier wars of the 20th century, the US operated from an advantage of scale – once it began to pay attention...
And so – for a time – we prevailed...
Not only as a producer of cars – but as a premium market...
Hence, all the European and Japanese car-making plants built in the US over the past several decades...
This time around, the advantage of scale is gone – but the capitalism is as myopic and tactical as ever...
10 years ago, the computer chips and image/other sensors needed for autonomous driving were less than 1% as capable(vs power consumption or cost) as they are today...
Further, the autonomous-car systems HW architecture is fully in place...
It’s like the personal computers before Windows and Office...
The economic debacle of ceding autonomous driving leadership to China is going to make the 5G debacle look like a simple flat tire, by comparison...
3
self driving sensor/AI tech can save more energy/time if used for smart traffic lights
1
This article completely misses what "US companies" are doing in China (see my long post).
GM has been self-driving cars there since 2016
In "joint" endevours with Chinese companies so that they can learn the ins and outs
Why would GM do this ? First of all, it sold more vehicles last year in China than in the US, so it is really a Chinese compay
Second, since the PRC government is 100% committed to beat the US in driverless cars , you can run over pedestrians without legal consequences.
The NY Times, in the age of globalization, needs to be more aware of what "US companies" are doing abroad.
1
They are just now figuring this out? It was obvious from even a cursory familiarity with the world of self driving cars that many basic aspects of their function had not yet been addressed. Not only not addressed, but not even frequently discussed. I don't understand how people who think about this full time could have missed that until now.
1
Well, duh.
1
Robotizing the roads seems to me to be a failed concept as currently conceived and promoted. That's what trains and trams are for. How about better driver education, to start with?
3
@WR
Like not cruise-controlling in left lane just next to truck (worst place as drive or steer wheels blow, you are dead meat).
Which prevents other people from passing truck.
TX signs: Left lane for passing only.
I hope they program that in.
Maybe now we can instead set our sights and venture capital money on green energy and zero emission vehicles. No one was pleading to have a self-driving car (except those like Salesky). In fact, many of us still enjoy driving and the independence it brings. What many are asking for though is a more sustainable way of life that reduces greenhouse gases.
3
Is anyone really surprised by this? When the press started writing about self-driving cars, I was saying there's no way this is happening in my lifetime.
1
As I have mentioned I am a legally blind person and no longer have given up my driver's license voluntarily. I am the last person you would want driving a car though I have threatened it eliciting a big groan from my wife.
The thought of having a driverless car car do the driving fascinates me. But then I remember how our GPS told us to drive off a bridge mid span and all the times this computer has left me in the lurch and I think better of it. Then there is the problem of a license. Would I need one? Would any state in their right mind give me one Well there are states that will give me a gun license . . .perhaps there is hope for yet.
1
All steering mechanisms in autonomous vehicles are electric. So, what happens when a vehicle loses power? No steering. It stops. Blocks traffic until a tow truck can move it. The safety regulations (ISO 26262) are specific and detailed for vehicle operations. Investors would do well to read the fine print. No autonomous vehicles will run in the US without a safety driver before 2024 or 2025 except on closed roads. Separate roads for autonomous vehicles may be a requirement.
"Ford and other companies say the industry overestimated the arrival of autonomous vehicles," aka, "Ford and other companies lied to investors."
1
Public Transportation everywhere. Put money to good use for all! Environmentally correct--shrinking the automobile footprint(well said M.K. Bernard) is where we need to be! Look at Prague--inhabitants can get anywhere using public transportation! Living on the outskirts of Phoenix (and having lived in Prague), it is so frustrating not to have transportation accessible. As a senior citizen, driverless cars are most appealing, but not as appealing as public transportation!
@Kristine
I like to drive, I like to camp, public transport doesn't cut it.
AI autonomous vehicles are fatally flawed because every vehicle is independent. Automated trains are systems with built-in feedback. If each car were connected to a network of other vehicles, many of the collision avoidance issues would be resolved. Furthermore, if traffic signals were in sync with every vehicle, all of the cars could accelerate/decelerate at the same time reducing the delay from starting and stopping at traffic lights. This has the potential to reduce time in traffic.
Cadillac does have a self driving model, that works on about 14,000 miles of divided super highways built to modern safety standards. There are no kids chasing balls into the streets, or stray dogs, as shoulders are part of the design.
Automated Vehicles could be reasonable on such a subset of road, but inherently, we we have narrow single lane streets with trees right up to the curb, and maybe a hill or twist, with the vagaries of sleet and storms. This is when an experienced driver knows that the car should be driven much under the official speed limit, with extra vigilance.
Road Signage is determined by the Federal Transportation agency, with deference to states, so regulations are sometimes allowed to be overruled. Some signage is political, such as Stop signs installed by neighborhood pressure, when it is perfectly safe to drive slowly and vigilantly through the intersection.
Modern planes can be controlled by autopilot, which work except when they don't, and only an experienced pilot can complete the flight. The same principle holds with street vehicles.
1
@Call Me Al
Yes, and driving VA to CA (3X now) one must pass trucks, which depending on state are 15-20 mph less than cars.
I try to explain to my back East family and they do not understand.
If I did not pass 100s of trucks daily, not sure but far slower and perhaps backwards???
Show me AI that can pass a truck in mountains one way each way, even in flatter deserts--there are too many people now messing it up now.
Driving skills have deteriorated significantly.
Every day on my way to work I ride my bicycle down a typical street where I live which is one and a half cars wide, has cars parked at random on either side, and where cars, pedestrians and bicycles all operate within inches of each other at all times.
This is not some edge case. Not just Japan but all over the world, it's daily life for probably more millions of people than live in the entire US.
driverless car sensor/computer technology has potential for big energy/time/safety benefits if used for smart traffic lights
1
Re: slower and costlier than they thought
In the US, YES. In China, NO
From 2017 (Two years ago)
1. "Shanghai allows autonomous tests"
"Matt Tsien, president of General Motors China, said, "Shanghai is in a great position to become a leader to embed connectivity technology into a comprehensive traffic management system, which will help increase transportation capacity within dense urban settings."
The United States-headquartered company and other automakers have been testing their intelligent and connected vehicles in Shanghai International Automobile City over the past 400 days ...
http://www.chinadaily.com.cn/business/motoring/2017-11/13/content_34469664.htm
From the South China Morning Post, 2019
2. "China formulates new policies for autonomous cars in bid to catch up to US
The latest sign of national regulatory support comes two months after Beijing became the country’s first city to green light open road test for self driving cars"
https://www.scmp.com/tech/start-ups/article/2132591/china-formulates-new-policies-autonomous-cars-bid-catch-us
3. From China Daily, 2019
"Pushing pedal to the metal for more self-driving cars on roads"
"General Motors is one of the first carmakers to develop the V2X technology in China.
http://img2.chinadaily.com.cn/images/201901/07/5c32b8fba310686029dec8fa.jpeg
Why would a US company bailed out by the US tax payer helps China to develop driverless cars in China while closing Lordstown?
Lower wages, and NO lawyers
1
Aristotle understood this problem better than anyone discussed in this article. He distinguished between motion and action. Motion is what physical objects do. Driving a car off the edge of the Grand Canyon can easily be accomplished by artificial intelligence. By contrast, action is what organisms, like people, do. Action is what achieves goals, like driving through heavy traffic to get to work. Artificial intelligence (AI) is good at simulating motion – it can control moving objects perfectly well. The problem is that it isn’t very good at representing the immense variety of goals (and their relationships) that a task such as driving to work must achieve. AI is faced with the problem of both understanding and improving upon the nature of the secret sauce that allows people to accomplish an immense variety of goals (and their interrelationships) in everyday life, however imperfectly. Since the goal of AI is to provide a safer mode of transportation than that provided by mere people, it must invent a superior sauce to that which animates people. I believe that AI is, and must of necessity be, too intellectually impoverished to do so. If I was invested in any company purporting to provide so-called “self driving” cars, now or in the future, I would sell immediately.
1
@JGB
Yes people do incredibly dumb stuff while driving, an it is not predictable to even an experienced driver like me.
I track everyone around me and note the unpredictable ones and am alert to them, they still do worse things than I predicted.
I don't know if AI cannot be trained in unpredictable stupidity, probably they thing it can, but I doubt it.
Our roads are in terrible shape.
What rural municipality can keep its lines painted in the north?
Who is trained to repair all these cameras and sensors?
Computer hacking happens.
And let's not forget design. Those roof top bubbles just look stupid.
Self driving cars are not coming.
1
‘Way in the Future’
Thank God
1
What this comes down to is the dog-vs-toddler scenario.
You are driving on the interstate when you come across road construction. Everyone is funneled into one lane with concrete barriers enforcing the temporary rule. There you are in one lane, cruising at maybe 45-50 mph. Suddenly from the right, with no explanation how they got there, out in front of your car comes a dog being chased by a toddler. Your dilemma: you can hit both or you can hit one of them. You choose the dog, without any conscious contemplation.
A computer is incapable of making this kind of decision. It would impartially figure the better of the less of two evils to only hit one, but it would be a fifty/fifty proposition whether it chooses to hit the dog or the kid. Self driving enthusiasts would argue that the computer could distinguish to a choice between the bipedal or the quadrupedal and then choose the latter, but processors are still not fast enough nor impartially capable in the fast time frame provided of weighing all contingent variables.
Self-driving will work only in designated highways, just like we (wisely) confine unsteerable locomotives and their following loads to train tracks.
1
A typical human driver would probably be eating/texting /smoking and run down both.
I am amazed that people thought this was an easy problem to solve. First, they have to figure out how to automate driving in rain, fog and snow. If the lidars don’t work, the car isn’t going anywhere. Second, the automated cars need to be as safe as commercial airliners. Even though today’s autonomous cars would likely cause fewer deaths than human drivers, people would expect these products to be perfect. Car companies would face much higher liabilities per accident than when only humans are involved. It doesn’t make sense, but that’s human nature. For these two reasons, it may not be possible to create fully functioning autonomous vehicles without additional technological breakthroughs.
4
The whole AI industry has ignored at its peril Gödel's incompleteness theorems. Gödel was Einstein's best friend for more than 30 years and is in a close battle with Aristotle for greatest logician of all time. He proved in the 1930's that in any logical system complex enough to support the arithmetic of the whole numbers that there would always be statements which could be made in the system which were neither provable nor disprovable in the system. AI qualifies as a logical system capable of supporting simple arithmetic. Hence there will always be situations which arise that will leave the computer software dumbfounded. Gödel believed these situations could only be resolved, if at all, by human intuition.
3
"Anticipate," is not usually a word I'd associate with a robot. ;) These things are obviously very problematic. At the same time, there are an amazing array of people who should never, ever be behind the wheel of a car. Frankly it is amazing that most of them have licenses.
5
It seems to be a matter of greed with regard to the auto industry's obsession with autonomous cars. After the bailout cars doubled in price. A Buick that I bought new in 2005 cost $12,500. When I was T-boned in and the insurance company totaled it, the replacement in 2010 was $24,000 and a new Lacrosse now runs about $35,000. I don't want to be a passenger when I'm driving alone; it's boring enough when someone else is driving. Why do these companies insist on giving us something for which there is so little demand? And... don't get me started on the electric cars that we're supposed to charge in our garages... where their batteries could easily explode and burn our houses (with us inside) down. If I want to be driven, I'll take the bus.
1
@Mary Trimmer
"After the bailout cars doubled in price."
Demonstrably false nonsense.
1
After more than 100 years of marketing car-driving to people as the key to freedom, sex, power, love, and being "up-to-date," does anyone seriously think that the public will give all that up? It's laughable!
https://emcphd.wordpress.com
4
I guess that this means that driverless cars will happen sooner than expected. I'm amazed at the number of techno-skeptics that support any Times article that contains opinions opposing any new technology. Why is driverless technology so scary? I see the opposition more as crazed libertarians who think that their ability to flaunt speed limits and other safety rules will be infringed.
6
Am I the only one who wants a continental high-speed rail system? Start it in the middle of "flyover country" in Tulsa, OK or St. Louis, MO, underdeveloped or struggling cities that could use the work, and build outward from there. The US is a vast, sprawling, nation with a disparately concentrated population, and could make use of updated rail.
I'd gladly pay more in federal taxes for that stuff. I never understood the appeal of self-driving (re: "hackable") cars...
8
Let’s begin with electric cars. Elon Musk is still working on a cost-effective battery. The US government should provide additional incentives to this industry. Don’t let China grab this market. Autonomous transportation could begin with mass transit, much like what is being tested in the freight trucking industry. Once these sectors are mastered, it can move to the mass consumer market.
2
There's a significant problem with software for any product or "solution".
It's only as good as the humans writing it.
I have had countless meetings and discussions with the AI and machine learning sect in SV and SF and to a person they refer to the transfer and elimination of multi-sector jobs by robots, AI, et al as utterly necessary. And they are not that concerned about the fall out.
Often some of these brainiacs like to characterize people as plebes and they merely want to replace or more accurately discard them.
I've been at this for a long time and what I see consistently are some people that are book smart yet willfully ignorant of how their concepts will impact society.
So what do I think of ANV cars? I don't see why I would utilize one and as for mass deployment, there's still so much work to do.
And what about software in general? There's a simple albeit longstanding tenet that comes to mind.
Garbage in, garbage out.
3
Amen.
Silicon Valley is good at solving problems with technical solutions that cause bigger, unanticipated problems.
Just implant microchips in all humans to proscribe their behavior around autonomous vehicles.
It's really not the big of a leap. Many people have already revealed (sold?) their souls to social media companies.
I think driverless cars will be a reality only when ALL the vehicles are driverless. That will eliminate most, but probably still not all, of the variables that human drivers introduce.
@Lynda What about pedestrians and cyclists?
I can't be the first to smirk that fusion-powered self-driving cars are a match made in Neverland. Wake me in 30.
1
I've thought all along that most drivers would prefer a car that serves more as a co-pilot and that does things like alerting you to road hazards and keeping you from driving straight into solid objects. Once onboard car computers demonstrate that they can be good co-pilots, then we can look into what else they might do.
2
Its a matter of data. The industry thinks it has huge amounts of data, but it actually has a tiny data base. To get thedata that humans have, developers would need 250 million autonomous test cars on the road. Each test car would have to look at ever facet of life in the US, as anything can affect driving decisions.
That may not do much good, because these vehicles are not adept at picking out important information, as humans do, so they would be flooded with trivial facts. Perhaps AI could learn, but its not a given.
With growing resistance to more data gathering, developers could not put 250 million autonomous test cars out there, looking at everything.
Bottom line: Autonomous vehicle developers are against an impossible problem. It wont be a decade, it will be never.
2
I am so happy to hear this.
2
We're wasting a lot of time on this (and unbelievable money) with partially deranged and acting-out futurists like Elon Musk et al.
Much less glamorous to work on fuel efficiency, battery technology, and alternatives to fouling and polluting petroleum.
4
Maybe not autonomous cars, but smart cars for sure. Warnings to break, automatic breaking, collision warnings. They are here or arriving. Accidents will fall in the next five years, completely changing the current personal injury-insurance-court services market.
I'm still trying to figure out what trillion dollar problem autonomous cars actually solve. I for one wouldn't buy one, if I had a choice -- I like driving!
And the idea of being a tender of one of the current semi-autonomous vehicle systems, being ready to take over when it gets into trouble, sounds like the definition of tedium and boredom. No thanks.
So that leaves automated taxis and trucks, I suppose. Putting more blue collar workers out of work is probably a great idea -- if you subscribe to the Silicon Valley mantra of "move fast, break things". And the thing you don't mind breaking is society itself.
3
@Abraham It's solving the problem of paying humans to drive. Corporations would love nothing better than to eliminate labor costs. Why anyone else thinks this project is a good idea is beyond me.
