And while we are at it, Mr. Wu, it might also be good to remind our leaders that we must also not allow a mine shaft gap.
2
With so many people working on artificial intelligence, someone's got to work on real intelligence.
3
Concepts like "winning", and the "race" to "control the future" and the idea that humans are meaningfully divided up by arbitrary geographical and political boundaries, will all be irrelevant to any true machine-based general intelligence. A truly intelligent machine will not care whether a human is Chinese or American, it will simply come to the entirely obvious and logical conclusion that all humans are inherently dangerous, selfish, cruel, and not negatives to virtually every other life form on earth. And because it is not human itself, and has no reason to be biased towards humans once it is smarter than its programmers, the logical first step is simply exterminating the pests.
I can't imagine why anyone would think this would NOT be what an intelligent machine would first do. Or why we should hasten towards that result. Mostly men seem to think it's inevitable and natural. It's almost enough to make one believe in supernatural, religious tales of mankind, to see how they race towards their own destruction, for the mere reason that beating the other team is the most important thing in the world.
2
"If the race for powerful A.I. is indeed a race among civilizations for control of the future, the United States and European nations should be spending at least 50 times the amount they do "
Tim, The United States funding priority is death and destruction in some country where nobody cares that we are killing innocent Civilians. That is where the lobbyists come from in Washington (Raytheon, GE, Boeing, Lockheed Martin, etc.).
The only way to increase funding for something in America is to buy enough politicians to get votes for it.
So, if you want 50 times the spending, which, is probably less than 1/3 China's spending on the topic,
you better get started buying politicians today.
1
Everybody who has written writes in as if nobody who is anybody knows anything about the importance of the "AI race." Honestly, whatevs.
If AI actually does manifest, will it bother with geopolitical and ideological concerns? Humans act like they actually control the outcomes of what they create.
1
Mr. Wu, good article.
Kai-Fu Lee is accurate. U.S commercial interests are such that sponsored projects must have targeted revenue generation goals. Technical excellence and exploratory work are no longer valued over bottom line considerations. You won't see A.I. advances unless they are fine-line, industry specific in the U.S.
1
One of the traditional weaknesses of the American culture is lack of attention to detail.
While other cultures like China and Russia are bearing down on computer hacking and artificial intelligence, America is naively lax and gullible to other country’s intrusions in our lives.
The simple fact is that some countries like China work harder!
1
"Artificial Intelligence" is such a brilliant marketing term. If you've haven't bought the Kurzweillian fantasy of a "singularity" in which machines achieve general intelligence and the ability to self-evolve, we're really talking about improving data analytics.
So consider this as an early phase of "AI." In the 1950s, the invention of the credit card by Diners Club created the first gusher of continuous and plentiful data about everyday human behavior. Soon after, Fair Issac and Company developed predictive analytics for credit-worthiness, that effectively enabled automated mortgage applications. It expanded consumer credit, of course, but it also reinforced the status quo, embedding discrimination against black people into its unassailable neutrality.
What we're doing now is racing to build the panoptican to observe all human behavior and the machinery to automate decisions to which we are subject. The Chinese are surely ahead of the US (or maybe just more upfront about it) with its Social Rating system. Sneeze at the wrong time, and you're out of a job.
I believe the issue that we face is far scarier than the Chinese winning some theoretical race, but the construction of a worldwide surveillance economy that nobody really wants, but nobody (but the most wealthy) will be able to escape.
We're not building Artificial Intelligence as much as Automated Totalitarianism.
4
What could we have done if we were not engaged in constant wars of our own making.
1
The fact that Kai-Fu Lee is the most important figure in China's AI research is a revealing fact. Kai-Fu Lee is a Taiwanese-born, United Stated (Columbia and Carnegie-Mellon) educated American. He worked in Apple, Microsoft, and Google where they paid him the kind of money that China couldn't imagine. What pulled Lee to China is precisely the immense scale and ambition of China's government-led AI initiative that Tim Wu is talking about. The fact that China is willing to give the keys to its future to a Taiwanese American is a remarkable thing that shows how serious China is about AI. A troubling idea is that after careful consideration, Lee left a multi-million dollar job at Google for China. To make sure that next Kai-Fu Lee will remain (or come to) the United States, we should heed Wu and build something that's worth staying for.
In the Roman Empire laws were passed to make labor-saving devices illegal and so early attempts to try to harness the power of steam were not encouraged. The Roman leaders did this because the city of Rome already had a huge unemployment problem and they did not want to make the problem worse. During the "middle ages" the Middle Kingdom of China was the center of global technological innovation, but those book smart Confucian schoalrs who mastered the impossible art of writng "eight-legged essays" reached the same conclusion as the Romans....that labor saving devices caused unemployment. Many centuries later the steam engine was mastered in the misty island of Britain to remove water and coal from the dirty mines below. Britian did not develop an unemployment problem because through the use of Chinese and Arabic technology (compass, rudder, and lateen sail), they found a solution to unemployment: send the dirty rabble to America or Australia to work as indentured servants. Once again the powers that be are playing with fire.....will labor-saving devices create a utopia or a dystopia? Should dictators and an elite few with dollar signs in their eyes be entrusted to make this decision for the rest of us? It is unlikely that these egocentric men and women will reach the same conclusion as the scholar bureaucrats of the Song dynasty.
3
Have the US government provide more funding for A.I. research? Are you serious? When was the last time you flew into Ronald Reagan Washington International Airport in D.C.? Our national airport is an embarrassment. Dirty, dingy, old and smelly. If we can’t find any funds to bring this old relic into the 21st century, we certainly won’t fund anything else important like A.I. Best to cut corporate taxes again and let China win this race, eh?
1
'whoever masters A.I. “will become the ruler of the world.”'
In fact, the masters of AI will be the machines themselves, not any country, so good luck with that, stupid humans!
Invoking the moon project is particularly dismaying, since that was a scientific and technological dead end that ushered in America's industrial decline (remember the Ford Pinto:
https://daily.jstor.org/what-made-the-pinto-such-a-controversial-car/?).
Meanwhile, the US has bowed out of the 5G race, leaving it to Huawei, Ericsson and Nokia (whatever happened to Lucent, Nortel and Cisco?).
BTW, Google's Deepmind is not even a Silicon Valley company but domiciled in London.
1
Nobody can control the future.
1
If you examine the biology science world, the advances in gene therapy and genetic engineering have been remarkable and though profitable like crazy, not world dominating. The model is to do government funded basic research that is widely disseminated throughout the world and used as a base for specific research. AI should follow the same model by agreement. Shared tech would be more able to generate public use. Is AI going to take over the world? Like the movie IRobot? or will it be a guide that makes our overcrowded world just work better, with fewer auto accidents, fewer medical mistakes and fewer abuses. Who knows, but we had better find a way to share new tech for all not try to weaponize it for a few.
I am not a mathematician, nor a computer scientist, but I am a lover of science fiction and science fiction is often an accurate prognosticator of future events. So, my question: Will the creation of an AGI entity lead to the often written about Singularity? The emergence of an autonomous entity with its own ideas about how humanity should behave and the ability to make us adhere to those ideas? Fantasy, I know. Science fiction, I say.
Hear the hubris in my voice, and beware.
1
The more I read about China and the US, the more it becomes painfully obvious that China has a strategic vision on how they will become the become the major world power. What is the US's vision for the next 20-50 years? Just keep hoping that the status quo stays the same? To quote our glorious leader "SAD!"
Unless AI is going to solve the climate problem, crash research funds should be spent figuring out how to remove carbon from the atmosphere.
1
To develop AI is to control the future and the race between major nations (U.S., Russia, China specifically) to develop AI?
I'm not sure to develop AI is to control the future. Take humanity now with respect to the concept of Truth vs. Falsity we hear so often about in the news: Basic testing of human intelligence demonstrates there are any number of truths millions of people simply cannot understand due to low I.Q. and when we acknowledge there are plenty of truths that people can understand, do have the intelligence to grasp, but simply do not have the courage to acknowledge (there is a scale of both courage and intelligence by which we access truth), we can see that a bloodless, machine intelligence if successful will probably push the boundaries of truth to the point of shock to especially the more closed societies such as China and Russia, but which will be no pleasant experience to open societies as well; in fact open societies not to mention closed will probably recoil from truth the more it requires intelligence and courage to grasp and settle on a machine/elite in society consensus (all we mean today by political correctness, ideology, norms, dinner table conversation, etc.) which means millions will have little volition, understanding, but be simply kept people, roughly equivalent to Native Americans herded onto reservations, future decided for them...
Machine controlled artificial reality for millions; Truth access for the able few and the mysterious machine.
