A few years ago, he wrote an article explaining that computers could never be good at poker, because that required bluffing, a skill exclusive to humans.
I wonder, couldn't one program a computer to record video and audio of the players you are up against, provide some probabilities of hand strength based on stance, mannerisms, facial expressions, and voice, and add that information into the mix with a betting AI?
That's not even necessary. Using just logic, probabilities and computation, research scientists have been able to beat very strong poker players head to head.
He is wrong but he has a point. Computers are currently good at head's up.
The ultimate test is playing in full table tournament against players with very uneven skillet. If there is one good player in the table, he will be able to recognize and exploit differences.
I agree with you. It is hard to get a computer to perform well many different skillets. Currently, I can only get a really good omellete and pancakes when the skillet dimensions are precisely known.
"[...] machines have algorithms, and they're getting better and better, but machines have no curiosity, no passion, and most importantly, machines don't have purpose."
IMHO the greatest uncertainty regarding AI, and in that sense the greatest risk, is what happens precisely when AI becomes so good that it can be made to have all those features. Intelligence probably is the toughest part of a more global objective that is to emulate whatever the human body and brain can do.
A world where machines can do whatever the human body does is vastly different from ours. It's hard to even imagine it, even though some authors are trying [1]. And some of the possibilities include the end of our biological lineage.
It will probably come from the opposite direction, i.e., integrating more and more cybernetics into our bodies until at some point we're essentially only our brains (biologically). Whether something non-biological can show emergent properties that the brain has remains to be seen.
In any case even AI completely indistinguishable from ourselves is likely only to be as much of a threat as our children are. Everything depends on how it is managed. Apocalptic scenarios generally can only happen in a tightly connected, online society with no offline fallback mechanisms for control. If we are that careless/stupid, then perhaps AI taking over our systems (and hence us) is not so bad after all...?
AI needs creative intelligence, ability to learn and adapt the very process by which it learns, as well as links to the physical world enabling it to control and perpetuate its existence. The entire process is in our (human) control. Until the very end.
The other direction is frankly more scary. Augmented human beings and other existing living beings. That is much more prone to go haywire and pose a threat.
I think both directions are basically going to the same destination. I believe we should worry more about the "artificial human" than about "artificial intelligence". To get to the artificial human there are several paths, biological (with genetic engineering), cybernetics (gradual replacement of body parts with non-organic artefacts), robotics (non-organic from the start) and software (brain emulation).
I believe all those paths eventually lead to the same kind of difficulties in predicting what kind of world they would bring.
Humans have no purpose, at least not objectively. None of these traits is uniquely human nor are they magical. If it's a material property, eventually it will be replicated through engineering.
As one of the first humans to have their job outperformed by a computer I feel Gary has taken it well in stride. I suspect the rest won't be so gracious.
No, in this context we're talking about things digital computers can do better than humans. Doing arithmetic was the first job where digital computers replaced humans, so my comment is quite relevant to the discussion.
He was still paid millions to continue performing his job. If after that match he was "fired" from chess I don't think he would take it in stride at all.
It actually sounded like he was still hurt by it, he was very derogatory about Deepblue, it had no intelligence, just cruching numbers etc. He was finding other faults with it except for the bit that beat him.
The quotes I've heard coming out about Alpha Go (from Go players) seem to be more repectful - that it's actually teaching professional go players new ways of playing.
Yes, but it took 20 years for Kasparov to find that grace. After losing game 2 of the 1997 match to Deep Blue, he accused the IBM team of cheating, by allowing a human grandmaster to override DB at a critical moment in the game. He repeated this bogus claim for many years, before finally coming to his senses.
You could argue that the automatic loom of the 1700s was the first computer that could outperform humans, and the people it replaced were so ungracious that it began the Luddite movement.
Why would anyone listen to him? He doesn't even have neither economical nor technical education. From what I know about him, he just knows how to play one game really well. Basically a sportsman. Why would I listen to a sportsman on an economical topic that is greatly affecting society?
Well because a lot of new and interesting ideas have historically come from people outside the field in question. However, having followed him on social media for a few years, I don't expect anything profound or insightful to come from him. He's basically a useful tool for certain world powers against certain others and will continue to get free marketing from them as long as he stays in line with the narrative.