4
@AGinn
I for one would love to have a selfdriving car for mundane drives like the daily commute. Especially after a night shift or such.
Moreover, a lot of elderly people with degenerative disease (Parkinson’s and such) who live in the suburbs or many smaller cities have to choose between driving (and endangering themselves or others) or not even be able to go to the store (let alone social outings).
1
Driving is a sport to people. So are flying and boating. The whole idea of taking the pleasure out of these things is ridiculous.
7
@Steve Bolger
...says the guy who lives in Manhattan. Your tune would change if you had a 90 minute commute, or were a long-haul truck driver.
1
The American driving public, does not want self-driving cars.
People in Europe, people in Asia, people in Africa and Australia and South America, DO NOT WANT self-driving cars.
NO ONE wants this product.
And yet, the auto makers keep insisting on trying to jam this product down our throats.
The world wants FUEL EFFICIENCY.
We want cars that can go 50, 55 miles per gallon.
We want electric cars that can travel 500 miles on a single charge, and then be recharged in about ten minutes.
I'm talking to you, Elon Musk.)
We want safe cars, that will get us from point A to point B without anyone getting hurt or killed.
We want reliable cars, that won't be breaking down every other week.
THIS, is what the world's consumers want!
The auto makers know that this is what the world's consumers want, but they don't care.
Instead of spending all of their time and money on what they know that we want, they keep insisting on spending all of their research budgets on driver-less cars.
Well, if I have to, I can still walk.
Thank God.
6
Currently we are trying to have a machine drive a car to replace a human. I think we need to make the goal smaller. We need roads and rules for those roads that are also designed for those driverless cars. Start with a small subset, like freeways. I would even take out of consideration roads with ice and snow. Computer Science has always solved problems by divide and conquer. First let's solve using driverless cars on freeways that are certified for driverless cars. Certified roads would even have a protocol for driverless cars interacting with road workers. The workers would also need place readable markers that the driverless car will easily interpret, etc.
2
A different perspective on self-driving:
My maternal grandmother recounted often one of the best memories of her new england childhood: on winter sundays, the extended family would pile into a horse drawn sleigh to ride a few towns away, where they’d take part in a huge family dinner. The cousins would play, the parents would imbibe, (perhaps a bit too much), and under a clear, unpolluted night sky, the family would huddle under one giant wool blanket for the ride home. She would lay there with her whole family sleeping around her, staring at the winter stars, listening to the sleigh bells quietly ringing—and nothing but these bells— as their horse brought her home, because, of course, it knew the way.
I think we could all do well to pause and reflect: what exactly are we trying to improve upon?
8
@Andrew Frederick this is perfect, and I've read about this before -- the doctor in Iowa who would be out on a late-night call and fall asleep on the way home, and his horse would get him there.
2
If all cars and trucks become driverless then driving jobs will become extinct. Car insurance will fall dramatically and many jobs will be lost in the car insurance profession. Freedoms will be restricted even more for all driverless cars will be under control. You will have the right to travel but if necessary this might even change if the government sees fit to curtail it Currently all internet searches by individuals are copied and saved as they are copied. Artificial intelligence may deem what searches we look for are appropriate or not. A few years back Cisco wanted to install chips in each one of us for tracking purposes. Next we will find out that our thought process will be altered with chips. To the benefit of whom?
4
That is too bad. It would be such a great help to the elderly and disabled.
5
@Friendly You realize that Uber and Lyft are a thing, right?
2
@Friendly
Who will help the disabled person in and out of the AV?
The auto industry struggling with autonomous vehicles? Good. It's time to think about how many people who will lose their jobs with no likelihood of comparable employment before profits. Quit while you're behind.
5
@Jay Orchard Exactly what I was thinking. 30,000 fatalities equals 30,000 funerals. Imagine how many undertakers and grave diggers would have to learn new skills.
The creation of Self-Driving cars is still not ideal to be used in public transport as it is fairly new to everyone. The technology and ai computers inside this self-driving cars are still in the process of developing and any new circumstance it detects will be completely new to it that could lead to accident and tragedies. The self-driving cars are still far away from the future as Tesla the leader of a technology revolution still hasn't fixed every mistake and flaws of this type of cars. The public along with car dealers should wait for self-driving car technology to improve and acquire more information to be 100% sure that it's safe to be used.
All it would take to kill self-driving cars forever, is one vehicle running over a group of school kids. People can understand when a human screws up, but they can’t understand a faulty algorithm.
Can’t believe so much money is being wasted on this boondoggle. I will never buy a vehicle from one of these companies.
5
DEDICATED CORRIDORS
For the foreseeable future, I see dedicated corridors for self-driving cars for long-distance travel, not in congested urban areas. You'll never solve the people problem, as it will be irresistible for some pedestrians to play around with these contraptions to make them do silly things.
Some advantages to this: just like a plane or train, except there's no schedule. You can leave any time you want. Second, these dedicated roads will be up to 7 times more efficient than our current system: no need to keep a 7-car distance at highway speeds. So just one lane of these new roads will carry the load of, say, 5 of the lanes we currently use. Think: not as expensive to build as you might assume.
Why is NOBODY working on this? No dazzle?
Once again, AI overreaches.
2
@fedup....I like this, but then am reminded that so much of what we have come to value in our culture is getting where we want to go when we want. We are spoiled. The corridor idea is great for commuting, etc., especially if my home and my work are along corridors. I would think the evolution of this driverless tech would eventually incorporate the corridors. But for now, you know what many people think of public transport....not for me!
Progress keeps trying to take a big step backwards. In this town and area, there used to be streetcars and interurban trams until the cars came along, of course. Investing in safe and fast public transportation, rail, maglev, or whatever should be the plan. But of course, unlike most other countries, not here.
3
We got to the moon in less time than it will take to get me to the grocery store in a self-driving car.
1
"The technology is available now to create a car that won’t hit anything. But such a car would constantly slam on the brakes. 'If the car is overly cautious, this becomes a nuisance,' said Huei Peng"
Here is another thing I think has not yet been realized, most of us would not drive in an autonomous car that drives like they do. We would actually be uncomfortable as our own passenger. When we have control, we take risks that we would not trust a machine to do. So we need a car that takes more risks than my first paragraph, but less risk than we do. It is a tricky balance.
2
AI has been vastly oversold in a number of applications, not just autonomous vehicles ... this may cause you to notice that overselling, overpromising, overhyping has been a trait of the technology sector in the past two decades. not just in computer technology, but in the technology of big pharma "breakthrough drugs," carbon capture climate solutions, 5G networks and dozens of other applications as well. technology is the new century's junk food.
1
The fear of automation shouldn't be for professional drivers, no matter what the the Ponzi scheme called Uber says.
Automation will be most disruptive against a highly entrenched, unionized work force - Government.
What is easier, more efficient, and popular - automating a 18 wheeler going through a school zone, or a infinite line at the DMV?
1
AI? Nowhere near. What we have are a collection of statistical algorithms running against big data at scale. There is nothing intelligent about it. These algorithms hail from the 1960s - there's nothing revolutionary except we can run them faster against bigger data sets. We've had no great insights into "intelligence" or "consciousness" or artificial anything.
4
@George Campbell
Agreed. By definition, a machine cannot “think.” It can only do what it is programmed to do, no matter how much thought goes into the programming.
1
However perfect driverless cars try to be, human behavior shall remain unpredictable, hence, a danger to him/herself, although a bit less frequent, as anything robotic will obey the rules (laws) of transit (i.e. stopping at stop signs; unable to cheat in 'speeding', for the few minutes it may gain or making it on time when starting late). But it may yet prove invaluable in public transport not amenable to rail.
Nobody is asking why we need autonomous vehicles. What is the problem it is trying to fix that Uber, Lyft, taxis, and mass transit already address? I wish the industry would focus more of its resources on electric cars -- an environmental problem ripe for fixing, now.
5
@Heidi It's just a way for corporations to lay off anyone who drives for a living. Sounds like utopia, right?
1
The tech industry's ethos of "move fast and disrupt things" doesn't work so well with cars. Although the current true prohibitive barriers to self-driving cars are cost and regulators' unwillingness to give tech everything they want, rather than ethical or safety questions. Once the industry figures out how to cut costs and to get regulators and local governments to look the other way, we'll get to have self-driving cars everywhere.
2
Of course this is going to take 20 years. Everyone who actually build cars or has any sense would tell you that the practical realties of self driving cars is waay off.
Why don't they put these things on the Las Vegas strip so people don't have to walk up and down ten miles to get from one end to the other?
Why? Because it's all about control, not utility. They want to yank that steering wheel out of your hands. You can't handle freedom.
1
Writing algorithms to deal with the extraordinarily wide variety of idiotic driver behaviors has got to be a real challenge.
I don’t even like using cruise control because I feel I’m not in control of the car so I doubt I would ever get into a self driving car.
1
The tech fix is just a dream away. I recall in the 50s the power companies said that nuclear-generated electricity would be too cheap to meter.
3
Here's the thing that galls me. The shameless hucksters shilling for autonomous vehicles keep pushing how we have to do this to save human lives. Give me a break. This is about your corporation making money.
If saving lives was really corporate America's motivation and they really cared about the 35K or 40K Americans who get killed each year in motor vehicle accidents (a number that's only been increasing with Americans' deepening addiction to smartphones), they'd be pushing for greater investment in public transit and accelerating the shift to less auto-dependent development. But of course that's not their job.
Hey, here's a shocking idea. What's good for Elon Musk and Jeff Bezos and Jack Ma might not necessarily redound to the common good. Maybe it's time we stopped letting the corporate cart pull the public horse.
8
What’s Uber’s path to profitability now?
4
I wish these articles would more objectively contextualize the safety of these self-driving cars compared with human drivers.
I’m guessing that these self-driving cars have fewer accidents and deaths per-hour-driving than the typical human driver. But when people only hear of specific instances of the cars” auto driving failing, it can make it very difficult to accurately assess the benefits of self driving cars compared to their costs. No one wants to get hurt by a machine in error. But I’d say most people don’t want to get hurt by a human in error either. We should hear the specific cases of both or overall statistics of both.
Otherwise this could be like an article trying to report the issues of seat belts by only mentioning some specific accidents where people survived because they didn’t have seat belts on. Would that mean that overall we are safer without seat belts than with them?
2
What are these "ride-sharing" services you talk about? Who is sharing the ride? Please use words and phrases that actually have meaning.
2
oh yeah ??
Who would have thought.
1
The availability of short distance transportation (eg:Uber etc,) with immediate connection to and from centralized mass transit ("L", light rail etc) is what is needed along with convenient scheduling (the "L" comes every ten minutes as I recall during heavy traffic times & is predictable); I walk 10 blocks when I have the time not when I have to be at a meeting in five minutes!!
1
I live in Mountain View, Waymo's home, and have, for more than 6 years, seen their vehicles driving around my neighborhood. Sometimes I pass two or three Waymo vans in one short trip. While I can understand the company's desire to conduct testing near their headquarters, haven't these vehicles completely mastered the not-so-challenging task of driving around my small California town by now? My concern is they haven't, and that simple micromaneuvers - such as those referenced in the article - still elude these highly advanced cars. I really hope autonomous vehicles can be achieved - humans make a lot of mistakes driving - but I'm skeptical of the progress being made as I observe the technology first hand.
4
Soon, someone will also have to let Jeff Bezos know that we are not colonizing Mars or space anytime soon either. There seems to be an inverse relationship between the degree of hype and actual success with all futuristic technology. Meanwhile billions are diverted (and lost) from “planet saving” today’s technologies that are tried and true (high speed rail, solar panels on every roof, etc) while we speculate on job killing automation. I guess companies and investors cannot ignore “new shiny objects”. Billions went to Theranos for this same reason and billions will be showered on the next greatest advance as “visionary group think” continues to be more newsworthy than common sense.
5
@Eric Blair
High speed rail. You can do a few simple arithmetic calculations (preferably in your head) on the back of a napkin, and understand that high speed rail only works in specific situations -- where hundreds of thousands of passengers want to go from point A to point B, and multiple points are in a straight line. Not I repeat not in a hub and spoke configuaration. LA to SF was a reasonable example of what theoretically could be viable. What happened?
So though you are right on your other points, it's really important to educate oneself before describing something as "tried and true."
A car that will always obey the speed limit is a hazard, nothing more.
2
Thank God. I will never get into a self-driving car, so please put off this arrival as long as possible.
5
Good riddance. One big concern erased from our dangerous roads.
3
Still, driver-assist technologies, such as lane holding or parallel park functions, are welcome additions, and may help compensate for driver inattention. Or worsen it.
@Andrew
I largely agree with you, but drivers do need to know how to handle basics -- for example, parking. Today's morons who depend totally on GPS to get from A to B are an excellent example.
2
Here's a situation a driverless car will NEVER be able to solve:
California law says cars must not cross double yellow lines, must remain three feet from bicyclists, and must pull over when there are more than five cars following.
On smaller two-lane highways, it is impossible for a car to remain three feet from a bicyclist without crossing a double yellow line.
To accommodate these requirements, a driverless car would need to drive behind the bicyclist. But this might cause more than five cars to follow the slow (5 mph?) vehicle.
Which California law would the driverless car choose to violate? Would it illegally cross a double yellow line, illegally approach a bicyclist too closely, or illegally allow more than five cars to build up behind it?
AI can never "judge." This is the problem with any driverless car--all it can do is react based on preprogramming. How can you program it to break laws? and in what situations? Judgement is more than brake, accelerator, and steering control.
162
@GP
The oh so brilliant Tesla guy spent hundreds of millions of shareholder money buying robots to perform tasks of which they were incapable. Not only that, but anyone familiar with sixth grade arithmetic could have figured out that the cost of the robots far exceeded the alternative cost of having humans do these simple jobs.
Artificial intelligence... artificial, yes; inteliigent, rarely.
14
@GP
I don't think you grasp how AI programming works. The computer is constantly optimizing for thousands of variables relating to safety, routing, legal compliance, etc by training on massive datasets from a multitude of precise sensors from real world scenarios - more than any one person could experience in a lifetime. It is far more nuanced than cascading if/else statements.
It's going to take time, but these machines will be able to make better decisions than a human 1000 times over before you even have time to move your foot from the throttle to the brake. Plus they'll never get tired, get distracted, or induce dangerous situations due to a sense of pride.
34
@SB
I'll keep that in mind next time my Windows PC crashes and has to reboot.
21
I fly airliners for a living - the autonomous car industry needs to look to the aerospace industry which has decades more experience in the automation sector. To be reliable and safe autonomous cars will have to be highly regulated, including certification, operation and maintenance. Why is flying on an airliner so safe?-because every facet is highly regulated. Pilot are highly trained and undergo continuous assessment and evaluation. Certification of an airliner requires the manufacturer to comply with highly detailed design codes and requires extensive testing in all conceivable conditions. Airliners have to be maintained to a very high standard, preventative maintenance programs are extensive and prescribed by law, mechanics and engineers are highly trained and licensed, every single component is certified and tracked. Before every flight the aircraft is inspected and signed out by a team of maintenance engineers, and pilots and engineers refer to an extensive document called a dispatch deviation guide to assess what defects can be accepted and under what conditions. Yet still things go wrong, automation does stuff you don't expect it to or fails. In many ways the environment autonomous cars operate in is a lot more complicated and critical. Can you expect your backyard mechanic to work on your car, replace sensors with cheap parts bought off eBay, and then expect it to be reliable? No, it will have to be regulated, and that's expensive. 95% there, 95% to go.
211
@Lloyd
I too am a commercial pilot and a high-performance driving instructor. Trying to equate autonomous driving with airliners is completely off the mark.