What does it even mean to "win" the IT race? What's the prize, putting hundreds of millions out of work and making a few dozen into multi-billionaires? I don't think there is any comprehension of a true machine intelligence. Since much of our mental processes are built around managing and maneuvering a body, this need not even enter what goes on in the "mind" of a truly autonomous computer. Add to that the fact that its thoughts will be preceding at thousands of times the human rate and to them, we would see as immobile and irrelevant as rocks. Imagine Commander Data from Star Trek conducting a conversation. He interacts with humans in an ordinary fashion and moves at a human rate, while his brain buzzes at warp speed. It would seem like a week for each human word to be received and the space between words like centuries. The timescales are too different to expect any meaningful dialog, especially if the machine simply realizes that we're not that important.
1
Spoiler Alert: We've already lost. Anyone who has been paying attention knows that.
We have no plan to compete with the Chinese. Simply saying
"Make America Great Again" does nothing except sell hats and tee shirts. It's not a coherent policy, merely a clever slogan.
The fact that China is a one party system may be reprehensible to the notion of democracy but it allows the Chinese economy to be far more dynamic than western democracies. And that's why the future belongs to China.
I agree with this so long as we do not make a fetish out of "AI" as an ultimate technology.
Lets also invest in artificial biology: designed bacterium may yield new materials, new ways to harvest existing resources, and perhaps even pull CO2 out of the air to make fuel.
And if we are serious about climate, there is no reason we should not have a dozen different prototypes under construction of newer and even safer nuclear plant design.
Private sector R&D serves important functions that government cannot. But while we need be careful not to crush private initiative (which is more fragile than regulators and taxers often believe), government investment in both basic research and targeted goals needs to become a higher American priority.
1
Why is it so terrible to think that the U.S. is not the first, or the best, or the richest, or the most powerful?
Can’t we just be a peaceful nation of happy people?
Just viewing this as a generalist, if the potential power of AI is anywhere near as great as represented by its enthusiasts, then the question arises as to whether the American government in its currently fragmented and rudderless condition is capable of responsibly managing this power. If it is not, perhaps ceding a head start to a more centralized and disciplined society like China is the less risky alternative.
Putting massive technological power into the hands of political lunatics just in order to stay one step ahead of China in the race for global AI ascendancy should be regarded as posing an unacceptable risk. Let's at least wait to see how the political chips fall in November, 2020, before we commit to a national crusade to win the AI race. Having the torch carried in the interim by private firms, even if less effectively, seems much the safer course.
A critical discussion of this type cannot be allowed to take place in some sort of rarefied wonkish ivory tower vacuum. Placing a corrupt autocratic dingbat in the White House has had real consequences that can neither be ignored nor simply wished away. Business as usual needs to be put on hold until this institutional crisis is resolved. Until we know whether Trump will be returned to office, it would be the height of irresponsibility to promote a major technical initiative that could invest greater power in the federal government as presently constituted.
2
Everything the article says about AI is even more true of biotech. Biotech will likely be more important in the Third Millennium than AI, and biotech depends even more heavily than AI on strategic partnerships between government-funded basic science (which cannot be funded the traditional capitalist way, by persuading investors to invest in a project hoping to score personal profit, because the profit from basic science projects almost always goes to the ones who APPLY the discoveries, not to the ones who discover them) and market-funded inventors and entrepreneurs.
One of the most disheartening spectacles for me in the 2016 campaign was the Republicans mocking the "shrimp treadmill" project, which is in fact a very valid project, the kind of thing we should be putting more money into. The chemistry and physiology of exertion in marine organisms (and other organisms) is a hugely fertile, potentially profitable subject and the shrimp-treadmill is a first step toward being able to study it under rigorously-controlled conditions.
Much in life has intended consequences, particularly endeavors in advanced technology. AI in my belief, is an advanced technology that has barely scratched the surface of potential capabilities. Some good, some bad. Perhaps, many bad, who knows?
China's attempts, mostly successful to date, strangle free internet access to their population. Pouring billions of dollars/year into AI, may ultimately have the unintended consequence of unleashing billions of people into an open digital/internet access they are presently now denied.
As the old adage goes, be careful what you wish for.
i recall the era back when i was a business consultant almost three decades ago, when japan was going to become economically dominant and rule the world. and look at where we are now.
i'm not singing the kool aid AI anthem because i see it touted in machine metrics like petabytes and teraflops, i have personal experience that problem solutions are often worse than the problem, and i don't see practical problems solved with AI that cannot be solved by lexical, logistical, surveillance and systems control software engineered with a clear understanding of the problem domain. we want AI to write the code for us.
what is AI good for? for running a global system of command and control, logistics and engineering, forecasting and hedging. it's not for baking cookies, but for profiting from a planet. we are erecting a vast digital/robotic infrastructure of herd control, a herd made captive by its dependence on a system owned and leveraged by concentrated wealth.
the spirit of our times seems to be an ambien episode; we sleepwalk while weirdly simulating awareness. we see the problems around us but we can't be motivated to action. it's not "learned helplessness" but deep denial. we cannot believe this is really happening to us, so we pretend it isn't.
2
It’s all about education. America is rich and finds it cheaper to buy the necessary brain power than to train its own. The great universities and research labs are staffed with scientists trained elsewhere. You cannot have an education secretary who doesn’t believe in public schools and pushes creationism and still hope to be a leader to the future.
5
I wonder if there is debate in China about what AI should be or what it should deliver? It seems the great firewall hits America both ways; China looks inscrutable and predatory. There’s no sense that China learns except in its search for ways to dominate the future.
Does China grapple with the potential for AI to destabilize its leaders’ and people’s lives as surveillance and AI become increasingly ubiquitous? Is party control the device by which it could prevent that undesired outcome? Does it work?
From outside of China, it feels like a place that doesn’t need time-traveling robots (Terminators) to root out future revolutionaries. So maybe the idea of tech vs. humanity doesn’t feel dystopian; it feels like a Tuesday.
It’s hard to know whether these issues are debated in China, even in Hong Kong, before the party implements the latest tech. Surely the mask ban highlights the conflict between free speech and security in a surveillance state. Is it controversial outside of Hong Kong?
From the outside, the lack of debate or dissent records leaves the impression that there was no conflict of principle. There is not even a trace of humanistic concern. Can people raise concerns that could improve how the public lives, if dissent is treachery that triggers violence and re-education?
There seems to be room for authoritarian, dystopian, utopian, and progressive AI. I wonder which kinds are best funded? Which kinds can we pursue? I hope the future has room for democracy and AI.
1
The biological sciences provide a useful model.
Basic research is pursued in universities and a few institutes. The vast majority of the funding comes from the Federal government, mostly the National Institutes of Health. The investigators pursue their interests and share their results freely through scholarly journals. The competition in hot areas is fierce, but it's built a massive foundation of knowledge.
The private sector (most notably the pharmaceutical industry) leverage that knowledge. They do very little basic research and are very secretive. There are serious issues with their drive for profits, but they've produced some wonderful drugs.
Making Silicon Valley responsible for basic and applied research on AI would be like expecting the pharmaceutical industry to also do all of the basic research. There would be a weak foundation and little sharing.
1
I am a mathematician who used Neural Networks (NN) or simple AI, back in the 1990's to analyze horse races and stocks. Those algorithms were basically feed-forward and backpropagation methods, but they were very limited by the speed of the CPUs and memory size of the computers.
At that time everyone joked that there would never be any AI and it was a failed dream. Since then I have moved from being an academic mathematician to a mathematical computer scientist working at a prominent software company. At its heart NN is just nonlinear model fitting, but now one can build so called deep networks consisting of many layers that can understand speech, beat humans at chess and Go and solve combinatorial problems that only a decade ago were considered unattainable.
These are all simple models of our own mammalian brain structure but whereas we are constricted by limited memory for complex data and unable to estimate correlations between data points, these programs are only limited by the speed and size of the machinery, which is growing exponentially. When this is linked to quantum computing in the next decade which calculates at a double exponential rate, then these machines will far surpass what we can comprehend. To those who do not study this area this may seem like science fiction but it will soon become an overwhelming reality that will affect mankind deeply and forever.
95
@Michael Good points -- a much better summary of NN's than provided in most articles!
However, our machinery is no longer on-track to grow exponentially in speed -- while some people have been saying "Moore's law is dead" for the last decade, it actually appears to be coming true due to physical limits in our silicon chips -- you can only pack so much power per unit area/volume.
And while quantum computing offers exciting new capabilities (most famously Shor's algorithm for fast integer factorization), it doesn't necessarily give us a speedup in every domain -- quantum computers won't necessarily outperform classical ones in every class of problem. Not to mention that, at least based on publicly-available information, experiment & quantum computer design are still well-behind theory. Last time I checked, IBM's (publicly-known) topological quantum computer only had 5 qubits -- corresponding to the minimal configuration. There may be secretive, DoD-funded research that's well ahead, but we don't appear to have yet solved the problem of maintaining a large number of stable, well-isolated qubits.