I haven't read Kasparov's book and the short article doesn't really dive deep into it so I can only comment on the BBC soundbite...
If people are "pushing back against AI", it's not the progress of technology they're against -- it's the economic consequences. People are worried about joblessness and no financial security for retirement.
It's similar to saying "embrace outsourcing because you get cheaper products" or "embrace H1B because America was built on immigrants and their skills".
You can't just speak of those aspirations as general platitudes without being aware of what the real worries are. People aren't xenophobic -- they just want to keep their livelihoods.
If you don't address the commoner's concerns, you'll be perceived as disingenuous.
> If you don't address the commoner's concerns, you'll be perceived as disingenuous.
True but I argue that the current hype about AI replacing jobs has much more to do with media clickbait than any real issue we should be worrying about right now.
If your business is selling ads next to cheap content, it's a great story - scares people into clicks. That to me is a real issue here. How do we prevent the media pandering scare-content like this?
For those old enough remember the "Y2K bug" that was going to have planes falling from the sky on the stoke of midnight 2000 there's a great analysis of the press's role in this at http://www.flatearthnews.net/chapter-one-bug-ate-world ...
> "Encouraged by these stories, some governments had spent fortunes in public money (and secured no better result than those who spent next to nothing). Journalists reported that the British government had spent £396 million on Y2K protection. They also reported that it had spent £430 million. And that it had spent £788 million. The American government had spent far more, they said - $100 billion, or $200 billion, or $320 billion, or $600 billion, or $858 billion, depending on which journalist you were reading. Anyway, it was a lot. Beyond that, the private sector had spawned a mini-industry of companies selling millennium bug kits, while publishers turned out bug books and bug videos, and estate agents sold bug-resistant homes, and a few families sold their houses and fled to remote cabins in order to give themselves a chance to survive the coming bug-related chaos. But this was not a story."
> "The sun rose on January 1 2000 like the lights coming on at an orgy. Everybody who had been so busy - the journalists, the governments, the bug-related businesses and the computer experts - all picked themselves up, hoped nobody was looking and quietly tip-toed away."
> (...) the current hype about AI replacing jobs has much more to do with media clickbait than any real issue we should be worrying about right now.
Hype? For sure. But the current advances in autonomous vehicles will have a very real impact on the working life of the common truck driver (3.5 million in US alone), I think we all agree on that.
I for one don't agree. If the autonomous trucks displace the truck drivers over the span of 50 years, I don't see the problems. Very few current drivers will be in the workforce 50 years from now.
Except that it had already happened. Sure, jobs will be created by low interest rates, but they are all pointless bullshit jobs that frankly we could all live with out. They are just an excuse to work. Thus the opiod crisis, increasing suicide rates, and Trump.
Kasparov says that machines don't have purpose, passion or curiosity. This is typical expression of meat machine privilege.
The purpose of a system is what it does (POSIWID). Feeling purposeful and not feeling purpose is just utility function giving feedback.
Is passion anything else than utility function configured so that the machine does not wander aimlessly and idle? If the machine has lots of freedom to choose its actions and it chooses to avoid 'side quests' and prefers one issue over all others, does it have passion?
How about curiosity? Consider phenomenon that is hard to predict and the machine can't recognize it. Is it curious if it has been programmed to investigate, poke around and learn. When boosting algorithm increases the weight for incorrectly classified instances, isn't that a primitive form of curiosity?
Kasparov also says that AI does not make us obsolete. I assume that obsolete means that if meat unit is removed from it's environment, others (society) don't miss it's contribution. What does 'us' mean? I interpret that as 'AI does not make all of use obsolete.' This is most likely true since for foreseeable future. If not for any other reason than comparative advantage. The oversupply of meat machines has already made many of them undesirable (overpopulation).
Interestingly, Kasparov is considered as a great intellectual in the West, while in Russian-speaking world he is known for support of some utterly controversial theories, e.g. Nosovsky-Fomenko pseudo-historical bunk (see, for example, https://www.youtube.com/watch?v=4Thfip4Owz8 [RUS]).