An autonomous car traveling at legal speed limits can be stopped and pull over in ˜5 seconds. An airliner simply can never be slowed below ˜150 mph, no matter what the emergency is, without risking killing hundreds of people, unless it is brought over an airfield or suitable landing spot under sufficient control.
Cars can pull over, occupants can exit in seconds on an emergency. An airliner needs to be kept flying at 150MPH or higher regardless of the emergency. That is a huge difference.
An autonomous car is closer to an autonomous elevator or an autonomous tram, than an airliners. The former can be stopped any time there is an emergency. The latter can't.
11
Certification of an airliner requires extensive testing in all conceivable conditions? Flown any 737 Max8’s lately?
5
@Lloyd exactly why we need flying cars, just like the Jetsons
2
The sign says: "School Zone, speed limit 25 mph when children are present." Can software recognize when children are present? The traffic lights are out: Can software understand the hand signals of a traffic cop directing the flow of traffic? What if it's snowing? The concert is over and 10000 people are leaving the stadium parking lot: do all the driverless cars know when to move and when to yield? Which one goes first?
80
@David
Good luck. Those who think the whole concept is way oversold already agree with you, and the Tesla fanboys generally feel a few random deaths are a small price to pay for saving the future!
11
@asg21 Yes, all of these situations can be programmed into autonomous cars. We have the technology to recognize children, recognize and understand hand gestures, drive on snow, and coordinate with other vehicles. For every situation that we can anticipate, we can code or train software. The article focuses on the fact that every day there are surprises that are (by definition) NOT anticipated. It will take a long time to develop solutions for each of these to the point where we can cover 99.999% of the surprises.
13
@Frequent Flyer " For every situation that we can anticipate, we can code or train software."
I would not be so confident of that.
11
I've been in the robotics field for decades. Anyone who's been in robotics more than 5 years knew how difficult this problem is and how unrealistic the promises have been. There's even a joke about it: "Show me an autonomous car driving in the northeast during winter and then I'll believe you." The money people should try talking to people *outside* silicon valley more often!
179
@AB - With global warming there is no reason to worry about driving in snowstorms in the future. Now if you had said 'sandstorms', then I might be able to agree!
10
@AB: Thanks for speaking my mind and saving me the challenge of having to articulate the absurdity of pervasiveness of the tunnel vision in Silicon Valley. I do so wish those folks would break out of that bubble and spend the time investing, researching and spending human and fiscal capital on more meaningful endeavors. Great minds may differ on what those endeavors may be, but few disagree about the futility of the autonomous car boondoggle.
11
@AB
So true. So far, autonomous cars have only been tested in optimal weather conditions.
The problem, however, is the inconvenient fact that in non-perfect weather, sensors get covered with mud and snow.
4
I personally believe that automatic transportations will be tremendously developed as soon as google creates a real AI. Then every car can have the AI system that allows to anticipate what other vehicles will do each other. Al might be able to anticipate what pedestrians will do by tracking their phones. I’m excited to see this technology since this will absolutely reduce car accidents and save so many lives. Hopefully, AI is not bad guy in the future like a movie though.
17
@Tae
You go, guy!
@Tae What is a "Real AI"? The Google search engine is real, and it uses lots of AI. Google translate is real, and it uses lots of AI. But what you seem to be talking about is a Hollywood fiction. AI is a set of technologies for building smarter software. There is no single human-like "AI"; instead we have a wide variety of engineered systems that do various useful things.
8
@Tae There are 50,000 deer-car accidents in Michigan every year. AI WILL NOT BE ABLE TO AVOID THESE DEER! Even airlines have mechanical failures with all the required maintenance. How does AI fix or react to an unknowable or unexpected mechanical failure? I agree that driverless cars will be safer. We need to allow for the unexpected or unknowable.
2
If the bar for autonomous automobiles is zero human fatalities, a standard not required in any other industry, then I guess we will have to forever accept the current 30,000 human deaths a year.
The people who doubt the ability of automobile autonomy to dramatically reduce traffic fatalities and congestion generally don’t know much about computer science.
48
@Craig
Bill Gates knew a lot about computer science yet every OS developed by Microsoft that I've ever used since DOS to Windows 10 has "crashed" on my computers many times. Just saying that personal belief in the reliability of computer science does not necessarily make it so.
There's another way to look at the math in this debate. There are approximately 225 million licensed drivers in the U.S. Taking your statistic of 30,000 fatalities annually that equates to approximately 0.015% of the entire pool of human drivers. The number of human drivers causing these fatalities is actually much less then that because there's often multiple fatalities caused by a single driver.
When autonomous vehicles can guarantee that their systems can do significantly better then that statistic there's a place at the table to consider deployment but we're obviously not there. I for one do not want myself or my family to be collateral damage in an unproven experiment.
5
@Craigi I do know computer science this application is decades away in software, associated hardware and infrastructure. Anyone that believes other wise is simply deluding themself.
3
The profit-based concept of having cars driven by glitch-prone and hackable computers is very dangerous to society. All computers will fail at some point, and computers can never be programmed to react to the subtle behavior of other drivers, pedestrians, or cyclists in the way humans do in order to avoid accidents. Nor can computers be programmed to avoid road hazards that litter our roads and force constant adjustments in order to avoid vehicle damage. Under real-world conditions of ever-changing weather and road surfaces, computers will not be able to safely navigate vehicles at high combined closing speeds separated by only a few feet. And people will never accept having tech companies decide who will die. There are more productive ways to spend our money in order to enhance vehicle safety.
39
@RC The question is not, "can computers fail", but "do computers fail more often than humans?"
Imagine a technology that monitors the distance between a car and the one ahead of it. If that car slams on its brakes, the human behind it has to become aware of it and decide to do something about it, which could take a second or two. Having a computer that would, say, start braking in a microsecond in that event could save lives and injuries.
The "hackable" deal sounds like a doomsday scenario - say, some psycho who would otherwise be crashing a plane into a building decides to suddenly make all the cars in the US swerve to the center. I dunno - theoretically possible, I suppose, but seems unlikely.
Far more likely that I'll be trying to get something out of my teeth or daydreaming about a cute coworker and fail to notice everyone's jamming on the brakes, is my guess.
4
Even dish detergent fails to work as well as its makers claim. Why did anyone think that a self- driving car would be somehow different?
Any reading of the scholarship on AI would have shown that the sucessfull AI recognition of emergent circumstance really only works well when the parameters of the possible are sharply controlled. And the consequences of imperfection are dire and include your life.
I recall one of the few news articles that actually asked an AI scholar about the technology and they replied something like "If you knew what I knew about AI you wouldn't imagine getting into a driverless car. "
Stop listening to those who want to sell something and attract investment about how great the product will be.
And journalist, please check the claims of industry against independent scholarship
62
@Noah M
yes, there needs to be more questions asked (or hopefully, existing studies revealed) that go beyond industry press releases.
2
@johnwe'?
Excuse me but why do we need self-driving cars? Call me a Luddite but I just don't see the
need. There are more urgent problems facing humanity right now, why don't we take care of those problems first?
5
@2B or not 2B -- in addition to which how many jobs (and people doing those jobs) will be eliminated? ALL commercial drivers. (There are approximately 3.5 million professional truck drivers and ~1.4 Million Delivery drivers in the United States).
The real money will be in laying those drivers off and relying on the computer to drive the vehicle. Likewise farmers won't be needed to drive their tractors, get rid of 'em!
Likewise Taxi services (including Uber, Lyft, et.al). Any delivery service, (Amazon, UPS, FedEx, USPS, etc.) Multiple millions of people, suresure we'll retrain them to other jobs, NOT since this wave will take place quickly.
Better hope the 2nd amendment still applies to you cause we'll all need it. (actually, I'm a non-gun owner).
1
Lots of hype and unbridled enthusiasm. I see a future for autonomous vehicles but it is 10 years out except vehicles that operate with a limited area and scope. Remember google glass? It was going to be the new thing that did everything and everyone was going to have one. I’ve been in tech for years. Many things start this way, fail, and then come back 10 years later when mature technology and a new twist make possible.
70
If ever there was a path down which we should not be rushing, it's that of the self-driving car. Silicon valley, as well as traditional auto makers, have cleverly created a sense of need and inevitability among the public, with the ever compliant media cheerleading the shiny new technology.
But does this really correspond to what we need? A society starved of investment in public goods can better use a massive investment in non-fossil based public transit in all urban areas and a shrinking of the automobile foot print where possible. But what we will get instead is more driving and a turnover of the entire fleet; a supply-side driven roll out of AI self-driving cars because neither big tech nor the auto makers have any value proposition in a public-transit centred transport strategy.
320
@M K Bernard
Furthermore, we will NEVER be able to deal with the climate crisis if we continue to pour investment into automobiles and the type of urban sprawl they encourage. Replacing the entire fleet of petroleum burning cars and trucks with electric vehicles will require vast quantities of rare minerals for the batteries, and a huge increase in electrical generating capacity. If we were serious about lowering CO2 emissions we would launch a sort of Marshall Plan for energy conservation and rapidly "shrink the automobile footprint" through investment in alternative transportation.
38
"But what we will get instead is more driving and a turnover of the entire fleet; a supply-side driven roll out of AI self-driving cars because neither big tech nor the auto makers have any value proposition in a public-transit centred transport strategy."
And indeed, the Uber-Lyft rigged ("gig") cartel has triumphed (at humanity's expense) by overwhelming streets and regulators alike with their crowding vehicular presence. All of those are small fossil-burners DESIGNED to destroy not just the taxi industry (which has its own problems but is not so brazenly evil) but public transit's more efficient high-capacity buses, making Uber-Lyft and imitators a prime instigator of the climate attack.
Alas, media scrutiny magically vanishes around these bad megacorps, who've only had to wrap themselves in "tech" and carry an "app"; and repeated dives into the MTA cookie jar by de Blasio and Cuomo have added injury to injury.
Let's support a much larger, better managed public transit network, not rigged employers' stock-casino ambitions. It'll require, in part, an actual left wing.
12
@M K Bernard
Car accidents are a leading cause of death in the United States. The father of my daughter's friend was recently killed in one.
Yes-- we do need to rush to self-driving cars in so far as they reduce the likelihood of an accident. Regulators have been monitoring this closely.
I agree that there are concerns about an increased automobile footprint. And more public transit would be great.
But if you're dismissing AI for cars as a 'shiny new technology', then you're not understanding the human cost of the current tech....
2
Five years ago there was talk of Silicon Valley ruling the world, or at least the United States.
And this is what they've given us -- Trump?
1
What happens when a self driving car gets a flat tire? Does it just stop in the middle of the road, blocking traffic? Does it try to keep driving and damage the wheel further, or lose control and cause an accident? And what do the passengers all do in the meantime?
74
@Jonathan Brookes . Solid tires?
For decades cars have been able track things like tire-pressure. I’m sure self-driving cars are programmed to pull over in the nearest safe location when they sense their tires are malfunctioning. In the cases where self-driving cars couldn’t get off the road in time safely, it’s probably the same instances where human drivers couldn’t get off the road in time safely as well.
2
Autonomous vehicles cannot safely, regularly, reliably respond to human behavior on the road. If they could, they'd be sentient. Autonomous vehicles is a still born concept. Nothing but hype, and completely outside of science.
87
@John
Yes and it reminds me of the time travel hype.
Well, then, it looks like the economic disruption that driverless cars would cause to the auto insurance industry, the car dealer system and the automobile TV advertising business will be postponed.
Which means that the USA will continue to experience 35k+ annual deaths due to auto accidents and hundreds of thousands of injuries and maiming's that did not result in death while untold number of engineers dilly-dally around in their workplaces.
Get to work you engineers!
18
I’m surprised there has been no discussion of cyber attacks on legions of computers driving our cars for and among us.
Even being air gapped didn’t protect from Stuxnet. I’m pretty sure the autonomous cars are networked, though.
26
The article notes GM said they would have a driverless ride service in select cities 2019. The article notes GM has not taken any recent opportuniies to confim that schedule. The last I recall GM commenting on the deployment schedule was to say not only were they on schedule, but ahead of schedule.
While GM has not confirmed the 2019 deployment, halfway into 2019 neither has GM taken recent opportunities to say they would not meet the 2019 deployment goal.
GM reports quarterly earnings on August 1. I look forward to what they have to say. Or not say....
You can buy a pretty darn good driver assist package on a Caddy today, available on all Caddy models this coming model year, then other GM vehicles next year. It's a pretty good system. "Cruise". Keep you eyes on the road, stay on a highway, and you never have to touch the gas, steering, or brake. Hands free, feet free. Drive across the US, no hands, no feet. Just keep your eyes on the road (the system watches you) and stay on the highway. It's a pretty impressive system that most people don't know about. Why GM has not promoted it more is a mystery to me...
Finally - I think we'll see autonomous trucks early on, for deliveries that involve highway miles. The truck starts and stops at designated parking areas immediately adjacent to a highway. You can be sure truck fleet operators will be grabbing that as soon as it's available. Walmart, Amazon, FedEx, UPS, USPS - anyone who runs a lot of trucks....
5
Culturally, no man will give up driving. End of discussion. Technologically, can't close that final 10% gap. Second end of discussion.
21
Autonomous vehicles, EV vehicles, SUVs in all sizes; somehow GM, Ford, Fiat, et al. have forgotten one critical factor when promoting future high tech vehicles. What have the all forgot? How much all this technology will cost consumers over and above the cost of todays gas vehicles. In the case of the EVs there is the problem of decreasing range as the batteries age through recharge after recharge after recharge over the life of the vehicle.
Costs will be much higher, sales volume will be fewer vehicles, and vehicle replacements will shrink.
5
@Bayshore Progressive
Batteries must be easily replaceable. There are the pollution and materials availability issues though.
The Dave thing was said about the model-t; back then they were no paved roads and horses were more adept at mud.
1
The main value that I see from self-driving cars is to get more seniors off the road, providing them with desired mobility w/o driving past when they should. It will also allow more seniors to choose where to live w/o requiring nearby mass transit.
23
Until these things can consistently and reliably drive in a snowstorm, they will be of no use to those of us in the Upper Midwest (or Massachusetts, or Montana, etc.)
Also, think about what happens every time Apple releases an iOS update. Remember that time when you tried to type “I” and the little box with a question mark came out? No imagine an unknown software glitch affecting thousands of cars at the same time. So while a self driving car may be safer than an individual driver, the risk becomes more concentrated and leaves things vulnerable to a systemic failure rather than the distributed risk we have today’s
17
It will require changes throughout the infrastructure of ground transportation. Those changes can be made (e.g. segregated lanes and new rules) but our society will have to focus and invest. Unfortunately that’s not easy for us to do right now. Self-driving cars would really be nice for many reasons, but people will still want to drive sometimes. It’s a difficult set of challenges and the writer is correct about the process taking decades before these become a common sight.
9
"Despite High Hopes, Self-Driving Cars Are ‘Way in the Future’"...
As it should be. Let's fix the roads, bridges, and potholes first before we assign driverless cars to navigate them.
39
Have you seen how many accidents these vehicles have been in--mostly in Mountain View and San Francisco?
https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/autonomousveh_ol316+
There are not a lot of AVs, but an inordinate number of accidents. They cause accidents due to their unpredictable behavior. I highly doubt the ones that are supposedly not in AV mode. They must hire some bad drivers.
7
This is not a balanced article. There are 37,000 US auto fatalities per year. AI does not fall asleep, get old, drink, daydream, or text -- all major contributors to auto deaths. AI will only get better. We need to accept some AI fatalities so long as we have far fewer human driving fatalities.
28
@Dan - what is unbalanced about pointing out that it will be far more difficult than they hoped to have computer systems guess what people outside the car will do?
It's simply true. These vehicles will be running on the roads were currently have, and that means they'll have to deal with kids playing, pedestrians, bicyclists, people on scooters and all sorts of other contraptions - and I've not mentioned animals running into the road.