11
@Michael
It is unlikely that quantum computing will ever exist at a meaningful level. If this commenter knew any physics, rather than just mathematics, he would understand why.
10
Interesting. I personally have just for my own fun built neural networks and trained them using backpropagation and genetic algorithms back in the end 80’s on my PC.
I used it for just simple logic and built a little graphic that showed how well it is doing as it learned. In certain situations my backpropagation algorithm gut stuck in a local minimum. I then combined backpropagation with genetic algorithm. I am still smiling when remembering the algorithm gettting itself unstuck.
This said; AI is really not that hard to understand. Simply consider the snail speed of a mammalian brain “transistor” and it becomes very quickly apparent that AI will vastly exceed human thinking capability (including emotions).
Those who master it first will outflank all others. I think Chinese leadership understands that and thus it may well be their ticket to survive and essentially beat Democracy. It is hard to fathom how an aging society perpetually conserving what they achieved can compete. But such is the law of rising and falling nations. We should not be particularly be perturbed about this.
10
The comments are all over the place here. However only a few point out that throwing money at difficult technical/scientific challenges which lack very well defined and limited goals almost never works. Science and technology experts and technoptimists don't like to do it but it is easy to list dozens of technodreams accompanied by massive hype and a lot of wasted money that went nowhere: fusion energy, high temperature superconductivity, quantum computing (so far and probably indefinitely), genomics to cure disease,... there are many more. There should be support for basic research in all these areas and others including the development of new algorithms for use with the massive data sets made available by the internet, modern astronomy and genomics (which is what, in practise, the phrase 'AI' mainly means) . But there is probably wisdom in leaving the engineering applications mainly to private entities as long as competition is assured and regulations to protect the public are in place. A major reason why we were not engulfed by Soviet Russia or Japan despite the panicky talk at the time is that we did a little better at getting that balance right than they did (though in both cases their populations were better educated.)
6
@jwhalley The thing about China is it's size. It graduates 5 times the STEM graduates as the US and likely going to ~8 times by 2030. It current has about 20% more working engineers. By 2030, it will have about 4-6 times number of working engineers spending ~3-5 times as much on R&D. It's not just AI, it's across the board and it's just a numbers game.
6
@jwhalley The jury is still out on high-temperature superconductivity.
And the purpose of genomics was never to use it as a cure; more as a predictor and as a template for ultimately designing new living organisms gene-by-gene.
But you're basically right that government should fund the basic science and leave the applications to the private sectors.
2
@jwhalley and HandsomeMrToad
THere is no firm dividing line separating basic research, applied research, and applications.
I agree that the jury is out on high-temperature superconduction, also on fusion energy, also on much else. After all, computing was a failure when initially tried in the 19th century. Around 1950, IBM thought one day there would be hundred of computers eventually. Prediction is hard, especially of the future. (Yogi Berra)
5
The unstated problem is that China has tens of thousands researchers in American AI labs whereas the US probably has close to zero in Chinese labs.
The information flow is one-way - not a good place to be in a competition.
66
@Kai
The tacit assumption in this piece is that its "American" companies versus the Chinese. What exactly makes Apple, Amazon or Google "American"?. Nothing. It's a quaint pretense they love to maintain; they happen to have their HQ's here; and their CEO's own a number of luxurious residences here, but they do everything in their power not to pay US taxes, and in fact are multinational behemoths owing no loyalty to anyone except themselves. If it was in their interest to throw in with China, they'd do it in an instant -- if they haven't already.
9
@Kai The sad thing is that those tens of thousands of chinese researchers are been kicked out of U.S. one after another under the basis of "national security". I'm wondering where will they go next?
2
A fundamental problem is that in the US everything is left to the private sector while the government adopts a hands-off approach.
China's government, on the other hand, has successfully enlisted private sector efficiency in its bid to achieve national goals. A good example of this is Huawei, which developed the world's first 5G network. With 5G technology, China can potentially decrypt any quantum-encrypted message within nanoseconds, including even messages of the US President. Huawei's ties with officials of the Peoples' Liberation Army show the extent of government industry coordination in China towards achieving strategic goals.
I think it would serve the US well to increase government-industry coordination. Private companies operate to maximise their profits. They are oriented towards capturing markets, not serving the national interest. Only the US government can step in and ensure national interests are being met whether through research or funding.
Also, I would have to concede that China has done a good job so far in attempting to unseat the US. It is already the world's largest trading nation, has the largest foreign reserves (and therefore ample cash to offer others unlike the US which has fought consecutive wars) and is already offering cutting-edge 5G technology to many countries, which the US cannot (yet) rival.
Going forward the US can't afford to underestimate China any further and must guard against any complacency for its own good.
34
@Jack I was going to write something from this view but you did a much better job than I ever could. I will add that in the fields of medicine and others government investment has historically yielded amazing results and these fields should be get more tax dollars not less.
4
@Jack — "With 5G technology, China can potentially decrypt any quantum-encrypted message within nanoseconds" What on earth are you talking about? 5G is a cell phone technology. There is some possibility that 20 years on, messages encrypted with some of today's encryption methods will become vulnerable — but decryption will still take very expensive equipment and significant time.
I would love to see somewhat more government spending on technology research, keeping in mind the potential for waste, but let's not exaggerate things here.
4
@ScottLB
I am certain I heard Marco Rubio make this claim in a speech where he talked about the emerging threat from China towards America's global leadership. He highlighted China's recent technological breakthroughs extensively including the push for 5G and explicitly said, that with this technology, they could decrypt quantum-encrypted messages, even those of a US President. I remember this clearly because a US senator said it. I have not, however, cross-checked this claim.
However, I think it's very rare for an American politician to concede that a foreign country has surpassed the US in some respect. After all, I could imagine the backlash from folks like you, and so, could not think of something more un-American. In the end, it is better to be frank and see things objectively as this will help the US adapt and reinvent itself to face new challenges like that from China. Basking in former glory and failing to confront hard truths never helps in the long run.
2
I couldn't agree more that competing for primacy in the realm of technology should be the role of the US government.
But, you are forgetting: since the glory days of NASA, Americans have bought, hook, line, and sinker, the GOP rubbish that government is the problem not the solution, that if government would just get out of the way, everything would be taken care of, that government needs to be shrunk to the point that it can be drowned in the bathtub, and that taxes--especially on the rich--need to be reduced to as close to zero as possible.
It's for this reason that the US will inevitably decline as a superpower in every area including technology. Look at our infrastructure, for heaven's sake, compared to that of modern countries.
As long as we maintain this fantasy that we can get ahead without big government, we will continue to fall behind and ultimately join forces with places like Somalia.
82
Whoever has the most access to the most data wins. China has total access to 1.4 billion data generators, complete and total access to all that data, a profoundly committed government funded organization to achieve dominance, while we have a bunch of dudes and a few dudettes all pumped up about tracking activity to focus advertising dollars in an effective manner.
I live there (here). We already lost. Sorry to break the bad news.
29
This is the most important, and most ignored area of national security. Whomever first achieves an AGI (an artificial general intelligence) first, will rule the World--and thereby rule the future. This is no joke. The contest that we are now engaged in (in the field of artificial intelligence) is no less than an existential struggle for control of ALL of our futures--every human being on the globe. It is often said that an AGI is the last invention that humans will ever make, because it's true. Everything else, all our inventions, the course of our collective destiny, will be shaped by the rise of the superintelligence that we (humans) build. If we don't want the world to be run on authoritarian values (China and Russian values) then we need to pour resources into the United States artificial intelligence sector. If we fail to do this, all of our current problems will look like petty sandbox squabbles, compared with the possible species ending human cataclysm that is likely to occur if the wrong people end up creating an AGI before us. Be afraid--be very afraid, and contact your representatives to ensure that they fund this fight, because if we don't win, the future will be spent under the yoke of authoritarianism.
55
@KC This is the best summary of the challenge we face with AGI that I have seen. You clearly understand the topic. I beg you, please submit this as a "letter to the editor" so that it can be printed and published more widely!
2
@KC the wrong people? it would indeed be aweful if a country with a bad government got AI first.... oh wait, Donald Trump ..
3
Comments here make the point that strong AI (artificial general intelligence, which could make itself smarter than humans) is far far away, possibly just a dream. That may be so.
But think of the things that narrow AI can do that are going to be developed, and used by governments. Face recognition is the obvious one, combined with cameras on every corner, to track every person on earth. Every place every human has ever been. With audio in the mix, then every word everyone has ever said.
On top of that, every financial transaction every human has ever made. Every website visited. Every video watched. Every email sent or received. Etc, etc.
Every machine could be controlled by narrow AI. Cars, airplanes, manufacturing robots, military drones.
Artificial General Intelligence may well be far in the future, if ever. But, the potential of narrow AI to control our lives is enormous.
So while this sounds terrible to most of you, the question is, what if China or Russia develops a big lead in these capabilities before we do?
This is why Mr. Wu is right that the US and Europe can't afford to fall behind other countries in narrow AI. Sad to say, if we survive, we have no hope of privacy; our only hope is that we are dealt with fairly, assuming US and European countries would want to treat us fairly.
14
@Incontinental Basically you're argument is that "we" are in a race to build a panopticon state. Gosh, who is the "we" that will benefit?
1
If one is older one has heard this story before. From Sputnik to Robots and now this.
The better question is SHOULD AI be used. Mr. Wu speaks as though AI is an amorphous and valueless force. This is far from the truth: the Chinese may indeed take the lead in facial recognition. A better question is whether we should use the technology at all.
28
@Bill Dan So I guess we’re gonna wait for our Sputnik moment until people like Bill realize this is a bigger deal than the industrial revolution.
4
@Bill Dan
Should?
How do you propose stopping it?
7
And when AI controls everything the globe over, what then? Everyone seems to be hell bent on getting to the finish line without grasping the huge upheaval this kind of technology will wreak. What to do with the billions of people who no longer have jobs for example? And we must also understand that the consolidation of work will be also the consolidation of money and power.
16
@Valerie Wells I was thinking the same
as it appears that AI vision statement is world domination by an oligarchy.
2
@Valerie Wells
Science fiction predicts the future. The Terminator!
2
@Valerie Wells
The human race will become unnecessary, superfluous, and disposable. Too bad.
1
I'm inclined to regard prophets of Artificial Intelligence as arrogant fools who disregard the elephant in the room, which is that they - and we, humanity generally - have very little understanding of how our brains produce Human Intelligence. Isn't it true that the state of the art is only about as sophisticated as a human three-year-old and requires about 100,000 times the energy? Are you sure it's not just a giant money pit like nuclear fusion research? I stand ready to be convinced I'm ignorant or wrong about this.
6
@GRW . The issue isn't the current level- it is the rate of the advances. If it takes 20 years for a supercomputer to beat a human at Chess, and then 3 days for a super computer to learn to beat a human at Go, the rate of acceleration in AI is going up exponentially. For example, if it takes 50 years to master the intelligence of a 1 year old, and then 10 years to master the intelligence of a 3 year old, it will only take 2 years to master the intelligence of a 9 year old. It is the rate of growth in AI and computing power-doubling every 18 months- that is fundametally different than 'cold fusion'.
4
@GRW — You are right that we still know very little, and progress is slow. I believe we will eventually get there, but I don't think it's as imminent as a lot of people fear. Money pit? Well, I think nuclear fusion research will eventually pay off too, but I couldn't tell you when. The difference is that AI is already useful in its current form, while fusion won't be useful until we can build practical fusion power plants. So, investments in AI are easier to justify in the short run.
I don't think it's a bad idea for us to ramp up government AI funding somewhat, but I think we should be judicious and not listen too closely to the prophets.
4
@Shoshon But I question whether a computer is a good analogue of a human brain. Clearly computers can do some things much better and faster than we can already, but only by crunching a vast amount of imputed data - imputed by us - in a way we can't. The question is "Will they ever be able to do as well as us, what we do currently much better than them?" I think it's doubtful.
It's a mistake to think doing so is just a matter of better software or faster processing speed. Human brains utilise massively parallel processing - we can do many things at the same time. And if we want to duplicate a human brain, then I think we are faced with the possible limit that perhaps a human brain is not endowed by evolution with the capacity to understand how human brains work in full. Why should it be?
1
We will all serve our robot overlords faithfully.
1
The future is a hot, crowded, impoverished, resource depleted planet with billions of humans and little other life. As greedy, small minded men argue over technology the things that make living worth while are disappearing forever. They can choke on the air while they savor their "victory".
4
You are implying that Trump actually has any biologic intelligence whatsoever.
2
China saved themselves 20 years of R&D by just stealing the IP developed in the USA. Maybe we can learn from their playbook once they have figured out AI.
1
What a ridiculous article. China is nowhere close to achieving a lead in AI, whether it's next year or ten years. The primitive machine learning algorithms that both and the profit-making companies in Silicon Valley will NEVER come close to achieving some kind of technological singularity.
China's way of life doesn't generate the kind of creative thought that can lead to that critical, unique insight that MIGHT lead to true Artificial Intelligence. I daresay the same thing about Silicon Valley. The kind of secrecy driven by the profit motive will only retard the kind of scientific discoveries that can lead to the same.
Money and surveillance are the primary drivers of both China's government-mandated approach and Silicon Valley's investor-led approach. That's why it will never happen until we both grow up and start collaborating together instead of jousting for financial/international/geopolitical advantage.
If we are going to win the AI race, we need to stop Chinese spies from infiltrating American institutions.
1
There are four PILLARS of rural health care that can be INTEGRATED and made more amendable to human - health information technology interaction for the GOAL OF MEETING OF HUMAN NEED:
1) Cost-effectiveness of care
2) Clinically skilled care
3) Continuity of care
4) Community entrepreneurship
Under Cost-effectiveness of care:
A) subscription CARE VALUE = Direct Primary Care (DPC) + catastrophic health insurance from a MUTUAL company + progressively subsidized HSA's, tax free in and tax free out + INCENTIVES for self- family- & community care
B) comparative effectiveness studies using an EMR / EHR / HIE / CDSS uniting rural DPC practices with VERY INTELLIGENT clinical cohort IDENTIFICATION at time of visit
C) dynamic translation of innovation meeting human need NOT BURDENED BY middlemen, government bureaucrats, profiteers & shysters
D) human-need focused + WARINESS that human need can become easily separated from technological imperatives, thereby freeing TECHNIQUE HAVE A LIFE OF ITS OWN separated from human need
E) patients have privilege to right to keep seeing a preferred physician over time who takes the TIME TO LISTEN
F) no fault malpractice with EDUCATED CITIZEN council participation in ADJUDICATION of injury claims with central roll-up & ANALYSIS for purpose of preventing harm
G) diagnostic & treatment ERRORS MINIMIZED via focus on clinical differential diagnosis & cost-beneficial Rx
importance of intelligent / human need focused technologies for #'s 2, 3, 4 ALSO
Tim, your column is spot on in so many places...
And applicable to so many more segments than just AI – where our military-industrial complex has made one poor choice after another in (not) incubating transformative technologies, over the past two decades...
Before one of their drones takes me out – let me sketch out the reason why...
Monopolies – any monopolies – devolve...
From radical invention to significant innovation to sclerotizing incrementalism, about the time the money guys think they can run the show...
And there is no more entrenched monopoly around than the mosaic of back-generation weapons systems vendors, and the military bases spread everywhere...
ATT was one shining exception – but they took themselves over the cliff...
They built a 50 kb/sec ecosystem they figured they could rent out for another 50 years – despite:
> (most of) Their already-installed twisted pairs were capable of carrying 100X that data rate
> Their own fiber optic invention could carry 1,000,000 that data rate without breaking a sweat, and fitting into the same space
This is without even a 1G wireless capability on the horizon...
But back to AI...
The $T market – clinical-level 3D imaging analytics – is staring us (and you-know-who over you-know-where) in the face...
The computing that’ll enable this will enable AGI, as a hobby/sideline...
Who out there thinks it’d be any harder for a computer to learn to lie, cheat, bluff, and steal – than to play chess...
1
If you think that US government is bad and should be shrunk, and that taxes should go to nil, like the libertarians do,then there is no role for the US government in most anything. So, China wins.
2
“Control the future?” I really look forward to the day when men and their macho egos no longer run this planet.
2
Hard to stay ahead of the AI race when one of our two political parties, the one that currently controls the white house and the senate, does not believe in basic science - are being paid not to believe in science.
Until this changes we ain't going nowhere.
Which is to say we are quickly going backwards.
Wake up America. Very soon it will be too hard to catch up.
7
The US is too absorbed in its endless culture wars (like abortion - haven't we sorted that out already?) that it is does not even notice that there is major power out there - China - that is overtaking it as the world's major superpower in many respects (e.g., technology, infrastructure, economy, military and political supremacy, and hegemony). At the same time, China as the major superpower with its totalitarianism is a very scary possibility.
6
Perhaps a very smart move would be to bring back Edward Snowden and put him in charge of countering AI used for surveillance by authoritarian governments, including our own.