Researchers have tricked some of the vision systems, one example being using a couple of stickers to make a car sensor think a STOP sign was a 40mph speed limit sign. The stickers didn't even change the red of the sign, or the octagon. Just the word.
So yeah, eventually it will be autonomous cars doing the inevitable killing, and the toll will eventually be lower. That's a long way off, and I'll be interested in how the companies program the logic when faced with running into people no matter which direction the car steers - because that will happen. How will they handle what is an optimization problem? Hit the fewest people? Avoid babies?
But this is serious stuff, this is automating large, heavy, fast-moving things - not speeding out a new way to tweak your news feed. Haste makes waste, and waste in this case is lives, too. Haste because of impatient investors is absolutely the wrong approach, but that's what we see a lot of.
18
I personally enjoy driving most of the time. But to be fair my daily work commute is the opposite way of rush hour traffic as I live downtown and work outside of it. If I am in a car I would generally rather be driving it than a passenger. That all being said there are certain times it would be nice to have the car drive me. One obvious one would be for to have my car take me home, or somewhere else, after drinking. As it is now, I either walk or take Lyft. The other would be for it to offer something I can't currently do. It the left 2 lanes of the freeway were dedicated to high speed driving with the far left basically anautotrain all moving in unison at 100+ mph and the second from the left the merging lane into that autotrain where it does a quick diagnostic test to make sure you car can handle the high speeds, has enough fuel/charge to reach your detestation, and can communicate with the other cars so it all works. Once you vehicle past the test autopilot would take over from there and spit you out in to the merging lane when you are ready exit the autotrain. I would be okay with those trade offs. But for most driving; and especially driving around the neighborhood, city, and canyon roads, I am more than happy to do it myself.
5
Clearly this author has zero experience in any Tesla while it is under self-driving mode. Yes it is only for highways now, but many californians use it daily to commute with not only excellent results but frequent instances of the car "saving" the driver from a crash These are backed up by the videos that Tesla's dash cam takes. I have seen many. Not perfect, but the biggest issue seen to date are the morons working around the requirement of Tesla's to keep hands on the wheel. These dopes tie tennis balls etc to the wheel to fool the car into thinking someone is paying attention. Misuse doesn't equate to poor design.
13
@Debra
"frequent instances of the car "saving" the driver from a crash "...
Those are the types of technological changes, irrespective of a "fully autonomous" vehicle, that will make driving safer.
3
This article seems like a FUD piece by manufacturers that are behind the curve.
11
Duh!
2
Hey, you think self-driving cars, which have been hyped by the legacy media for years, don't work, how about electric cars? Complete disaster. Actual sales going from miniscule to less than miniscule, even in places like China where state bodies have to end up buying almost all of them. Check the Danish or Swedish papers: the governments pay people to buy them (free power, licenses, parking, tolls, etc.) and still, few people want the hassles of owning one. It's all the media trying to inject the general population with feel-good nonsense; until battery technology hugely improves, forget it. Not to mention, of course, they're either close to or are net lifetime pollution-positive: it's just that the Poles and others -- safely downwind-- end up having to live with the pollution. All of these issues quail before the main point, though, which is that a huge part of the general population, and those journalists who consider it part of their responsibility to tell that part of the population what to think, actually believe that these vehicles cause no pollutants to be emitted into the atmosphere! Once again, Mencken wins; that giant sucking sound isn't from south of the border after all (pace, Monsieur Perot) but instead comes from the vigorous thumb-sucking going on at this very minute all over our country ...
7
@BostonReader:
Cars with gas engines are a disaster. Too much to go wrong...and does. Think valves, injection systems, filters, mufflers, radiators, pistons, catalytic convertors, belts, transmission...and the list goes on. EV's don't have any of that. I drive my hybrid now but my next car will be fully electric. No more dirty smelly gas engine to contend with. Bye bye gas. Hello clean reliable electric.
16
Elon Musk, though a very smart guy, says lots of things that should be taken not with a grain of salt but with a 50lb bag of salt. I thought the SEC had put a muzzle on him to prevent his further attempts to manipulate Tesla stock, but I guess it isn’t working.
Further, while this particular piece deals with technological challenges but omits the mention of other factors including liability for accidents, potential regulatory obstacles, and last, but far from last, the need for the car to operate in all situations and circumstances without driver intervention because the operators of these cars will never maintain enough situational awareness to take over driving instantly.
13
And for the actual numbers from Tesla:
“In the 1st quarter, we registered one accident for every 2.87 million miles driven in which drivers had Autopilot engaged. For those driving without Autopilot, we registered one accident for every 1.76 million miles driven. By comparison, NHTSA’s most recent data shows that in the United States there is an automobile crash every 436,000 miles.”
Extremely easy to find...
12
@Andrew
Wouldn't a more relative comparison be:
How many accidents on a freeway or highway with autopilot not engaged vs those driving with autopilot? I say this because unless I am mistaken most accidents happen on city and neighborhood streets not highways and freeways. And Autopilot is designed for and primarily used on highways and freeway not city and neighborhood streets, correct?
10
@Still Waiting for a NBA Title
Honestly don't know. It is called the National Highway Traffic Safety Administration (NHTSA) and apparently it took a federal lawsuit to get them to release any numbers at all.
I would say Tesla is likely the only car company with the ability to easily track accidents.
PS: Knicks fans feel your pain.
2
@Andrew
Lies, damned lies and statistics...
What they're not telling you there is when the autopilot screws up you're meant to take control - so when the crash occurs it doesn't register as being on autopilot. Depending on whose figures you believe there is no evidence that Teslas are safer than average.
From https://drivemag.com/news/how-safe-are-tesla-cars-5-facts-and-myths-explained
"But when it comes to deaths, Tesla commented that it registered one fatality for every 320 million miles driven, compared with one for every 86 million miles driven for vehicles of all manufacturers. `Normal` cars would kill at least three times more people than Tesla cars; Autopilot included, the company concludes.
But Tesla's comparison included all types of vehicles, not just passenger cars. A more legitimate comparison would have been only between passenger cars, which, as The Christian Science Monitor noted, would mean only one death in 428 miles driven by humans. This would make a Tesla on Autopilot more dangerous as a passenger car that's driven by a human driver."
9
A deer bolted out of thick woods in front of my car on the road yesterday. If I hadn't been paying close attention to actually driving, I would have hit it, and possibly been killed. An autonomous car could never have detected that deer, because of the thick woods from which it ran. On Interstate highways, drivers turning on semi-automation and ignoring the road, now clog the passing lane and congest traffic.
This has got to be stopped.
Railroads should figure out how to carry much of what trucks now carry on the highways. Technology can make railroads much more efficient - precisely because a railroad is so much more controlled than a road. The Auto-Train between DC and Florida is a solution for cars - those cars are truly driving themselves to Florida.
The whole road infrastructure was designed with human drivers in mind, from the signs, the pavement markings, the varying types of roads for different levels of traffic from 14-lane urban freeways to rural dirt tracks, to the advertisements of business that have located along roads specifically to get human drivers to make a spur-of-the moment decision to stop and spend money there. Consider the business impact of this.
We want to be free to decide how to drive safely, and when to exit and pull over when hungry or fatigued, or for something simply because it strikes our curiosity. We are not cargo! We want to drive.
25
Haven't there been several hundred thousand automobile-related deaths on the nation's road in the last decade? And we have one death attributable to a autonomous vehicle? With the status quo so grotesque, I'm astonished so many people are opting for that option.
9
I would not be caught dead in a driverless car! Or, around it.
6
More to think about, courtesy of the IEEE "Cars That Think" newsletter that dropped into my email today. The first article it links mentions that people might have to pay more for cars that protect their privacy, otherwise, you are data to sell.
"“Car companies are now realizing that they’ve been sitting for years on troves of customer information that give them an edge on turning their product into a bundle of services,” Alvarez León says.
He points to a recent interview with Ford’s CEO Jim Hackett in which the executive said, “We know what people make. How do we know that? It’s because they borrow money from us. And when you ask somebody what they make, we know where they work, you know. We know if they’re married. We know how long they’ve lived in their house because these are all on the credit applications. We’ve never ever been challenged on how we use that. And that’s the leverage we got here with the data.”"
https://spectrum.ieee.org/cars-that-think/transportation/self-driving/surveillance-and-the-selfdriving-car
Think about the changes - people who love to drive? You are going to be cargo in a vehicle, making no decisions. Likely, you'll be sitting in the same lane, not passing or being passed.
In a residential area or urban locale or shopping area? You will be cargo in a vehicle that will be going slow to avoid harming pedestrians, pets, etc.
And your car may be beaming ads at you, unless you pay to turn them off.
6
The technology development of the autonomous vehicle is very far ahead of the needed development of new public policies related to this technology, needed upgrades and revisions to the highway and roadway infrastructure, and the insurance industry addressing issues related to the technology. This may not happen on any scale until after 2030.
7
No matter what you think about the complexities and potentialities of self driving, I come back to this question: what problem does it solve? Engineers, being engineers, are constantly trying to remove "the human factor" from systems they create. What if humans are inherently better at some tasks?
On the safety issue, we actually need to know how and why crashes occur, both statistically and individually. With dashcams and other sensors on cars, we are just now coming into the golden age of understanding crashes. When we have a full realization, then we can implement corrective measures. Until very recently, we have relied mainly on investigations of crashes and the stories about why they occur. For the first time in the 100+ yrs. of history of cars, we can now know for certain.
Has anyone stopped to think how much it would cost to change over the entire fleet of cars and trucks to self driving? It would take about 20 yrs. as cars age out of service and the cost would surely be at least $10,000/car, probably more. That works out to about 2.8 trillion dollars. To those rushing into this field, that means profits. For the rest of us, it means money out of our pockets. If we have that much to spend, is self driving the best way to spend it?
After long and careful consideration, I believe self driving will be widely implemented...in the next 25 yrs., but mainly for selective, limited uses, like a closed central business core of a city or on Interstate highways.
8
Throught the entire article, and in most everything I've read & heard on ALV's one topic is glaringly absent: insurance & liability. Who (or what entities) are liable when bad things happen involving ALV's (and in the mixed world of many ALV's along with human drivers, they will happen. Big time.) Have any of the car companies addressed this? Are they assuming liability for the autonomous driving system? Have the car insurance companiew weigned in? If nothing is done to address this it will just be breeding a shark-tank of lawyers waiting to sink their teeth into the issue when it hits the fan!
11
Please let this happen no sooner than one minute after I have died of a cause other than an automobile accident.
When they coined the name "automobile," they didn't mean it literally.
6
put them on a track and run them in a circle like rides in a kiddie park for arriving to point a departing from point b
1
Thank goodness.
2
If humans can reliably drive a motor vehicle (remember: 80% of the drivers think they are above average in skill level!) it is only a matter of time before fully autonomous transport is available.
Very few people encounter anywhere near the entire gamut of possible scenarios on the road, eventually ALVs will have a store of capability for this wide range. As human skills erode from lack of use, ALVs will dominate. After all, very few humans can reliably "drive" a horse anymore; soon essentially none will fly military assault planes anymore; the list goes on.
The ALV needs 3 things: inertial frame (where am I and what is my vector motion and mass); external sensors (vector objects and identities); and a prediction engine (how to accomplish the goal with all these moving (vector) objects, which includes predictions of the future and guessing at/peering into the unknowns. This last is the hard part. Humans frequently get it wrong, which contributes to so many crashes. ALVs will eventually be better, and won't text, and won't be mentally disabled by legal and illegal drugs, won't be distracted by other people in the vehicle, or the cell phone, the radio, the...
10 years is perhaps optimistic. Boeing can't even get MCAS certified after several months, and MCAS is simple by comparison.
1
Human and robot drivers of personal vehicles will never share roads. Full stop. Humans will not tolerate millions of (insurance liability driven) obsessive speed limit observing, intersection light clogging, robot granny drivers. A formula for rage and vandalism. I'd expect inner cities to zone all-robot districts, with personally driven cars required to be fully automated, human driving disallowed. Outside those zones, robots will be banned. By contrast, large rule-bound robot buses, tractor trailors and delivery trucks free of white pills, faked logs, tailgate crowding and illegal parking, will be welcomed by humans on shared roads.
6
@J111111
Bingo.
2
What consumer group or constituency actually wants driverless cars? Are we suddenly falling short of adults with drivers licenses? Wouldn't it be nice if new technology addressed something that society actually wants?
26
@Scott G - ah, but don't you care about the poor investors, and the techies who want to "disrupt" and get rich from it?
8
@Scott G
When I hit my late 80's and have my driver's license taken away (like my mom and in-laws had happen to them) I'm hoping a driverless car is available.
1
What you say about autonomous automobiles is correct. But it is a fair bet that within five years, our major cities will have autonomous flying taxis in commercial operation (Google skai).
While autonomous automobiles have to be able to anticipate an almost infinite variety of possible collisions, autonomous flying vehicles have no problem being in constant contact (location, speed, direction, flightplan, protocol, etc.) with every other aircraft of every type in the vicinity.
Moreover, hydrogen fuel cells eliminate virtually all the problems associated with heavy batteries.
1
@bmz
"Moreover, hydrogen fuel cells eliminate virtually all the problems associated with heavy batteries, replacing them with entirely new problems (storage, explosion containment, etc., etc.)." - FTFY
3
@asg21 hydrogen fuel cells have been safely used for decades in our space program as well as on the highways. It is actually safer than gasoline (a hydrogen leak simply rises and evaporates into the atmosphere).
2
The only way to make self driving cars viable and safe is to remove humans from the equation. In San Francisco I’ve seen bicyclists blow through red lights at maximum speed like no-one else is on the road and completely oblivious to the idea that their own safety is at risk. Unless we can make self driving bicycles talking to the self driving cars there will still be deaths on the road. Human stupidity will ensure it.
8
Much like "Artificial Intelligence", a pipe dream, and in the case of AI, a misnomer.
6
So 4 people have died in Tesla driverless cars since 2016. How many hundreds of people have died in that same period in human “controlled” cars? Which one is riskier?
4
@HV - humans drove over three trillion miles in the United States last year. How many miles of Autopilot operation went into slamming people's cars into stationary highway dividers? How safe is the technology today when a different driver in his Tesla was also being steered towards the divider that had already claimed one life? In a fully-autonomous car, he'd not have been able to steer out of it.
If our consumer protection agencies weren't in the hands of corporate interests right now, I'd be looking to penalties for Tesla calling a driver assist product "Autopilot". Or is false advertising a technological advance nowadays?
6
@HV
Probably a waste of time, but here goes: If 2 Teslas crash out of a total of 5,000 Teslas, is that better than 4 ICE cars out of a total of 5.000,000 crashing?
2
In other words admit you can’t get a car to think like a foodie snippy nosed yuppie and predict, anticipate you were just gonna cut off somebody because you want to be the first one to the upcoming red light. (But, spent boatloads of money researching that fact first).
2
@Capt. Pisquat
Right. You're describing actual behavior - should we just pull the drivers' licenses from everyone that does something similar? Do you ever exceed the speed limit? Looking forward to following autonomous vehicles that are programmed to never do the same?
2
If all or most cars are autonomous there will be no problem with erratic drivers
1
@Pat - and luckily there won't be any people, pets, other animals, etc. anywhere near all the roads.
Interesting conclusion some researchers arrived at when considering self-driving cars in urban environments - they'd be helplessly slow if pedestrians jaywalk frequently enough. The cars would stop. The passengers would sit. The walkers would safely cross the street (or walk down the middle of it if they chose).
There's a lot of shifting needed to flood our pavements with autonomous vehicles. Slowing them in most places is part of it.
2
We are seeing a repeat of what car companies did at the advent of the car culture in the US. First they promoted the product as something that we all needed, then they convinced the government to compel us to pay for what they told us we wanted. Back then, they told us we wanted cars, then convinced the government to force us to pay for highways. Now they tell us we want self-driving cars, then (because that cannot work with today's infrastructure) they will convince the government to force us to pay for the infrastructure needed to support their products. They will say this is all in the name of safety and convenience and economic growth. In the beginning and the end, we all just wanted a safe, easy, and cheap way to get from point A to point B. In many cases, the solution was and is public transit. In other cases, "human-driven" cars work just fine. But car companies intentionally dismantled public transit, and now they are trying intentionally to dismantle "human-driven" cars.
10
@george
"Back then, they told us we wanted cars, then convinced the government to force us to pay for highways."
Nominated for most delusional post of the day. Good one!
2
@asg21 Read the history of GM's lobbying efforts, my friend.
1
Thank goodness! With any luck, I'll be long dead by the time this becomes a dangerous reality, where the only concern is how much money will it make.
4
I've always been sceptical. How will these vehicles handle a gravel road or a snow/slush covered road in winter?
9
This article is far too facile and fails to recognize the biggest impediment to autonomous which is cost. The current additional cost for autonomous control systems is +$10K. Not only is that price far too expensive for mass adoption, the price is unlikely to drop much over the next 7 years or so.
A second issue with autonomous, which is being driven by electrification, is the additional weight of the systems. Combined with excessive battery pack weight, the range of autonomous vehicles cannot meet the ranges necessary for the US market. Furthermore, the energy draw from the computational control systems significantly limits an EV's range.
This is further complicated by the fact that autonomous is reliant on EV technology. Cobalt stores are an obvious challenge, but current nickle mining rates are also thoroughly insufficient to meet the demands of mass adoption of EVs. While it's possible to mine more nickle, the technology required to meet demand will take 5-7 years.
All said, there should be continued investment in these technologies.
3
@Chris - not sure where you get that autonomous depends on the vehicle being electric. Autonomous is sensors and control interfaces, not the power train.
But I'm curious about your confidence that control technologies aimed at a huge global market are somehow going to remain stuck at $10,000 for the next seven years. To quote you, that sounds kind of facile. There is intense R&D being done on all the sensors, software, networking. Similar to the investments pouring into electric vehicles (already a lower cost-of-ownership compared to ICE cars)
For example, newer lidar packages pricing under $1,000 compared to tens of thousands a few years ago.
https://news.itu.int/the-price-of-lidar-is-falling/
Note the article I linked also thinks that level 4 or 5 automation won't be spreading until the mid-2020s, not earlier. But not because of cost.
1
What certification process does a driverless (DL) need to pass?
What additional safety features must it have, features that a driven car does not need? AI is complex, and still in its infancy. Just ask Watson.
Health devices have to pass ISO standards. Imagine letting your surgeon use medical tech on you that is not compliant.
Until a DL has been certified safe by an independent agency (not the car maker), it should not be on the road. Period.
Remember the Boeing 737 Max. FAA off-loaded certification to Boeing.
What is the independent agency for DLs?
None exists.
We are talking about releasing into the wild a tech that is not safe. Because no entity can demonstrate it is safe. CEO bluster is not disinterested certification.
So go ahead, early adopters, push for DLs before they're ready.
Stack up the corpses. When a board member of the car maker becomes one of them, then maybe regulation will be the headline.
6
My husband bought a model 3 Tesla last fall. It has self-driving features for pulling out of the garage and parallel parking which is cool, and handy. However, every time I'm in the car with the auto-drive feature engaged, I am reminded of how far away we are from safe, self-driving cars. Throw in road construction with the lane-lines mixed up and the car cannot handle it. There are so many variables on the roads that it will be a long time before a computer can come close to matching humans when it comes to navigating roads. We're on the way though, maybe my kids will have self-driving cars in twenty years?
11
@LW Don't bet on it.
1
@LW In the 20 years, the time value of money will kill the investors ;-)
1
This article completely misses the point with the anecdote about the jaywalker fatality.
The only relevant statistic is the number of deaths/accidents human drivers cause per mile driven, vs the number of deaths/accidents computer drivers cause.
Currently we have 35,000 fatalities annually from human drivers. If computer drivers could reduce that to 5,000, we've just saved 30,000 lives.
It's very simple math.
4
You cite a handful of deaths in self-driving cars while ignoring the tremendous number of deaths from human errors. How many early airplane pilots were killed? Should that have halted aviation? Self-driving will improve over time as experience is gained.
It's not a question of whether self-driving cars will be perfect and never have accidents; that's impossible. The question is whether they will be more or less safe than human drivers. Considering the crazy things human drivers do like tailgating 5' behind a vehicle, texting while driving, looking in the back seat to talk to children, making left turns from the right lane, etc., I'm confident self-driving cars will be safer.
5
I am old enough to remember when I was a child they were telling us that Nuclear Fusion was just around the corner and it would usher in a new era. Still Waiting.
The arrogance of the tech industry in thinking just because they want something to happen that they can make it happen.
I live in San Francisco and I watch these folks testing the cars in my hood all the time. I have driven in front of them, behind them, and cross paths with them.
The AI just does not work. When following them the are like drunk drivers weaving down the road. They "sense" a long car in a diagonal parking as a hazard when it is between two compact cars. When driving opposite them, you wonder are they going to cross the center line?
On a left turn a woman getting who had just got out of her car and was walking along side it then to the curb caused one of these cars to stop with is tail hanging out into an intersection. An oncoming car almost hit it's stuck out tail when it stopped for no reason. There was no oncoming car in its direction of travel, it also would have cleared the woman just fine or it could have moved over a bit, over to the left while making its turn. But no, it just stopped.
Sure, there is a promise of less death and destruction if the technology works. But we are decades away from that. To the people who talk about all the death and destruction like drunk drivers, I say that is a non-argument. Let them take a Uber or a Taxi.
I for one will stick with my 1947 Desoto.
27
I’m really shocked at the tone of this piece and the negative comments. I’ve been driving a driver-assist Tesla for a year and right now am on a 700 mile road trip. I’ll admit it’s a little unnerving sometimes, but so was learning to drive. I call it “practicing”. In order to master anything new, even driving a gas-powered car, practice is required. Is the technology perfect? Definitely not. But the potential of this technology is staggering. As this technology becomes more widespread, cars will be “talking” to each other (no, we won’t need wired roads as someone here commented). Behind the wheel it becomes obvious that these cars are MUCH safer than human driven cars - this has already been proven! It’s an easy leap for me to envision fully self-driving cars based on my experience. There are countless performance benefits to this technology. When I drive ma gas car now, I feel much less safe. These cars react much more quickly than we can. If you’re on the fence or questioning, I urge you to try one out. I can almost guarantee you will favor the new tech.
6
@Christine Feinholz
"It’s an easy leap for me to envision fully self-driving cars based on my experience."
That's because you're ignoring the costs of replacing the current installed base.
4
I was recently in a head on collision caused by circumstances I do not believe a self driving car could have avoided. A construction crew was digging a trench on a street perpendicular to the street I and the other driver were travelling.
The street is a steep, narrow residential one with a blind curve created by a hill. The crew had parked a truck on the left side of the road (from the perspective of the parked truck) facing uphill. I was coming downhill. In front of the truck was a sign stating "Road Work Ahead", as well as cones to guide me around the truck. The placement of the cones and vehicle forced me into the middle of the road a bit. As I was in this part of the road, the other driver came around the blind curve accelerating up the hill. He had no way of knowing I would be where I was and I had no way of knowing he was coming due to the hill obstructing my line of sight of him. He was travelling maybe 25 mph and I was able to slam on my brakes while laying on the horn. So I was barely moving when we hit head on at the driver corners of the vehicles. So a low speed collision and no one was injured.
Neither one of us blamed the other for the the accident. We saw it as a construction crew mistake in traffic management. I wonder if both vehicles being driverless would have avoided the collision.
4
The "we used to be really hopefully about self-driving cars but now reality has set in and it won't happen soon" is a narrative as old as the idea of self-driving cars. At times I think it is the only narrative. Consider: "Even self-driving leader Waymo is struggling to reach full autonomy " from Arstechnica in Dec 2018; "Through All the Hype, Self-Driving Cars Remain Elusive" from the NY Times itself also in Dec 2018; "Uber’s Vision of Self-Driving Cars Begins to Blur" in August of 2018 again from the Times; "Automakers Discover What They Don’t Know About Moving People" from Bloomberg in July 2018; and "Self-driving cars are headed toward an AI roadblock" from The Verge also from July 2018 -- exactly one year ago when we were supposedly so optimistic and confident that self-driving is around the corner.
4
If highways have to somehow be 'wired' for self-driving cars, self-driving cars are doomed.
In the NY area, it takes about 40 years to get pot holes in 100-year-old roads filled. Then it takes about two years for enormous trucks to tear the roads up again.
Because fixing public infrastructure has not yet been fully 'monetized' to benefit big political donors, most heavily traveled roads look like they're in a war zone.
11
"Recently, one of the company’s cars encountered a bicyclist riding the wrong way down a busy street between other vehicles."
Um, that's EVERY DAY on the roads in any given city.
30
The only way to make self-driving cars work with current technology is to remove human-driven cars from the equation. Humans can be too eccentric for machines to figure out.
When I rode a motorcycle, I learned to keep a map in my head of where the cars were around me, so if something happened, I knew where the escape points were. I learned to loathe unpredictable drivers because I couldn't anticipate them or the others who responded to their actions. A self-driving car would have an even harder time with people who change lanes frequently, try to squeeze past other drivers, tailgate, get distracted, and all the other odd things that about 10% of humans do behind the wheel. These cause problems for humans; they probably baffle machines completely.
11
The Icelandic goddess Bjork spelled it out for these "experts" back in 1993:
"There's definitely definitely definitely no logic
To human behavior"
And yet they thought they could simply write a logical formula to predict it...which in itself is an illogical conclusion to come to. Oh, the irony!
BTW check out Bjork's music video for Human Behavior. It's a surreal masterpiece of the medium!
5
"The industry’s unbridled confidence was quickly dented when a self-driving car being tested by Uber hit and killed a woman walking a bicycle across a street last year in Tempe, Ariz. ..Elsewhere in the United States, three Tesla drivers have died in crashes that occurred while the company’s Autopilot driver-assistance system was engaged..."
And during that period of time, how many pedestrians and others have been killed by drunk drivers? How many have been killed by distracted drivers on their cell phones? How many have been killed by drivers who've fallen asleep?
Simply listing a tiny number of people who've died in crashes by self-driving cars does not prove that self-driving cars are dangerous. Indeed, I would guess that if The Times looks up the data, those self-driving cars are vastly safer per 100,000 miles driven than cars driven by humans.
Let's see the full story.
15
@Sean The question really must be: what are we doing to ensure that the software in all of these "self-driving" vehicles is robust?
Currently, the answer is "nothing". We're all crash test dummies for a multi-trillion dollar industry and I, for one, didn't volunteer.
https://www.huffingtonpost.ca/patricia-herdman/self-driving-cars-safety-software_a_23704668/
6
Do the self driving cars break for wildlife? What if a deer jumps out on the road...or a dog?..or a squirrel? Does the autonomous car try to avoid it?
3
@Wendy
They definitely brake for large animals that represent a blockage, just as they brake for people in the road. I don't know what the minimum size animal to trigger that is.
1
I feel like there being forced on the public like FB. Total bad idea all around.
11
It's funny seeing all of the Luddites at archaic car companies team up with other Luddites at other archaic car companies to make self-driving cars together. Then they realize the technology involved, and need to get a chip, battery, and screen manufacturer involved as well. It's like Iron Maiden splitting a royalty check between their 5 guitarists, comes out to peanuts.
Or, just license Tesla's tech. My holdings would appreciate that. They're closer and know it better than anyone involved. Despite 3 crashes, ask any insurance company if they want you either in a Shelby Mustang or a Model X.
2
@Chad
"Or, just license Tesla's tech."
Excellent idea - then they'll have an entity to sue when problems occur. Or don't you think problems will occur?
5
At my very first driving lesson at the age of 16, my father said these wise words to me, which have governed my decisions behind the wheel and somewhat in life: expect everyone else will do something stupid.
If the self-driving cars can program that in, we are good.
5
@Marian
One of my favorite sayings is "Expect the unexpected, but don't count on it."
3
This is a good article, but it failed to mention the inherent racism in software. These vehicles are literally programmed to see white people only, and can potentially never see a non-white one. I assumed software would see 'body types' (such as a dog) versus skin tone/facial features, but that is not the case.
@Jonahh
"These vehicles are literally programmed to see white people only," Do you have some reference to support that assertion? They brake for people wearing heavy coats, hats, and ski masks.
3
@Jonahh
The TV show "Better off Ted" had an episode where the automatic doors didn't work for black employees, so each black employee had to have a white "minder" to join them as they moved around the building.
1
I have a maxim, perhaps original: "Artificial intelligence can never overcome natural stupidity." Self-driving vehicles will eventually arrive but hopes that they will be perfect are futile. About the best we could expect is that they will be a bit better than the present condition.
1
@Bodger
"Self-driving vehicles will eventually arrive but hopes that they will be perfect are futile"
Of course they won't be perfect. The question is whether they will be more or less safe than human drivers. Considering the crazy things human drivers do like tailgating 5' behind a vehicle, texting while driving, looking in the back seat to talk to children, etc., I'm confident self-driving cars will be safer.
3
@Phil
"I'm confident self-driving cars will be safer."
Why? The autonomous vehicle will have to respond to all of those crazy things, won't it? There's a saying from software development - "It's hard to make a foolproof system 'cause fools are so ingenious!"
2
Excuse me for saying I told you so, but, well, I told you so. I have been sounding alarms about this hype for years and few people were willing to listen. In fact, my concerns about self driving, when shared with reporters and others in the field were rebuffed.
I am a technologist and consultant but self driving cars are not in my area of specialization, still if one just thinks about it for more than five minutes, the idea of a car piloting itself through the complexities of roads, people walking, dogs, children, rain, darkness, cloudy days...on and on the complications multiply into infinity.
One post that I made here on the Times online was ridiculed but it was a simple proposition: a self driving car would always have to wait when there was contention for space on the road. If car A wants to go and car B wants to go into the same space, who decides? Do the two cars then just sit there waiting to see what the other will do? Who mediates? If they both go, they crash.
With computers and smaller and smaller processors, you can solve a multitude of problems but, also, ask this: what problem does a self driving car solve? The engineers, ever doubtful about human capacities, say it will be safer. This is unproven and can't be proven until billions of miles have gone by.
I've thought from the start there are two places where self driving might work: an isolated center city area where everyone knows what to expect and on Interstate highways. Otherwise? Maybe nowhere.
10
US roads, bridge & tunnels are falling apart & outdated, and yet billions are being spent on the dream of a vehicle that drives itself. Mass transit is a joke or non-existent in many places across the country, but an entire industry is trying to build more individual vehicles. Europe, Japan & China have high-speed modern rail and we have little (relatively) old AMTRAK and are working about putting millions of high tech self-driving vehicles on those broken down highways.
Dream of autonomous cars? More like a nightmare.
52
Weather and maintenance of sensors. In cold weather it is common to scrape the ice or snow off of the winshield. Will this be easy to do on the sensors?
Sensors will degrade or wear out over time. The function of sensors is very important; a matter of life and death. At some time, ( 5 years?) sensors will have to be inspected, certified, or replaced. This will be expensive because the tests are physical, not digital. A radar system requires a physical test involving moving objects. $4000 tuneup? After 10 years? Many cars will be scrapped early due to sensor cost.
14
@AJK
I've already seen a few articles on auto websites (like Jalopnik) describing the surprisingly expensive repair bills for cars involved in "fender benders" that damaged some of the sensors that are required for driving aids such as avoiding other cars, "intelligent" cruise control, etc.