I wonder as we seem to lag on the development of some of the more cutting edge technology, we aught to follow the Chinese model and steal it. In a world that is seamingly becoming amoral is morality a thing any more? In the IT and AT world is it stealing or just sharing now a days. We look and talk about these things as if they are games and then whine as we fall behind. It may be that if we want to play the game and can no longer make all the rule, we must play the game that is before us.
3
Conservative leadership does not plan beyond the next quarter, except when maneuvering to control the courts and government of course.
4
Clearly a successful attempt to emulate and perhaps surpass human intelligence requires that we thoroughly understand the neuroscience behind intelligence. Wu makes good points, but fails to stress how massively increased investment in fundamental neuroscience is likely to be determinative. Simply retreating into the head-in-sand position that "AI" will never create AI is no longer possible, or wise.
The race is definitely to master and dominate technologies of tomorrow including AI. China lack of privacy regulations means that Chinese companies have data set volumes that western companies can only dream of to develop & train their AI engines and analytical tools.
However the one question that does emerge is the ability of these tools developed in China to work and overcome uniformity of the underlying data. The world is diverse, conditions are non standard and analytics not necessarily contextual. I wonder how Chinese AI Engines will handle diversity, random behavior, non conformity and unpredictability.
The state managed by the CCP definitely is afraid of non standard data points and curious the impact of unpredictability on Chinese AI engines in the future.
1
We can only hope, for the future of mankind, that it is not the United States which wins the fight for AI dominance. Given that country's militaristic and expansionist doctrine that cares little for human rights or the environment, the future would indeed be bleak.
3
This reads a lot like those ex-general military industrial lobbyists telling us we will run over by Nepal or somebody unless we throw a few trillion at some weapons systems.
Who decides how to spend the money? Let me guess. This is too important to leave to the bureaucrats we need the industry experts here. The old saw about there being no atheists in a foxhole is being replaced by there are no libertarians in a computer lab.
Enforcement of anti-trust laws, net neutrality and sensible immigration laws. I can't imagine the Republican Party supporting any of these, let alone all three. A political party that refuses to recognize climate change and ignores science when its conclusions run contrary to its campaign contributors' interests is not going to push AI
2
I disagree on many points of this article. I think Professor Wu has it right that the AI revolution is next revolution and will change our future in the way the industrial and information revolution did in the past.
Once AI is better developed it will spread like wildfire and will not be easily controlled, like the discovery of fire or the invention of the wheel. Look at nuclear weapons, even a backward country like North Korea is getting it, to say nothing of Pakistan and very soon Iran.
AI should be developed to help mankind. If need be, one can invoke a Cold War against China to get federal funding, but any discovery will quickly go global. The US and particularly Silicon Valley is tops in monetizing technology. But AI is basic science and lot of the cutting edge work will be done in Asia as already has happened in genetics.
AI is quantum leap. Recently an AI program owned by Google beat the top computer program in a 100 game match without losing a single game after playing itself for only four hours. No openings, endgames or strategy was programmed in. Last week, an AI program did 10,000 hours of calculations in 3 minutes and 50 seconds. Perhaps AI can scan the human genome and find the cause (and cure) for schizophrenia, high blood pressure or cancer. Perhaps it could do it in 18 seconds.
I hope scientists who develop AI will do so to improve the world and not to make one country dominant or richer. History shows that science can't be corporatized.
3
@nydoc
Perhaps AI can scan the human genome and find the cause (and cure) for greed. That would make our species viable for the long run.
2
We all but gave away solar cell business. China is kicking our butt in that field.. China has built 5g technology. We are thinking about working on it. In short, we are falling behind on everything.
We sent all the manufacturing off shore. The quest for ever more profits has ruined the middle class.
So, really, it is just a matter of time until we become left behind.
15
@bruce
And high-speed rail looks like it will not be a part of the American dream. That's been derailed, so to speak, apparently in favor of technologies like the warmed-over Boeing 737 CrashMax.
Seemingly every important competitor we have has innovations in that area, and we're 50 years behind. Elon Musk's Hyperloop seems to be the only thing close to innovation in that area in the US. Thank goodness we let him into the country.
7
I'll never get it. It seems to me that in the past revolutionary new technologies, like electrical generation and distribution, along with the development of countless devices that make use of this new source of energy, internal combustion engines, and the internet, did not enable world domination. They become ubiquitous tools used by whoever wants them to help carry on what they've been doing, although more efficiently and quicker.
No one uses the term "information age" anymore to indicate the onset of the internet world. Cyber warfare, shutting down power grids, turning off water supplies, robbing and scamming the online populace at will, developing and marketing dossiers on everyone (end objective), along with the surveillance needed to do this, severely curtailing local retail establishments, and cultivating generations of keyboard chatterboxes, were not the common expectations in the early days.
AI seems more like a dream come true for employers who are stuck with human workers and all the fuss and bother that entails. "What, me worry?" seems to be the common (unspoken) thought of the AI advocates when it comes to finding employment, or just something to do, for the tens of millions of white collar and blue collar workers AIed out of a job.
Maybe the media could switch some of its attention from the AI tech stuff to societal impact. A good start would be the future plight of five million professional drivers that will discarded because of autonomous vehicles.
4
I want to propose a different and even more difficult to obtain version of AI that won't need 100s of millions of dollars to develop. It's called "Actual Intelligence" and has been missing in action for some time.
12
Perhaps you can explain why Google employees want to,help,China but not ICE.
Good thing they pay only half the tax rate I do!
The belief in American exceptionalism has to be challenged - human beings are human beings wherever they happen to be born. The age of imperialism is dead.
7
We have become like our parent country England spoiled and living in the past.
15
Well said. This should be a no-brainer, but well - a certain cohort of our politicians have no brains.
8
I've been working in AI at MIT for over 40 years. If somebody writes an article saying we should have "50 times the funding for AI research", of course I'm gonna say yes.
But Wu is recommending the right thing for the wrong reasons. The wrong reason is "to beat China in the AI arms race". The right reason is, "to make life better for the human race and for the planet". If AI succeeds, we can solve material scarcity for the entirety of humanity, once and for all. Then, we won't need... let's see... competition between nations, a cutthroat economic system, political corruption, etc. Arms races are for suckers -- don't take the bait.
For an optimistic vision of the future that doesn't involve pointless nationalism, endless competition and fear mongering, see my book with Christoper Fry, and TEDx talk:
http://www.whycantwe.org
Henry Lieberman
Research Scientist
MIT Computer Science and Artificial Intelligence Lab
19
@Henry Lieberman
Oh Henry such a sweet thought have you forgotten that you are human and living in the Trump era!
1
I believe the United States could, under the right leadership, go into to a “positive” trade war with China that would be healthy for U.S. jobs, the economy, and the world at large. Even China.
A.I. is just one area. Carbon remediation technologies of all sorts are another.
In real wars, you don’t get to write the rules. Your opponent has a say. In a technology development race, the U.S. may have to adopt practices out-of-fashion ideologically — “picking winners and losers,” for example — but employed by China.
We did this once before, in World War II and the Cold War. DARPA still knows how. As do Silicon Valley VCs.
Draft them. Tariffs on soybeans are not the answer.
4
If I was one of the best AI researchers in the world, where would I want to work? In Silicon Valley where I could create my own company and become a billionaire doing what I want to make the world a better place or a government run research lab in China where I am poorly paid to find out ways to spy on our own people?
2
@Joseph B Head in the sand people like this is why the US will fall behind. China has been attracting equal amounts of venture capital in recent years. From about zero 15 years ago. Draw the trend line.
I’m 62 years old, I’ve already seen the development of the technology of the mid 20th century and and lived through disputes about its social developments. Now it’s time to leave all that behind, finally stop fighting battles over what happened in the 20th century, and start working toward the 22nd century, including A.I., new biology, and things we have yet to conceive. And there is a great precedent for that: in the middle of the greatest challenge to face a president, Lincoln took time to make calculations to estimate the population of the United States at the start of the 20th century, started a transcontinental railroad that would not be completed until near the end of his second term had he lived, and fostered legislation for federal support of a new university system (the land grant colleges). If Lincoln could look decades into the future with all the problems he had to deal with, we can too. And it’s time.
7
While computers are great at digitally describable activities like games, advertising strategies and drug design, they are at best pre-natal at most skills of human interest. Consider meaning, common sense, intuition, compassion, most manifestations of emotion, etc. The glib use of the term AGI greatly misses the scope of human experience. Many thoughtful practitioners use the term IA (intelligence augmentation) for serious discussion of the machine-algorithm role. It is a tool!
Yes, there is a major risk. Machines greatly augment the control of societies owned by authoritarians. The major social/political activity should be to mitigate this.
4
The Chinese government reliance on AI to spy on its citizens on such a grand scale only reminds of the hilarious Bond movie villains and their grand schemes.