6
Self-driving vehicles, if/when they arrive, will not just be for the ultra-rich.
They will be for:
1. Commuters who want to work during their daily commute
2. Elderly people who are increasingly homebound by a fear of driving
3. Handicapped people who have difficulty driving
4. Trucking and delivery services addressing the unavailability and/or cost of human drivers. (Which raises a host of other questions.)
And others.
It’s still unclear if self-driving vehicles will ever reach the necessary safety levels.
If they do, however, they will be used by many people for a variety of reasons.
5
Pedestrians need to have some control in those arenas where autos and pedestrians mix. Control could be passive with pedestrians possessing electronic beacons which would amplify and ensure their detection by vehicles all the way to active control where pedestrians possessing electronic switches that could coldcock an autonomous vehicle if it got within a set distance of a pedestrian.
Anyone not caught up the in the irrational exuberance of the tech industry, and the breathless media hacks that feed the frenzy, knew that self-driving vehicles were way, way in the future, if only because the rationalizations for why they were so necessary were weak and the problems so evident.
14
The lesson for any application of "artificial intelligence" is that it is a lot harder than it looks. Navigating a static landscape is easy; dealing with unpredictable people isn't.
Start with autonomous trains, whose landscape consists of fixed tracks, signals, and at worst the occasional pedestrian or vehicle wandering onto the tracks (usually at known crossings).
Everyone who made a confident prediction of autonomous vehicles should have been more humble. Maybe in 5 years, maybe in 20, perhaps never.
6
Why, exactly, do we need driverless cars? Only because a clique of ultrarich uninformed influencers say so, and a not insignificant fraction of the public is naive and fashion-conscious enough to buy into this nonsense. Wouldn't it be much better to invest all of this talent and money into better mass transit? This would be orders of magnitude more energy efficient. As a side benefit, it would clear up the roads for those of us who enjoy driving. At 66, my hope is that driverless cars will take at least 30 years to become mainstream. In the meantime I'll follow my motto: " If it ain't got three pedals it's a golf cart".
29
@Ajax Because most of the cost of operating a vehicle is the driver. Anyone unable to drive (old, young, disabled) could be transported for a fraction of the cost of a manned taxi, in circumstances in which buses are impractical.
3
@Jonathan Katz This is a bit contrived. Sort of like saying that because some people need dentures we should all have our teeth pulled.
7
I'm waiting to see how driverless vehicles do on a city street covered by an inch of ice with 6 inches of slush on top and intermittent drifts, traffic confined to a single lane with opposing lines of cars taking turns going forward, while cars slide randomly past red lights or spin off to the side. Looks like I may have a chance to watch the experiment here in GR during the next year.
17
As an engineer who specializes in risk mitigation this is a topic of deep fascination to me.
Tesla's approach frankly terrifies me. The human attention span isn't capable of long stretches of inattention punctuated by moments critical, time dependent action. Tesla knows this, and is fully willing to blame the driver. At least with Uber (and I can't believe I just wrote that) they had the teremity to halt their program after the death of pedestrian. Knowing what I do about human factors, the death of a pedestrian and the death of the driver aren't that different. Tesla has killed 3.
I won't trust driverless cars until the steering wheel is gone. Until then, the driver is just the patsy for the machine's mistakes.
31
@Jen. Well said. The problem of maintaining situational awareness is a well know problem in the aviation business. Though it might not seem so, pilots have far longer, in most cases, to take over control since they generally are not driving about 19 feet from the tail of the plane in front of them. With autos on the road, there is very, very little time to take over control, assess the situation, and take corrective action.
8
I'd rather they concentrate on cars that make it impossible for drivers to text while at the wheel.
65
Hmmm....
Maybe we can get the driverless cars and make them travel in unison to reduce the chances of mishaps. And maybe we can put them on specially designed pathways to further reduce the chances of mishaps. These pathways can also have built in recharging or electrification. A final step would be to keep maybe just one human in the picture controlling the operation. Then we would have reinvented the railroad.
29
@JMM
That's brilliant!
1
This article, and the persons interviewed, seems to imply that human unpredictability is the only remaining obstacle to fully autonomous cars; that the actual driving part is in the bag. In a cursory search of the internet just now, this is not what I am seeing. Autonomous systems still have profound recognition problems. Snowflakes and raindrops, pigeons and seagulls (stationary and flying), graffiti or band stickers on stop signs or street signs or masking tape on pavement, all can confound even the most advanced systems. Autonomous vehicles have trouble with lane lines, (and where these don't exist, cars using other vehicles to determine the flow of traffic often follow lead vehicles onto off ramps leading away from the desired route). Tree shadows, hills and bridges, the list goes on and on.
20
The ONLY people clamoring for self-driving cars are the rich people who stand to profit from them. If they get the mortality rate down to .01% that will be fine by them.
And those of us who find ourselves in that .01% who are killed will be just one more "externality" to them. No thanks.
22
Best piece of information I have seen today. Driverless cars are not a concern that I want on the road around me.
8
First of all, human drivers are horrible at handling all these "unusual situations". That is why tens of thousands are killed by human drivers every year, two million are injured, in the five million accidents caused by human drivers.
On top of being bad drivers, humans drink, text, fall asleep, etc while driving.
The bar for autonomous vehicles is not perfection. It is simply to be less bad than humans.
Tesla is the only company chipping away at the problem and adding value to their owners today, with a constantly improving autopilot.
The other automakers are simply panicking and reacting.
Will Tesla get to full self drive next year? Probably note. Within the next 3 years? Probably yes.
And we will all be safer because of it.
6
It's strange how all these articles on self-driving cars ignore Tesla.
The fact is, a Tesla on autopilot is already safer than most drivers under 24.
For most of the situations described, most human drivers will do the wrong thing. The idea that autopilot needs to be perfect is a fallacy; they need to be better than the worst human driver.
5
@Mannyv. I have a Tesla and love autopilot. On long freeway road trips it does 95% of the driving, and does a better, safer job than I do. Nonetheless this experience has convinced me that fully driverless cars are a long way off.
5
I see multiple difficult-to-solve issues. A few:
1. I am in my new Autoauto on a two-lane highway. At a sharp bend I see many backed-up vehicles. It stops. OK, its database knows there is a bridge beyond my sight that likely is raised, so it will wait. And, wait. How long does it wait not knowing that the bridge is stuck open. Is it allowed to make an illegal or off-pavement U-turn? What if a policeman arrives and guides me to make that turn? How is the usual programming overridden or I take control?
2. Related to 1, does the authority managing the bridge immediately send a message to some central data bank available to all autonomous vehicles so a timely decision can be made? Requires a ubiquitous communication infrastructure.
3. Signage for detours due to construction or accident is often haphazard. Will even the smallest government organization also furnish the needed information on a timely basis?
6
So there's two robots on a date in the back seat of a driverless car. Oh, the electricity between them!
There are some limits to technology, are there not?
2
Watched a presentation from a leader in AI development, his opinion was self-driving cars were not coming. Too complex once you add irrational human behavior in. the exact reasons mentioned in this article.
That was three years ago. Humans haven't changed, shoot look who we elected.
23
The article does a good job of pointing out some of the major obstacles to this emerging technology, but I wish it would also have discussed what may be an even greater threat than failing to detect a pedestrian...the security of these proposed vehicles from outside interference, i.e., hacking into their systems.
Even non-autonomous cars are at a growing risk of such intrusion, as they become ever more dependent on software to control every facet of operation and are in constant (and blind to the driver) communication with a variety of servers. We still can't figure out how to fully shield huge corporations and government operations with their elaborate networks and systems to actually prevent (not just "deter") hacking and control seizure.
We also need to design in-car systems that will not fall prey to the bane of every user, slowing or freezing of their computers, even momentarily. Barreling through a red light or into a pedestrian while a spinning ball or rotating hourglass suddenly pops up saying "I need a few seconds" is totally unacceptable. We are, I fear, a long ways away from that sense of security.
6
I truly believe the real "end" to all of this is completely isolating the driving lane from pedestrians. Not to prevent cars from hitting people, but to prevent people from walking into cars. There just isn't any other way. Human error(humans making mistakes) will always, always be the primary driver in whether or not autonomous vehicles are "safe".
This is why interstate trucking seems like such an appealing early target for autonomous cars. There are no pedestrians and bicyclists on interstates.
I really miss actually driving a car all by myself. I loved driving a stick shift. They are common throughout the world, but rapidly disappearing in the US. I enjoyed working on my car, tuning my own engine, changing the filters. I took pride in making the engine purr like a kitten. Everything now is fuel injected, electric, or computerized. My timing light and spark plug gap gauge seem so antiquated. I liked taking a good old fashioned key with me when I left the car, not a silly fob that gets left behind in the cupholder more often than not. Cars used to be fun to drive, but today's cars insist on driving for me. What an incredibly boring experience driving has become.
12
This is an excuse and the real reason is litigation. Who would be at fault if a self driving car causes an accident? A software engineer, a mechanic, a CEO at the car manufacturer? Yes, I am being sarcastic about the CEO! As it is now, responsibility for accidents is mostly outsourced to human drivers. Taking on that responsibility will deter auto companies from self driving technology for some time.
3
@GGI
Excellent point, GGI. This is similar to the problems with the Boeing 737-Max. The thing crashes, who is at fault?
Along with software engineers, I'll bet some ambulance-chasing lawyers are already working on this.
4
Luddites often have really strong opinions on technology.
@P.S. - so do people who understand the complexity that financially-eager tech wizards gloss over.
Fully autonomous in all situations is, to use a technical term, really hard. Tesla's algorithm steered a fast-moving car into a highway divider, not 'seeing' the mass of stationary concrete as an impediment to the car. Uber's algorithm 'saw' the woman with the bicycle 6 seconds before killing her, and took no actions until less than 2 seconds before impact.
The amount of three-dimensional computation, and the set of complex, interacting rules that typical human adults apply all the time to drive safely is a testament to the fact that humans are, from birth, training to operate in a messy, complex environment, using a massively complicated neural system.
For Tesla to say his new chip will replicate that in a year or two is a call to eager investors, not a technical reality, and pointing that out is not being a Luddite, it's consideration of what's involved to get to full level-5 automation.
Humans start with baby steps (after wiggling and crawling). Google's the one major player in self-driving to be taking things very very deliberately. I don't praise them often, but in this case, they're more realistic than the jackrabbits.
18
@P.S.
"So how long did you attend school?"
"Just until I learned the meaning of 'Luddite' - I use that in all conversations that are too technical for me to understand."
3
@asg21 these same sentiments have been on offer for every major innovation in transportation; from trains to cars to flight to space travel, and the luddites have always been wrong. But you're right, here at the NYT it's very clear that we are to agree with the majority and if we don't it's because we don't know enough to understand that they are right. Thanks for the subtle reminder.
1
"developing software that can reliably anticipate what other drivers, pedestrians and cyclists are going to do, will be much more difficult, he said."
Humans too, as proven by the number of road deaths and injuries, can do no better than AI, and probably worse. Waymo is the leader, the rest are far behind, but self driving doesn't have to be perfect, just better than humans, an easy goal. Waymo is already there.
2
Driving home from DC to Kansas 5 years ago I realized that the first big market for autonomous vehicles will be long-haul trucking. Autonomous driving is well suited to freeways, and so long as truck depots are placed along freeways, there is little reason to fear the crazy-human driver/cyclist/pedestrian factor.
And the payback period of the autonomous-driving technology will be relatively quick if one considers the following: there’s no driver to pay, no mandated limits on driving times (as computers don’t need to pull over to rest), no need for a cab that can fit a person and certainly no need for a climate controlled cabin with seating and windows, etc. which will improve fuel efficiency.
Autonomous-driving technology is decimate parts of the labor force, but if the objective is cheaper, safer, more-energy efficient transportation, long-haul trucking will be the first major transportation segment to see its successful implementation.
1
@It’s News Here I worked at a large retail store which had an 18 wheeler trailer used for storage parked in the receiving bay. When we got deliveries the driver had to disconnect his trailer, snag ours pull it out then make the delivery, then all that was done in reverse order as he left. There is no way a driverless truck could handle that situation.
4
@Richard Large truck depots already have specialized vehicles and drivers for moving trailers within the depots. And those depots are already swapping out cabs for use in delivery to retail stores and other locations. The point is that the long-haul portion of truck transport is ripe for change and the industry will adapt to accommodate the new needs that come from the enormous cost savings associated with autonomous long-haul trucking. But as you point out, making deliveries to final destinations will not be part of the near-term opportunity for trucking.
The title of this article says it all.
Despite the constant hype (the press conferences and announcements are just PR tools to make these companies seem relevant and progressive), we are many years away from self-driving cars.
Imagine what it will take to be sure that every road in America (and/or the world) has a visible center stripe for sensors to read, or that the various municipalities, insurance companies and manufacturers figure out the liability issues? Not to mention snow, rain and ice conditions.
We're a good 20-30 years away from even crude implementation.
Don't get me started on electric cars...
7
I don't think it will ever happen. Just read up Integrated Information Theory on the Internet to understand why. To do what even an average human driver can do today would require massive amounts of computational power and all the deep learning in the world, and then some.
As many has pointed out, the best they will be able to come up with in the short term will be vehicles travelling within a controlled or predictable environment.
Alternatively, we have to remove all humans from roads, but that would be impractical, wouldn't it.
7
I’ll add that they’ll have to be constantly updated to be able to beat Adversarial Machine Learning techniques. There’s a video out there from some researchers showing that you can “disguise” a stop sign by taping a few pictures of bottles to it. Unless there’s a big leap in deep learning technology, we can probably always expect there to be some way to exploit the models that the cars employ.
2
Nobody ever asked me or anyone I know in a poll or in some legal process whether we wanted driverless cars. The Times, saying the country is eager for this, is acting like an advertiser for the industry.
39
This has been obvious for years. Autonomous vehicles will not be seen in wide-spread use until the 2030's, if then. The problem is more than technology: it's reality. Autonomous cars have to drive among human operated cars, humans on bicycles and humans on foot. This is beyond a technology problem. However, out on our highways, where pedestrians, bicycles, traffic lights and stop signs are not present, we could see autonomous, or at least, self-driving cars with a human behind the wheel, in the next ten years. But keep in mind there are myriad other reality issues, such as regulation, insurance and procedures when accidents occur. What if you're riding in an Uber and your vehicle collides with another vehicle. You are a witness. Are you required to stay? File a statement? Fault must be assigned, in most cases, but can humans just walk away? What if you're injured? These types of issues are the tip of the iceberg. So, that Uber and Lyft stock people bought is worthless. Both companies are highly dependent on autonomous vehicles to achieve profitability. Seriously.
2
This is great technology, and I love the driver-assist features in my car. But the key to safe autonomous vehicles is true situational awareness, and it will be a great feat for an autonomous vehicle to match an attentive human driver.
The example I often think of: I’m driving down the road and see some young kids a block ahead running around and playing with a ball off on the side of the road, but not in the road. I slow down as I approach, so in case one of them should happen to run into the street, I can more safely stop. An autonomous vehicle would maintain its speed and then slam on the brakes only when the potential hazard is realized. There are so many scenarios where preventive action based on situational awareness can reduce accidents, or at least their severity.
Granted, not all human drivers are as attentive as they should be, unfortunately...
21
@Bill C
Excellent point & example. Years ago I had the opportunity to participate in a Commentary Driving course at my place of work. It taught situational awareness ..... looking ahead; scanning; what do you see; how could what you see impact your safety and the safety of others i.e. what if scenarios. I considered myself a good driver then ... this course made a lasting positive impact.
1
No computer knows what is going on in a person's mind, which means that a computer cannot possibly anticipate what another driver (S) is going to do. Also, there are many other elements that come into play: weather conditions, traffic density as examples. One person has already been killed due to self-driving car. I prefer to rely on my own faculties than put my faith in this technology absolutely.
4
@Andrew Henczak
The argument seems to be that since there are stupid drivers, it's acceptable for autonomous vehicles to kill people, as long as they're not doing it too often.
1
Our cars can automatically parallel park for us. Adaptive cruise control will adjust your speed if you get too close to another car. Your car will automatically return to your lane if you start to drift. About to hit a person or object? They'll stop quicker than you could. Our cars will alert you if you're about to merge into another car. Backing out of a parking space? Here's a 360 degree top-view image that shows you everything around your vehicle.
For every person who says they fear autonomous vehicles and don't think the technology is needed, do you welcome the above safety features? The cars on the road today, while not fully autonomous, are already driving themselves.
You will not wake up tomorrow to see empty cars driving themselves around. It's happening gradually and drivers are getting used to it.
2
@Jake the problem with "driver assist" technologies like lane keeping is that the better they get, the more drivers rely on them, and the less attention drivers pay to driving. That's not a theory, that's a fact. The worst case is when the assist technology works 80-90% of the time... driver stop paying attention, but there's still a pretty high chance of failure.
5
@Jake
Beautiful - if a Tesla can do all of that now, why is any further development needed? Let's stop fooling around and remove that steering wheel!
1
Good. I can’t imagine ever getting into a self-driving car. But then, back in the day, I resisted ATMs, preferring to deal with humans, until the tellers got so snarky I preferred machines. Now I resist self-checkouts at the market, because I don’t want the humans to lose their jobs.
But self-driving cars offer too much upside for tragedy. With any luck, by the time they make it to market, I’ll be too old to go anywhere.
5
The article points out the big concern - Drivers and self-driving cars on the same road which is why four years ago I predicted that in 25 years human drivers will be outlawed on freeways in LA. In LA our problem is simple - We can't build more freeways and traffic continues to increase. With autonomous vehicles traffic jams will be eliminated and traffic volumes significantly increased. Other benefits? Here in LA, we could go from 150+ accidents on an average day to fewer than 10.
More importantly, as autonomous cars become more available fewer people will want to own cars. Why pay for a depreciating asset that is parked 90% of the time (one that mostly drives itself)? Car services will be cheaper and even more convenient than today, think shared jet ownership for cars. Cars with drivers will be a lot more like horses are today. Some people will still have them, but most won't.
1
Autonomous cars were always a pipe dream, at least for the vast majority of us alive today. Experts in automation have been saying this for years. If only the media would occasionally take the time to listen to the critics rather than cheerlead the charlatans.
4
@GT Tesla has auto pilot that works. Today. You can watch the videos on youtube. There are limitations yes, but for your daily commute, or a long haul drive under most conditions you can use this today.
And it collects more data and improves continually. The future is here, its just unevenly distributed.
@Kevstev
Exhibit A of totally unrealistic expectations. You're going to be very disappointed.
1
To everyone asking "why?", the answer is 33,000 fatalities per year. We should be investing as much as we can in this technology.
3
I would love to see them tested more in winter environments. The public exhibition of a self-driving vehicle on a closed course in Minneapolis last winter was foiled by snow blowing across the road. This is 90% of winter driving in Minnesota. Not falling snow, just snow blowing across the road. This is nothing compared to falling snow, packed snow on the road, narrow plowed streets, icy roads, and so on. So, until they can manage those situations, it's not going to be realistic. Whenever I read these articles, they always talk about primarily urban environments creating challenges, but winter weather needs to be dealt with.
6
With my eyesight gradually failing it is clear to me that all the driving I actually do today could be easily done by an autonomous vehicle, and more safely --- were it not for aberrant behavior by more able drivers. And certainly interstate highway driving could be dramatically improved outside of urban zones by
optionally autonomous vehicles.
1
You can’t expect autonomous vehicles to cope with the insanity of the roads in a nation where police don’t even enforce the current lax traffic laws and drivers are only required to be tested once behind the wheel at age 16.
10
Too many cars. Self-driving cars won't change that. We have to move away from the car culture.
That being said, as a frequent pedestrian and bicycle rider, I trust the future of self-driving cars over a distracted, frustrated, over-aggressive, under-trained human driver any day.
3
@Woody Guthrie
Wonder if the cyclist killed in Tempe by the autonomous vehicle felt the same way before she was hit?
2
Are self- driving cars what people really want? Or is it reliable well-planned public transit? Sitting in traffic and long commutes by car make no sense at all to young people. As more people move to urban areas, driving to work in individual cars is not a solution. Austin is good example of an otherwise great place to live, facing worsening traffic congestion every day.
65
@LA
You aughta come to Nashville. Talk about congestion!
Frankly, I am shocked at the number of "Fred Flintstone" comments. For some really good laughs, change the subject from autonomous vehicles to electricity, the internal combustion engine, airplanes, putting a man on the moon, cell phones, and countless others and then read the naysayers comments. My sides are hurting.
2
@Bill Harvit
Another example of how Spanky was elected. Cluelessness squared.
2
Chuckle - one of the more recent examples of attempting to deploy "technology" for the sake of deploying technology.
6
All you have to do is collect data and say the words “artificial intelligence!” three times while waving around a stack of gullible investors’ money and magic happens!
16
@Miker Best comment ever!
2
I hope self-driving cars never happen-the dumbest idea ever. At the very least, I hope I'm dead by the time they show up. As it is I don't want a car that beeps at me and distracts me with all the new "safety features" for distracted drivers. I don't want drivers around me to rely more on technology, I want them to pay attention and follow the rules of the road, which they very often do not. We need to get back to the idea that every driver is responsible for their car and what it does and not abdicate our responsibility to technology.
13
Finally an article that states the obvious for anyone that has basic knowledge of computers and the present state of the art. They are not even close. Like nuclear fusion it will always be a decade in the future. Not only by the limits of technology that are still daunting but also particularly because of liability constraints. Can you imagine what will happen to a company responsible for mowing down a whole kindergarten of kids at a crosswalk?
35
"Way in the future?" Thank God. Honestly, when did masses of consumers say to the auto industry, "We want self-driving cars!" We didn't. We can't even get vending machines to work well consistently. Granted, our current system is quite hazardous and polluting and it needs to change, but I have little to no trust in automated anything, let alone cars (or buses, planes, or trains.) Yes, cars are our single most lethal weapon when drivers are drunk, high, using cell phones, eating burgers, disciplining kids, juggling pets, ignited by road rage, etc. Thus the case for creating self-driving cars...I know. I simply don't trust the technology it will take to make them safe, and I don't trust all the distracted drivers out there. My psychologist would probably say it's just a matter of my not wanting to give up control. She's right.
8
I don't work with autonomous vehicles, but I work with big data and am using and aware the of current state of deep learning and other 'AI' methods, and I thought these "self driving in one year" predictions were silly.
Here's what I think happened, it's basically two parts - first, the executives and promoters didn't understand the complexities involved with pattern anticipation algorithms, and thought the entire problem was developing adequate sensors and throwing some sort of undefined computation at the data - the engineers would figure out what kind of 'computation' would need to happen and it would just work!
The second problem lies with the engineers - they likely got excited about the progress the sensors and object/movement pattern processing were making in controlled environments, started to give radically optimistic
progress reports to executives, and here we are.
Real AI that deals with, matches and anticipates human behavior is *extremely difficult. It's not "here's what all the sensors see, now do this because the rules say to do this when we have this input and everything external follows the same rules all the time."
We're putting these machines in *our world, not vice-versa.
51
@mcomfortBest comment IMHO.
1
Wow so many negative comments.
I look forward to the day when I'm no longer required to be intensely focused on driving to prevent my death. I would much rather climb into a car designed more like a lounge, tell it where I'm going, and sit back and let it do the work.
Humans are unpredictable? Get em off the road.
At least half the 40,000 deaths per year are currently caused by drunks. Get em off the road.
Per my observation, a lot of highway congestion is caused by folks that are not familier with the local situation. Folks that miss an exit, or cross 4 lanes to make and exit. Get em off the road.
Need to text while driving? I would rather you do that from the back seat.
I have no doubt that computer managed highways will be much safer.
In the future, folks that want to drive themselves will pay more for insurance. As they should because they will cause more damage on a relative basis.
Can't happen soon enough for me.
213
@may21ok
Maybe a long ways into the future. Even the envisioned cars still include the assumption that in certain situations the driver will take over. Modern commercial planes, some of the most sophisticated machines in common usages day, still expect the pilots to take over in certain conditions where the computers simply can’t process what’s happening. And pilots, unlike 99% of drivers, are highly trained. Reading, texting, relaxing, or whatever, it’s still a two ton machine moving down the road, not a “lounge”. There is no way most drivers could suddenly look up from their screen or book and in a split second comprehend a problem AND take the appropriate actions.
28
@may21ok
"Get em off the road". Perhaps with infrastructure and public transportation? We have encouraged the use of cars over many years and because we do not wish to pay for things that do not directly benefit us, we have no mass transit as they do in most other countries. Not to mention our aversion to "the public".
There is no other way to get to work or to visit relatives than our streets and highways. How do we "Get em off the road"? Autonomous vehicles will not reduce congestion.
It seems you may really mean "Get em off my road".
9
@may21ok
When I first imagined self-driving cars, they traveled on specially designed roads—the route and the driving would be interconnected. I actually had a dream circa 2002 about something called a "google car" that was self-driving. It was small and looked more like the cockpit of an old fighter jet, with glass visibility front and above.
I heard someone talking a couple of years ago about converting roads and parking lots into solar panels—that these hot, reflective surfaces waste a tremendous amount of energy. It seemed to me that self-driving cars could have been developed in conjunction with a complete reenvisioning of our transportation grid, with a whole new concept of energy and navigation. As a sexagenarian, I wanted to live to see that. I still hope to, but in these dumbed-down times we haven't had a new national vision of transportation since Eisenhower.
I share anxieties about AI and self-driving cars, but I guess I imagined it as form of regulated transportation that still gave the individual free rein to GO. Jump into a bullet-shaped, efficient "google car" and go where you need without regard to anyone else's schedule because of our glorious network of roads. This "vision" preserved all the independence of personal cars with the efficiency and economy of public transportation.
Just a silly dream. But I'm still thrilled at the prospect of what this could mean to me in my old age and to the mobility of people with various disabilities. Think of it!
2
In the meantime, can we please focus on improving public transportation and reducing the air pollution generated by millions of cars?
35
Famous quotes of the NYT: "Experts are skeptical that Tesla can pull it off." This opinion has been wrong so many times before that it should be a viewed with its own skepticism.
1
@Roget T
Don't forget - too much Kool-Aid is bad for your teeth!
1
Driverless cars need all other cars to be driverless and probably a wifi infrastructure to help communicate with other cars. It also needs two fail safes. The first to immediately move the car to the side of the road in case of w/e and the second to back up the first one.
This will happen, and it will save lives as soon as people realize driving is not a right.
1
The term "self driving cars" is a bit of a misnomer. All of these cars are all still operated on computers that are programmed by humans. The machines competency is based on the accuracy of the programmed software, and the human behind the wheel. The cars today should be more accurately called driver assisted.
When we reach true machine learning, where a system reaches general intelligence on par with a human, and can "think" on its own, then we'll have a true self driving car. And the propensity for human error behind the wheel, and some bad code, will be vastly reduced.
General intelligence is the hardest nut to crack in AI. And once it's reached super intelligent machines won't be too far behind.
On the 1-10 spectrum of self driving cars, we're probably at around a 5 now.
2
With all this money and time spent on something that is very likely going to be less than satisfactory, a really nice train system (or a few train systems) could have been built instead: safe, reliable trains can't be beat....
Why is America so wedded to the automobile??
137
@Gemma Trains in Europe are wonderful, for tourists. Possibly we Americans are more anxious about sharing space with strangers, reluctant to surrender the autonomous personal capsule.
3
@Gemma Not even close. We don't have the population density of Japan. Or even Europe. California's rail to nowhere is nearing 100 billion dollars with no end in sight.
2
I can't wait until self-driving cars are the norm and no one is at fault when the vehicle goes awry and the response to a fatality is a routine mandatory mediation between the victim's family, vehicle manufacturer and the firm that wrote the control code for the vehicle (assuming they haven't outsourced it to a contractor offshore). The mediators likely ruling, using a typical parallel for what kind of damages get awarded for software malfunctions today, only a public outcry would get an award from a prescribed table for the death and the penalty would be that the manufacturer has to give the remaining vehicles of that type a free software patch. No one is responsible and everyone goes home happy except the individual below ground, sort of like new era no fault.
5
Tesla floats fully self-driving cars as soon as this year. Many are worried about what that will unleash. - Washington Post
@Markus
Since it won't happen, I don't think there's any need to worry.
2
The hard truth we all EE guys new about for very long, but had trouble to confess even to yourselves or our priests...
The best hope for driverless cars is on the limited access freeways, where some restrictions on human drivers are already in place. These highways could be restricted to driverless-enabled vehicles only. So, for example, the human driver of a driverless-enables vehicle could enter the highway ramp and release control of the vehicle to the driverless option.
This would be a great help on long distance trips, and provide an incentive to add a driverless option when purchasing a new vehicle. Additional technology could be added to restrict enabling the driverless option for use on designated highways only.
4
@Ron Bartlett
Great idea! Let's restrict it to roads that accidents never occur on, OK?
1
Humans can't predict what other humans will do. That's why you have hundreds of thousands of car crashes in the USA each year resulting in 40,000 or so deaths. When I'm on my bicycle I wish all cars were self-driving.
36
@J lawrence There might be some truth in that, but at least if you do get hit there is little doubt who is at fault. When the offending party is an offshore contractor who caused the accident by writing some bad code for a vehicle manufacturer sub, things are going to a lot less clear cut.
6
@Kevin
No, completely the reverse as you'll be able to sue a far more deep-pocketed entity.
My hope for self-driving is will be able to get the horrible drivers off the road or at least no longer behind the wheel. Of course in my mind that doesn't include me but I do have decades of heavy urban driving without an accident.
What is understated in stories about self driving cars is the serous and large problems they hold the promise of solving.
Human drivers are terrible and cause 10’s of thousands of deaths and even more permanent injuries. They drive drunk and sleepy while texting. They grow old and slow to react but still drive on. There ate also those excluded from the independence a car offers by sight or physical disabilities.
We are (rightly) horrified when a Tesla’s system makes a mistake or an errant pedestrian crossing a busy road is hit by a self-driver but seem to put out of minds the carnage that the current human-driving-a-car combination causes, and the potential to cut that dramatically, and we don’t seem to consider the blind elderly, and infirm who will gain dramatically from them
5
@Mark Marks
So how will Tesla take care of all the bonehead drivers driving normal cars? Do you somehow think that their behavior will change as more Teslas hit the road?
3
This is good news to me. I’ve often wondered how self-driving cars could ever navigate in my city. Here in Portland, every intersection is a crosswalk, whether marked or not, and pedestrians have the right-of-way. So pedestrians are included in the generally polite traffic dance. When I’m driving I have to scan the sidewalks for peds getting ready to cross, and yield to them. When I’m walking I sometimes step back as a signal to a driver to go ahead. Drivers often give us a hand signal that they see us. And, the usual right-of-way order for cars at a four-way-stop is changed: if a ped is crossing in your direction, you can also move into the intersection.
In my IT career, I observed many instances where executives and investors stubbornly underestimated the sheer complexity of a problem. (Fortunately that didn’t prevent us from going to the moon.) But I hope I never get the opportunity to “participate” in testing these cars!