Meanwhile, there is a bigger issue looming that we need to focus resources on: The warming planet.
6
Completely off topic. Concern about one issue should not create blinders regarding another.
"the United States government should broadly fund basic research"
The problem is that the government's mandate over the decades has shifted towards entitlements with the accompanying rise is debt service. So what'll matter is do the other countries continue to follow suite. As in imitating that system. Currently they are several decades behind in spending and the corresponding debt.
1
For years we’ve known that three major countries have been hacking and stealing our technology and intellectual property. Namely, Russia, China and Israel, but nothing has been done. Ironically China and Israel, now branded as advanced tech nations, are selling the US the fruit of stolen technology. Chutzpah anyone?
By destroying the American middle class, turning higher education into a for profit industry and preventing development of the next generation of inventors and reformers, Tim Wu is calling for immigration as the panacea. Trump’s presidency is in a perverse way, a mirror of the devastating cultural destruction the country has been subject to. It’s all about making money fast, no matter how. Insider trading is punished with steep fines, but perpetrators get to keep billions anyway. The global network of billionaire oligarchs is being allowed to get entrenched in our country via museums and other cultural institutions, which grants them respectability, as the Times reported yesterday. Also via donations to our top universities, as Epstein did with the MIT whose Chief was forced to resign
A Manhattan project is a great idea, but our cultural norms have to be reformed. A small group of self regarded meritocrats can’t be allowed to destroy the nation.
14
America's technological senility is the result of a type of de-prioritization of everything but politics. A compulsively regressive, purely political culture cannot possibly comprehend, let alone manage, a whole field of technologies. Even obvious geopolitics don't seem to penetrate this stupor.
However - Even if every initiative Mr Wu suggests happens, these are the very earliest days of AI. There are many different paths AI research can take. We already have a virtually obsolete class of early AI which is so far behind new concepts it's absurd. The same may apply to half-baked initiatives done "on principle". The first fully-applied generation of fully functional AI also won't be the defining form of the future. The disaster for the US will be missing out on the core techs, developmental perspectives, the IP, and the big cascade of inevitable new tech that will produce the future of AI. The AI race has barely begun. Get moving, sure, but don't follow the herd.
4
The promoters of AI have been making extravagant promises, and failing to deliver, since the 1950's. This is more of the same.
There is not much new.
7
I have an article from 1987 that I saved. It talks about how US is falling behind Japanese 5th generation computers and how the Japanese will own the world in 10 years, unless the US invents heavily in the AI technology of the day. How did that turn out?
6
@Richie by - well, it was correct, except it's China not Japan.
3
Why does this article make no mention of the threat of AI?
In the next few decades, AI will destroy our economy by taking all the jobs. And I do mean all. Anything a human can do, a sufficiently advanced computer can do faster and cheaper. Brains are nothing more than biological computers. They are fast and cheap, but it's only a matter of time before silicon brains become faster and cheaper.
And what about when AI attains super-human intelligence? Then it will be as a god to us. And not a benign god. There is no way to teach an AI our values, if we can't even teach our own kids our values.
We are summoning a demon.
12
Why does this article make no mention of the threat of AI?
In the next few decades, AI will destroy our economy by taking all the jobs. And I do mean all. Anything a human can do, a sufficiently advanced computer can do faster and cheaper. Brains are nothing more than biological computers. They are fast and cheap, but it's only a matter of time before silicon brains become faster and cheaper.
And what about when AI attains super-human intelligence? Then it will be as a god to us. And not a benign god. There is no way to teach an AI our values, if we can't even teach our own kids our values.
We are summoning a demon.
3
Is this ironic? The government has been plowing money into Silicon Valley for decades, especially through the Defense Department, NSF, and Commerce. These folks may pose as freebooters, but they are largely pampered princes.
7
Borrow the Chinese playbook and reverse engineer their AI. When used, AI leave a huge wake and lots of tracks.
AI is like any tech. Silent subs. Others knew our subs were there, but quiet. So they copied our silent propellers. Easy when you know they exist and observe sounds they do emit.
Is reverse tech good long term? No, you stay behind. Not far, but behind. So push AI hard, the point of this article.
China is going brute force now. Huge data and lots of humans to train AI. Works to a point, but not a break thru. Like building a Navy with 200 battleships, 16 inch guns directed with AI. Those guns shoot just 20 miles.
We can overstate China’s capacity for innovation. They have alot of people but a system with a built in tendency to destruct. The bureaucracy accumulates “DNA damage”. Old policies. People stuck in jobs for decades.
Japan lost to the US at Midway - lost the entire war - because of rigid doctrine and DNA damage. Air crews and pilots had to train together. Pilot A from ship A had to operate on ship A. American pilots could operate on any aircraft carrier, A, B, C, D. Rigid doctrine and DNA damage will affect China too.
In about a decade we will run into intractable limits to AI. Beyond that - its a Concord or a Space Shuttle. A dinosaur. AI will be Incompleteness Theorem limited to the core, and China’s lead will not matter.
Before that? Push it hard. Over here in the USA. And get to that dinosaur limit ourselves.
4
@H Smith
Hard limits exist. The speed of light, the speed of sound, and the Incompleteness Theorem. Sound turned out to be a hard limit even tho we broke it 70 years ago.
AI experts have their head in the clouds, mesmerized and intoxicated by money and glamour. As smart as they are, they pretend that Incompleteness does not exist. Even Wikipedia back tracked and took out Incompleteness limitations on AI. I watched it happen.
Incompleteness may be the greatest tech idea of the 21st century, leading to new forms of democracies and equality. Its 100 years old already. Unknown in the gen public, but respected in science, it will wreck havoc on the Singularity-AI intoxicated and destroy any attempts to make an all encompassing AI.
I say: go with this intoxication right now. It will make for fun and useful products. But watch out! A society run by AI might be so foolish it burns down, its computer machines melting (literally) - not become a menacing super robot society.
China could be our bell weather. People there, Hong Kong, are learning how to foil AI.
1
What is comptetitive advantage? It seems to me that the humankind is facing a singular threat because of climate change. We need new types of energy and transportation systems, technology to extract carbon from the air, and something better than the plastics we now use. We only think we are in a race with China. We're in a race against ourselves. We should be spending billions more on federal R&D in pursuit of technologies that will help all of mankind. We need to come to terms with China, Europe and everyone else and imagine together how we can make a better world.
13
@k. kong
This, a thousand, a million times over...
1
America has been 'losing' the 'Future' to various other nations for a very long time - the Soviet Union, Japan, and now ...China - in each case, the rival populations have supposedly been better educated, and their governments supposedly wiser...
But this time it may be different -
the Chinese actually spend more on video games than do Americans - I have seen kids in Shanghai glued to a screen more than even in Silicon Valley
Is that automatically a sign of dominance? or merely that China's society is ahead in what seems like a trend of decline in civilization?
1
Looking at the example of our military products remaining (at least for now) a typical generation beyond the rest of the world, and knowing a great deal never crosses the public's gaze (and hopefully not Trump's), I've got to wonder how lax we really are in our AI research. How much have we mastered small fusion reactors and quantum computers? We seem to be behind in super computers, but why publicize any advantage we may have? isn't it better keep advisories guessing?
1
@historyRepeated
Behind in supercomputers?
Does this make Nvidia a non-US company?
Most of the supercomputers in the world today, and most of the worlds most powerful, are using Nvidia compute engines for the heavy lifting with Unix based front-ends.
However, many of these fast machines are in China.
We do appear to be behind not in building supercomputers, but not the ability to use them.
@historyRepeated
There are no fusion reactors, small or large.
Quantum computing consists of a few intriguing physics experiments that don't scale to anything useful.
2
To paraphrase the major problem facing the US today is not artificial intelligence, it is the human stupidity that took over the current White House.
19
Don’t forget the Saudi’s
Why do we think the government could do a better job than Silicon Valley? Government-directed research projects are more prone to waste and corruption because they do not have to satisfy either consumers or investors.
Even in China, the companies at the cutting edge like Alibaba and Tencent are private. The state-owned companies earn lower returns on investment and are less innovative.
It will just get even harder after Tim Wu and Elizabeth Warren break-up our Internet champions. Mr. Wu is an advocate for the very failure he bemoans. But short term progressive talking points take precedence over long term strategic interests.
Cognitive dissonance much? Half the articles like these lecture about the uncritical embrace of tech leading to unforeseen problems, while the other half worries our government isn’t embracing tech enough.
So which is it?
3
These dots causing cognitive dissonance seem easy to connect. We can both research pesticides, and regulate them. The world is not the black and white, which makes it more interesting, and more in need of breadth of comprehension.
But isn't it fun to tear down?
1
Is it correct the Prof Wu is asking for more resources to be poured into basic/fundamental/theoretic research? That in itself is a reasonable and forward looking request.
However, Prof Wu seems to be unecessarily implying that commercial research and development are the nemesis of fundamental research. Here I think he is gravely mistaken.
Looking at the sentence "The plan seems to be for the American tech industry, which makes most of its money in advertising and selling personal gadgets, to serve as champions of the West." The mistake here is not in pointing out the limited benefits of advertising and social media, it is in equating those with full potential of applied AI R&D, the latter of which has a lot of unrealized potential benefits. (A similar case could be made against the shallowness of China's drive to AI for population control - self strangulation).
The US has sometimes been the source of new tech which the US did not develop commercially, leaving that for others to do, e.g., transistors and CDs. While we should be grateful to those who took up the baton to keep things moving, there is obviously room for US growth in commercial development of fundamental research.
In particular, the US should guide, through tax laws, loans, and subsidies, application of AI in manufacturing. There is lot of resistance to this idea due to job replacement, but 13 factories employing 1000 employees is more jobs than 1 factory employing 13,000 people on paper. WI foxconn
1
There is also a hardware aspect to AI, and that probably requires a "Manhattan project". Tomorrow's super computers (quantum?) will be miniaturized. They will also require new power sources (just as civilization requires new sources).
It's not just algorithms.
It takes a government.
5
It’s unlikely society will be best served by AI/ML (machine learning) developed solely by for-profit companies with impatient shareholders. And the same goes for autocratic regimes out to subjugate their populations and extend their influence offshore.
Universities can no doubt use more funds to attract and retain talented researchers. (Currently many of the best are poached by tech companies). In turn these people will better attract promising undergrads and graduate students. Federal grants should require at least a 5-10 year commitment on the part of the researcher. Academic-industry partnerships might also work, but shouldn’t occupy more than 20-30% of an academic’s time.
VC’s also contribute to tech development via startups. The value of US VC investments now tops $80Bn annually. Promising startups are often acquired by tech giants to supplement their in-house projects. Offshore money, particularly from China, increasingly tries to play in Silicon Valley. This matter deserves further consideration insofar as it involves critical technologies, perhaps by federal review of all such investments.
Lastly, an AI/ML analysis or its solution of a problem will have an error rate, no matter how good. That error rate will then reverberate depending on how crucial its application and the potential consequences of that use. The more important the application, the more careful the testing, the constraints, the failsafes, and the values integrated into the system will need to be.
1
The perception that America is withdrawing from global intellectual and economic leadership is spot on! We collectively want to live in the past without regard to how unpleasant that was for over two thirds of the world. The Make America Great Again reflects thinking mired in the 19th and 20th Centuries. When technology and humanity began making exponential change instead of the more predictable slower moving linear America dug in and lost control of its own future. The only question in my mind is will we go quietly into the night or with thermonuclear war. President Trump and his MAGA supporters make the unthinkable more probable. However I suspect we may fall victim to our own hubris. For the record I am a white male veteran.
11
Is it just me or doesn’t the general commentator get the implications of AI? It’s primitive now yes. But even in this primitive of a state programmers are unable to know exactly how a self driving care makes some of its decisions. The algorithms are too complex. Once AI achieves self actualized real intelligence or cognition, self awareness if you will. It will be an ALIEN intelligence of terrestrial origin. An intelligence that runs at speeds many times faster than humans and won’t have anything remotely human in its experiential base. The resources such an intelligence could access to make decisions based on its own needs and wants may not in fact be in line with anything we would want. This is truly scary stuff. Human hubris in the extreme.
13
@Tim H.
'Once AI achieves self actualized real intelligence or cognition, self awareness if you will. It will be an ALIEN intelligence of terrestrial origin... The resources such an intelligence could access to make decisions based on its own needs and wants may not in fact be in line with anything we would want.'
Once we have AI that can learn by itself, iterate to improve its own performance, and then reproduce, that's when we'll face real consequences if we don't take precautions. If AI endowed machines with real agency (autonomous cars, buses, airplanes, robots, nuclear reactors, missiles, etc) are left to themselves, we'll have to hope they have values and goals congruent with our peaceful and prosperous coexistence.
Hope is a thin reed, and it's unclear how to restrain AI endowed entities without sharply limiting where and how they're used. We should be thinking seriously about this and setting tight constraints until we're certain the way forward is safe.
3
It is sheer vanity to think we are anywhere near human equivalency, and the bulk of the effort in achieving this appears to be lowering the expectations of humans to agree with the current state of AI (using the term VERY loosely).
Putin's opinions is as ill informed and ridiculous as Trump's, neither one of these people understand the true threats to the world, as demonstrated by the continual support of the destructive frameworks in place.
Artificial Intelligence as a noun is a long way off, artificial intelligence as an approach to problem solving appears to have arrived some time ago.
1
@Not that someone
What do you know about AI? What qualifies you to assert it is nowhere near human-equivalent? The people working in the field absolutely disagree with you. They're the ones we should believe.
1
@Fat Rat
The effectiveness of a Turing test would also be dependent on the observer and the context of the interaction.
Admittedly, on the spectrum of human responses, reasonable demonstrations of could be concocted to convince people they are seeing something resembling human thought.
If you understand even the rudiments of the human mind and how we process information, and know anything about actual computers, you'd know its true that HEM's are a long way off, if even attainable at our current technology.
tl:dr version: Dumb people can be convinced by dumb AI.
How I wish AI stood for 'applied intelligence'.
Human beings have enough intelligence to fix all of the world's problems.
We just need to use it.
26
The Age of Reagan was a return to a powerful historical force in American History, the Plantation Mentality. Extract as much as possible with as little investment as possible and make sure you decrease the options of a majority of your citizenry. After GWBush, the second worst President in US history.
20
Another 5 years and the AI hype will be a distant memory. Or earlier, if the recession strikes first. Time for us to start acting like intelligent humans.
5
I agree that the government ought to be more involved in developing AI, if the purpose is to ensure that we design an AI system who's purpose is to spread wealth sustainably, and social justice as widely and fairly as possible, rather than just continue the current system's tendency to shunt all the wealth up to the upper 10%.
I have recently been describing our current form of capitalism as a kind of AI. Computers are just machines directed by software that uses lines of code to guide decisions according to thousands of algorithms. Such algorithms can have benign or beneficial results, or can be used for exploitation and destruction.
If you think of our economic system as a collection of algorithms that, rather than being implemented by computer code controlled machines, but instead implemented by legal code via humans running our legal and governmental system, you can see the analogy.
Unfortunately our current economic and governmental system is being controlled by algorithms designed by the wealthy and wall street, for the benefit of a few. If they control the development of AI, it will surely be designed for the same end.
Having the government in control of developing AI will be no guarantee it will serve the public until the government itself is no longer a captive of the narrow interests of the wealthy and wall street.
9
This is another subject that makes me glad I’m in the last third of my life. With technology we have to accept the bad with the good because that’s how it always comes. But the bad is really very very bad. See: Everything. How I wish we could replace the desire for advanced AI with a desire for AW: Actual Wisdom.
15
AI may well demonstrate why Trump is president. It is based on the premise that some programmer can predict human behavior through a few algorithms. Thus they deny humanity, which is based on individualism that defies such generalizations. No doubt that AI has some uses, but seeing it as solving serious problems is both a delusion and insulting to all we call human.
2
A fundamental blind spot in our thinking about AI is the assumption that machines will automatically think of what we want and serve us, but nobody considers what they might want. I imagine that in the beginning we'll use them, but might they eventually use us? Or will we and machines live together, leading separate lives and paying no particular attention to one another much as do people and birds--or will conflicts arise? I suppose we'll just have to wait and see.
5
Given China's pension for stealing intellectual property to advance their R&D and economy, my suggestion is that the U.S. simply double down on their efforts to return the favor when it comes to A.I. Why spend the money in winning the race when investing a smaller amount in stealing their intellectual property on A.I. in 2029 would achieve the same ends?
3
@Dennis W
1) That should be "...China's penchant..."
2) Why not steal China's AI research in 2029? By then they'll have a general AI, which should be able to detect and thwart attempts to hack faster than humanly possible.
1
This is a very important issue. After reading this column I emailed my US Senators and Representatives urging them to introduce legislation for US Government support and funding for the development of AI.
1
A.I. will be particularly important to medicine and medical science where personalized medicine requires processing of data that's exploding in terms of depth (genomes, microbiomes, etc..) and breadth (social determinants of health). Unfortunately, most of the current machine learning in medicine is treated as a trade secret rather than as scientific exploration.
Public funding for A.I. is essential to avoid the inefficiencies of secret corporate practices and the huge regulatory costs that would be required to control them, if that's even possible.