9
"The hard part is anticipating what they’re going to do next.” This goes in both ways. How can a pedestrian tell what an auto-driving car will do next? There is no eye-contact with non-existing driver. Nobody understands human consciousness at all, and thus we cannot create a conscious auto-driver. Do not let these cars drive on ordinary cities with pedestrians, period.
63
Autonomous vehicles are a pipe dream. I wouldn’t get on a self-flying airplane and I wouldn’t get into a self-driving car. Manufacturers don’t talk about what would happen in snow, since that is an inconvenient problem. The best we can hope for are smarter vehicles that will provide assistance to drivers, just as commercial jets for pilots do today. Autonomous vehicles may be useful for a very limited range of uses but it’s hard to see how it will ever go beyond that. Perhaps a better public policy would be to raise standards for getting and keeping a driver’s license and heavily penalizing distracted driving. The government should force manufacturers to severely limit on-board entertainment systems that encourage distracted driving. That won’t happen, of course, but it’s what should happen.
10
The biggest fraud is Tesla’s “Autopilot” which has already killed several people by driving them straight into highway dividers or semis, etc.
2
@Jay Lincoln, I don't think we should expect Tesla's autopilot to be perfect. I think we should expect it to be safer than if a human were driving. So to me the question is, even if several people were killed while operating in autopilot mode, how many would have been killed if they had not been driving in autopilot mode. So last year, if 3 people were killed in Autopilot mode while 50,000 were killed while humans were driving, which is better?
3
@Jay Lincoln Absolute numbers of deaths are not informative. What is more meaningful is numbers of injuries/deaths per miles driven, per hours driven, per miles of highway road versus local streets etc. These numbers must then be compared to human-driving rates of injury. Then we can have a discussion.
1
@Marty
Your comparison is skewed. You have to include some notion of mileage driven and the number of cars on the road. For example, if there are 50,000 times more drivers on the road driving 50,000 times more miles per year than autopilots, it may be that 3 people killed in autopilot mode while driving very very few miles is actually way off the charts.
1
Yugo first. Pun intended.
4
You don't have a lot of faith...
Business will not allow anything to stand between it and cash.
Think of RoundUp, the herbicide currently in the news for damaging lots of people.
For decades foolish scientists tried and tried to get formulas that would kill, say, dandelions and leave corn alone.
One night, while watching The Simpsons down at the lab, they their "Doh!" moment: The problem isn't dandelions, or the RoundUp, the problem is corn.
If the stupid plant can't keep itself alive in the face of the roughest, toughest, kill-everything-in-sightiest chemical on the planet, well, let's fix the corn.
So they did. Now, Americans can hardly find corn not genetically tinkered with to survive a chemical onslaught that would clean up the Amazon (Not that Amazon, the other one. Unfortunately.) in a fortnight.
Same thing here: if the cars can't deal with people, or snow, or potholes, the problem is the potholes and the snow and, especially, the people
Time to redirect the engineers, bring in the docs and the shrinks, and put the blame where it belongs.
Doh!
3
What a dumb idea. The best thing about automobiles is that you get to drive them yourself - anywhere, anytime you want, fast, slowly. You can join SCCA. Fun. People don't see what they have - take it all for granted. Shooting themselves in the foot. Dumb idea.
5
“Self-driving” is a broad term
encompassing everything from features already present in vehicles to fully autonomous vehicles, which are likely decades away. People need to be very specific about what they’re talking about when they say “self-driving,” and tech journalists should nail down those specifics before we get another flying car-like delusion.
The corporate media, as usual, went along for the driverless ride with the corporations hoping to profit from driverless cars.
The government aided and abetted by passing legislation that was heavily criticized by Consumer Reports as being short on safety and too accommodating to the companies promoting this risky technology.
The whole affair illustrated the problems inherent in a capitalist system in which corporations and “deep-pocketed investors” virtually control the media and the political process. Potential profits trump human safety - we can thank the heavens for our country’s extravagant tort system which would’ve held corporations responsible for any consequences.
Apparently- and luckily for us - the corporations calculated that the liability losses couldn’t be easily offset by the profits.
9
What's the point of self-driving cars? What existing problems will they solve if they get perfected? What improvements to our transportation system might they provide? I've never seen a justification for self-driving cars other than they pose exciting technological challenges to create them.
Transportation in the US is plagued by two problems: too many vehicles on the roads and the smog and greenhouse gas emissions they all cause. Solutions are well-developed and evident in many places around the world - improve mass transit!
5
@Steve, the first point I see is safety. Robots don't text while driving, they don't fall asleep, they don't go to the bar, have 6 drinks, and try to drive home. How many people are killed in car accidents today? How many fewer would be killed if cars could drive themselves? That's the point in my opinion.
6
@Marty
So robots are going to go to bars to prevent drunks from driving? I agree that that would help, but I don't see how a BEV take-up rate of far less than 1% will affect all of the other bonehead drivers.
1
I’m driving a Tesla model 3 with autopilot cross country and it does drive itself in many circumstances just fine. The biggest issue for me is my own fear of letting it go but gradually I learn to trust it. Of course it’s not perfect but no new technology ever is, besides it doesn’t need to be perfect only twice as good as humans. So far I’ve had a few “phantom “ breaking incidents where the car slows down for no apparent reason and I’ve duly reported those to Tesla on the spot. It also prevented me from switching to an occupied lane at 60 miles/hr and the autopilot wasn’t even on. I believe autonomous vehicles (cars, buses, trains, etc.) will happen and a decade it’s not too long to wait.
4
Auto/Technology companies are, it appears, attempting to solve all of the issues before the technology is released. How about developing a simplified version – similar to the ubiquitous cruse control.
Program auto-driving (a) to limited access roads only (known via most GPS mapping), (b) to over 35 MPH only, (c) stay within posted speed limits at all times, (d) be disconnected when driver touches breaks or accelerator.
A few other safety items could also be programmed in: shutdown transponders for highway workers; required presence of a seated driver (perhaps seat sensors); disconnect in extreme weather (snow, heavy rainfall) which can give false data to system.
State governments would be wise to get ahead of this technology and create the appropriate regulations to help ensure the safe implementation of these systems.
@Stevenh Exactly what Tesla does today.
Most of what you suggest is already done by Tesla .
One of the easiest ways to determine if technology companies have drank their own Kool-Aid is to see if lesser challenges have been solved by these companies effectively. I'm not even sure if the car industry and regulators have completely solved the problem of airbags randomly decapitating people?
Trains run on rails in a single direction of movement that the operator has control over. A train does not have steering or turning ability, nor the infinite possibilities that exist when thousands of vehicles, pedestrians and fixed objects interact on roadways. Any unknown obstacle to a train would usually be at limited points, such as vehicles at rail crossings, pedestrians at stations.
Putting automated trains and switching in systems where there are far fewer human and engineering variables to solve would be a logical first step for transportation technology. Once that's ubiquitous, up and running without incident or fatality for a number of years, then try cars.
2
At the limits of AI, the neural networks engaged in "driving" have no idea what driving a car actually entails. As Roger Penrose aptly described it decades ago, we are spending a fortune on the "emperor's new mind" only to find that it is made from unobtainium.
17
The shuttle described in the article seems like the best use of driverless tech as we need to reduce the convenience of single passenger car use. It allows public transit to create a microgrid of smaller vehicles that can reach more people touch points. The only question is how well behaved would the public be without a driver around. Can be bad enough when a driver is on board.
A self driving car does not solve the problem of congestion. There is a way to not have to drive yourself, it’s called public transportation. It works in lots of other countries and somewhat works in some US cities. It requires investment but it benefits all. Why not?
239
@Sara S - You are absolutely right about utilizing public transportation. Their are various public transit modes that effectively move the masses (on the surface and underground) commuting in and out of the urbanized areas. It is the individual motorists choosing to drive who are the root cause of traffic jams and congestion.
Investing in the modernization of public infrastructure should always be at the top of the government's to-do list. Governments at all levels are responsible for executing such capital projects through public authorities and transportation agencies providing oversight and accountability.
16
@Sara S Hot, dirty, and unsafe, and that's just the bus stop or train station.
Slow and inconvenient as well. public transportation also makes the movement of personal goods difficult.
I'm afraid the it also doesn't look good on any cost/benefit analysis outside of a few areas of high density.
7
@Sara S Self Driving cars may not "solve" congestion, but they can absolutely minimize it. When an entire micro-area of traffic is communicating, the cars effectively act as one. That means the cars self-adjust speed etc for the most efficient integration into and out of lanes, on ramps, etc.
Humans, on the other hand, are completely irrational; swerving in and out of traffic, ultimately causing more congestion.
While I wish it weren't so, improved public transportation is a pipe dream in the US due to the way this country's infrastructure has been built. Furthermore, Americans, on average, live too far away from their places of work. Fixing public transportation would require a near complete infrastructural overhaul, which is close to impossible.
6
I've just wondered whether any society is mature enough to accept changes induced by automation. I think it is obvious that self-driving-autonomous-cars will not only change the car technology itself, but also radicalize traditional economy and transportation system. A lot of jobs will disappear and conflicts involved with new technology will break out. I'm not sure how many countries are prepared to handle social and cultural transitions associated with automatic cars.
Editor’s note: This comment has been anonymized in accordance with applicable law(s).
1
It used to be turn the key and step on the gas to go.
And step on the brake to stop.
Now it is an endless series of mystifying computer messages, admonitions, warnings and beeps.
I am altogether baffled by the new cars.
So are millions of other senior citizens.
The difference between them and me is that I have quit driving, and they have not and are still out there with you.
1
I neither want one. Nor do I want to be on the roads, where one might be driving. Into me.
Full Stop!
6
@TheraP, I'd rather be on the road with self driving cars than with drunk drivers or texting teenagers.
40,000 people in the United States died from auto accidents last year. I'd rather be on the road with self driving cars. It would be much safer.
Our current model of software running on silicon will not close the gap that will permit a 4500 lb minivan to drive down a residential street full of playing children with nobody aboard. The kind of intuitive thinking that the human brain does probably requires a different computing model, perhaps biologically based and certainly massively parallel and quantum in nature. Right now, quantum computing is stuck, like fusion research. So we are at least 30 years from fully self driving cars, if not longer.
8
If things go the way I would like to see them go, Self Driving Cars are the cars of the future "and always will be". I believe the automobile industry and Silicon Valley are trying to further exploit and disempower automobile owners by forcing them into a passive lease holder role in private transportation and that this trend is something we should resist.
25
Having looked in some detail at the technology, I have long believed that fully autonomous cars are at least 20 years away. Some will arrive sooner, but use will be limited and regulated.
We are already seeing some basic capabilities, such as automated emergency braking, intelligent cruise control and automated lane keeping. I just rented a car with the latter feature. It wasn’t perfect, but the car could drive itself, or at least keep in its lane, on some highways and even secondary roads. It’s a long way from full automation, but that will happen.
It is interesting that so much of the testing of AVs is done in urban and suburban environments where the challenges are so complex. I understand the thinking here, but It makes more sense for self driving cars to be operated on major highways where there is less going on. Cadillac and Tesla can do this (in limited ways) already. Others will follow.
1
Back in 1985, I was really hoping that by 2019 the focus would be on flying cars powered by batteries not self automated driving cars. So much for being the Jetsons putting thoughts in my head.
5
Self driving cars will never exist unless all roads are wired, and autos can communicate with other autos. Even then it will be a pedestrian massacre.
The amount of citizen/government interface required will negate even the thought of privacy.
In America infrastructure is a four letter word.
40
If Tesla is gathering data from hundreds of thousands of vehicles, about both traffic situations and the way their human drivers reacted, while other companies have only hundreds of vehicles on the road, I think I can see who is going to win this race.
4
@James Van Zandt
So long as that race doesn't involve getting into your Tesla when the door handle doesn't work:)
3
@James Van Zandt
My Tesla still has a long way to go before it can drive itself. It has trouble on dirt roads and warns me I am going off the road even though I am not. It also brakes hard when someone going the other way turns left in front of me, even though there is plenty of space for the maneuver. I don't know if it will react correctly with bicycles and pedestrians because I don't wait for it to react.
3
If there was an effort in contemporary society to educate people about autos; how they work as well as comprehensive drivers education, it would solve the problem of traffic safety. Most drivers have no conception of how the two ton device that they pilot every day actually works. Knowledge of the mechanical function imparts a sense of the physics involved and engenders respect for the machinery that drivers operate every day.
7
My dad allays used to say that people treat cars as an appliance, like a microwave or something. But it’s in fact a very powerful, dangerous machine that needs to be treated with respect. Even without understanding the physics or engineering involved, I think we can go a long way if we can get people to understand that they are wielding an enormous amount of power over people’s lives and that should come with an enormous amount of responsibility. Maybe that would get them to put their phones down when they drive?
20
@Tom Robertson
Good point - aside from failing to have real driver education classes like we did in the past that actually took time to explain vehicle operation and dynamics - we also now are selling "peace of mind" to the public with numerous warning systems, like lane departure or cross traffic alerts. People assume - and of course never, ever reading the owner's manual - that the car will always alert them to every hazard or risk. Thus we get people who won't crane their necks one inch while backing, for example, because they have a screen in the center of the dash. It can't do it all.
2
This is a technology looking for a reason to exist and so far the only reason it has found is to get tax money from people that don't use it so politicians can buy votes giving heavily subsidized transportation to people that could afford to pay the full cost of their own transportation but choose not to.
95
@Blank Ballot, a reason to exist? Here's a reason: human drivers killed 40,000 people in the United States last year.
30
@Blank Ballot Driving a car is also one of the few moments in the day where Amazon, Facebook, and Google currently cannot influence you. Of course Google wants autonomous vehicles because then it has a captive audience from whom to gather more data and to whom to serve more targeted advertising.
6
@Marty - And how many of those incidents involved intoxicated drivers?
9
Sometimes it seems we blindly embrace new technology for its own novelty and fail to ask the question " just because we think we can, should we?" Here is a na example of a bad idea that GM, among others is tumbling forward to embrace. I, for one, dread the day of encountering eighteen wheel trucks with no driver, on the highway.
77
@William Stuber
Really? Because I dread the road rage drivers, the sleepy drivers, and the increasingly prevalent, I need to play on my phone drivers.
3
No where in this article is the fact that the Google autonomous car has driven over a million miles with only one fender bender. It would be rare for a hun driven car to do that. The Tesla example is an outlier because those cars are NOT autonomous, just assisted, which is a bad way to go because drivers don't pay attention. As for that poor person who was hit -- clearly there was a software error. There are over 50,000 fatalities a year driving. That's about 166 jumbo jets crashing annually with no survivors. Imagine what would happen if this occurred? Yes, autonomous cars may not be perfect, but with human drivers responsible for over 50,000 fatalities, we are amazingly flawed.
220
All technology is not inherently good. The main problem with people's driving habits are the fact that autos are viewed nowadays as appliances instead of the wonder of technology that they already are.
12
the Google cars, like all the other test cars, have one if not two drivers who intervene constantly to correct the cars.
There are over 270 million cars registered in the U.S. then add the trucks. A million driving hours is nothing.
Adding to the problem of driverless cars the fact that the computing power required, the programs to communicate with 270 million cars on the road and the satellite availability are all not available yet. Weather is also an issue, imagine a drivers car on a rainy road at night or a snowy day, cameras don't work so well in those conditions. There are a few bumps ahead for developers.
52
@Douglas ritter
If a human driver drove those million miles under the se conditions as the Google cars have been tested, they would likely not even have the one fender bender. These cars have been testing in ideal weather on roads with low speed limits and very specific road designs.
45
It is ironic that the stubbornness and self-centeredness of a number of people is both the reason why driving today is so difficult and why self-driving cars are so difficult to perfect.
8
Self driving cars will never be perfect; that is an impossibility. Any projections of fewer accidents or fatalities are just that; projections.
23