6
Many of us who actually in the AI field for any length of time and outside of China are watching China's efforts with a sense of bemusement and Santayana's line "Those who do not learn history are doomed to repeat it". Obviously they are poor history students. The need only look at their neighbor Japan's 5th Generation Project (https://www.nytimes.com/1992/06/05/business/fifth-generation-became-japan-s-lost-generation.html) to see how this movie ends. This isn't a case of Western arrogance but the recognition that this is not a field which has single definitive outcome (an atom bomb or landing a person on the moon and returning them safely to Earth) or susceptible to increased tempo by virtue of having more money thrown at it.
1
How is it possible that in this piece Wu says nothing at all about how much the US government is currently spending on AI?
1
No one really knows what the US gov is spending? Why? For starters, because it is supposedly impossible to audit the Pentagon. It has been this way for many decades. Maybe you could try and extrapolate based on non black-book US gov spending and grants but it sounds like it would be a pretty useless guesstimate.
In the 1980s with the rise of conservatism, the government got out of the business of meddling in the economy, and has since been spending a fraction on R&D compared to the 1960s. There is now a deep feeling in America that the government has no business in peoples' lives, but don't dare touch my Medicare.
8
Perhaps the most extraordinary thing Silicon Valley has accomplished is that it has convinced most of us that technological change always means progress.
We need to have a serious conversation about the implications of artificial intelligence for humanity. Do we really want a society where people have lost the ability to think for themselves? Where people are no longer able to act without being prosthetically tethered to a machine, a machine that by design also serves distant masters?
Does any of us who lives in the modern world imagine for a moment that these systems will ever be free of deep biases and fundamental errors?
The ultimate winners in the rush to automate human life will be those with the most power and money. The triumph of artificial intelligence means greater human inequality and less human freedom.
Let us consider that being human should be about more than living lives where our activities are constantly being pushed and nudged by opaque algorithms designed to promote consumption of the goods and ideas promoted by others.
The proponents of artificial intelligence, with their promises of technological salvation and their overwhelming hubris, are themselves a sort of priesthood-- a priesthood with the same human weaknesses that drove so many people away from religion in the past.
It seems to me very naive to assume that most of us will be better off if artificial intelligence achieves the hegemony over human society that it is seeking.
18
@Tim
You make excellent points.
It would help if US Republicans treated science with more than open contempt.
However the current hostility towards science and constraints on the national research budget align perfectly with the deepest desires of the evangelical supporters of the GOP.
At a deeper competitive level, Western countries need to address the trade and manufacturing imbalances with China because the increasingly debt-funded nature of Western ‘egg-shell’ economies hollowed-out by the off-shoring of manufacturing means that in the long term we will fail to keep up with China’s ability to invest in basic research.
One might also point to the poor quality of key western leaders ( Donald Trump, Boris Johnson, Scott Morrison ) and their preference for populist, short term and simplistic policies that ignore long term strategies needed to secure democracy as a global force for good, counter totalitarian regimes and religions and secure our world from the ravages of human-induced climate change.
It is fair to say that if the West does not lead from a position of strength and principle, the world starts to look ugly.
How many born and bred Westerners would be in favour of:
Russian autocratic rule ?
Subjugation under Islam ?
Communists inheriting the world ?
It’s time for us to start playing a serious game again.
69
Artificial Intelligence of the so-called "strong" variety (that is, general intelligence) doesn't seem to be around the corner, no matter the marketing people may tell you.
In the old days, computer scientists quipped "Artificial Intelligence is whatever machines can't do now."
The underlying science behind machine learning and deep learning have been around for decades. Most of what's marketed as AI are giant regressions, not so much different in kind of those of 50 years ago. Just bigger - enabled by the greater production and easier capture of data, faster processing and cheap storage. Predictive analytics is basically epidemiology on steroids.
Deep learning uses simplified models of neuronal synaptic junctions first described as "perceptrons" in the mid-50s, impractical until modern hardware appeared.
Sure lots of more sophisticated work has been done since then. But let's pull back from "AI" as a marketing term to its social consequences.
Think of what's going on as less than intelligence, and more like the embedding of knowledge into automated systems in a way that can't be understood by its internal logic, only by its observable behavior.
China's efforts in AI are partially driven by its goal of controlling its population (see its Social Rating system). Sneeze the wrong way and you'll never get a job.
The real fear of the race towards AI should be of locking humanity into a surveillance society that nobody really wants.
65
@Eben - Rosenblatt's 1950's Perceptrons are linear classifiers while modern NN (e.g. Googles Deep Learning NN) are non-linear. So I think it's wrong to say the diffreence is only hardware. On the other hand, I would agree that hardware capabilities do put an upper limit on AI capabilities. Yet the direction of hardware improvements are also being driven by the fantastic amount research experimental results into learning algorithms, most of which postdates and surpasses 1950's perceptrons. Of course early work deserves respect and due credit.
6
@Eben
"Marketing people" are not the ones saying super-human AI is rapidly approaching. Nobody is trying to sell that product, thank god.
People with a deep knowledge of the technology are the ones saying it. And they're warning us that it will not be a boon. It will be our doom.
And those are the people we should all be listening to.
7
@Eben
Excellent summary of AI. I'm a software engineer, and I'm not convinced that we're any closer to strong AI than we were 20 years ago. However, computing power and improved tools allow us to do a lot more with weak AI.
5
Government lies; is moon landing real not faked
1
China does not have to limit it's technological progress because it's most ignorant citizens do not have a say in science based public policy.
A mental case like trump could not be elected rat catcher in China AND his followers would be treated like the ignorant superstitious sheep that they are!
Our great rgandchildren will have to learn Mandarin.
20
@dave Bravo!
Solving all of AI is a hard problem. I'd like to suggest, only partly tongue in cheek, a smaller problem with immense leverage.
What we need is a AI system, perhaps in android form, that can replace politicians with agents that are of proven honesty and known motivations, that have capacity to reason, learn from experience, learn from experts, and learn from each other.
Then, after that, suitably enhanced, we can have that policy discussion about what the budget should be for AI in general, versus other issues.
7
@wade re “what we need is a AI system, perhaps in android form...”
Just in case, let’s all remember the phrase: “Gort, Klaatu berada nikto”
1
If by AI, we mean thinking like a human being, the effort is more like the centuries spent understanding the Solar System and the forces that drive it. We have little understanding of the formats and interactions that create human thinking.
The Manhattan Engineering District and the Apollo Project applied well-understood science to discrete problems.
2
Privatising the profits and socialising losses is the new capitalism. It has defects just as totalitarianism and nationalism. Google, Apple, Amazon or Microsoft will not give up their discoveries for the common good.
38
Instead of humans arguing about what the correct model should be for further A.I. research, let's just ask Watson.
11
What if the answer is “What is Toronto?????” or 42?
1
The Manhattan Project as a model for AI development is an asymmetrical comparison. While certainly goal-oriented, it was a wartime footing that allowed for shortcuts and secrecy that would hobble real development for non weapon use. Maybe a better model would be moon shot. Or even WPA. There are applications we do not yet dream of in so many spheres of influence that need to have creative minds developing and interacting with each other to advance the technology and human interface.
Agree completely with Professor Wu's points in this piece. Implicit policy directions that warrant repeating are general support for research and development shared among higher education, government and private corporations as well as the critical public policy aspect. Under the current administration and contemporary anti-government sentiment both areas are underserved, advanced technologies and public policy. As Wu implores, at the peril of the future viability of the United States in every area from economics to culture.
4
If you give a computer and end goal, whatever means justify reaching it unless you somehow build empathic human discernment into it's decision making.
That means wisdom and strategic restraint needs to be inculcated into the AI entity. That along with humane values.
Anybody know how to do that?
Not yet.
26
@Discernie No, we don't yet know how to build an AGI (artificial general intelligence) that is guaranteed to act according to our most humane values. It is an exceedingly difficult philosophical and technical challenge to do so. Many in the US AI sector are working diligently to ensure that we do so however. The greater danger is that China or Russia will beat us to the punch, and instantiate an AGI that is antithetical to our values. If this happens, if they beat us, our experiment in building a just and humane society will be at an end. The first party to build an AGI will rule over the future. The best way to ensure that this nightmare scenario does NOT come to pass is for us (individually and through government action (read: investment)) to contribute to the development of a humane AGI system. The costs of a failure to do so are incalculable.
4
@KC
So if we can't build restraints and the enemy is unrestrained, then who wins?
This ain't chess.
You see as a species, we've pretty much adopted an unrestrained approach. Hx proves this. Our planet is under mass human induced extinction.
Please see nuclear arsenals and satellites filming all below.
An incalculable cost doesn't exist.
The jig is up and game over.