More pop AI garbage. As a rule, pretty much any article about AI from someone outside of the field of AI or CS in general is going to full of these vacuous, overly broad arguments. Stephen Hawking is guilty of this too. Basically any mention of "AI" and "dystopian" in the same sentence is a red flag for this kind of stuff. I wish that as a website we could just ignore these articles.
No, BBC. Gary Kasparov is not an AI authority. Please don't treat him like one.
You shouldn't judge an article based on who wrote it, whether it's Garry Kasparov or someone "in the field" of AI. Sure, people are more likely to listen to Kasparov because he's an intellectual megalodon. But by dismissing anything not written by an "expert", you're making the exact same mistake.
I did read the article/watch the video and it still is crap though. I agree that we shouldn't make dissemination of knowledge contingent on holding degrees or having accomplishments in that area, but I still have never really seen a good article on AI by someone outside of the field itself.
I really love this article about Fan Hui and AlphaGo, [The Sadness and Beauty of Watching Google's AI Play Go](https://www.wired.com/2016/03/sadness-beauty-watching-google...). It's the first time I've read a balanced intimation of what it can be like interacting with artificial intelligences. I come away from this and the Kasparov article feeling that AIs will be friends as well as partners.
I think your criticism misses the point of this article--it is an article about what it is like to be displaced by AI, not on AI itself. This an interview by perhaps the most prominent and intelligent person to be so publicly displaced by AI. I would submit that AI researchers have zero expertise in what it is like to be displaced by AI. Garry Kasparov on the other hand, graciously and publicly accepted a lifetime of work and his position as the best in the world becoming obsolete. That he came out of that optimistic and interested in continued technological progress, including in AI, is interesting. Also, his observation that humans are explorers who have passion and purpose is a really useful one for (especially non-technical) people trying to find their way in world where AI is a reality.
> But by dismissing anything not written by an "expert", you're making the exact same mistake.
Uh, no. You're simply favouring people who have dedicated their lives to studying and researching something, over people who had a spare month and thought "You know what? I'd love to write an article over something I know absolutely nothing about".
I love to read about amateurs and autodidacts as much as the next person on HN, but the simple truth is that most people talking about a field, who have not majored in it, are not going to be either of those people. Autodidacts who put the effort in are rare these days, what is significantly less rare are people who write articles about a subject, who haven't read even the basic texts on said subject.
This is true even for specialists who are majored in another field, such as: Stephen Hawking, Stephen Wolfram, Elon Musk (Although I admit it is dubious to think of a CEO as a specialist in this sense), and now Garry Kasparov. Rationalwiki has a list dedicated to people who are Nobel prize winners in on field, talking absolute bullshit about another field: http://rationalwiki.org/wiki/Nobel_disease (That is, bullshit in the sense of nonsense, not in the insult sense).
Actually, I would probably go so far as to say that this attitude is strictly against autodidacticism, because where autodidacts favour knowledge about a subject, the other favours arguments from authority and complacence of thought.
You put it better than I did. I would liken the current media frenzy over AI to past media frenzies over quantum mechanics. There's an overabundance of non-experts giving incorrect or overly philosophical opinions on the field. These often lead readers with no background on the subject with incorrect assumptions on the matter (e.g. "quantum mechanics means consciousness is a quantifiable physical phenomenon").
> "Uh, no. You're simply favouring people who have dedicated their lives to studying and researching something, over people who had a spare month and thought "You know what? I'd love to write an article over something I know absolutely nothing about"."
Do you not think the people that have dedicated their lives to studying AI are likely to be more optimistic about its uses than the population at large?
I dislike the 'opinions as news' trend on the whole, but there is room in public debates for non-experts in a specific field. The knowledge required to develop the technology is different from the knowledge needed to be able to engage in debates about its uses.
AI has the power to take jobs away from humans. This is not a controversial view, mainly because it's already happened (such as in car manufacturing). In my opinion, there's not enough debate about the implications of the increasing sophistication of AI (not enough debate amongst the general population, there's plenty in tech circles).
When they list the fears they forgot an important one: discrimination.
Of course, discrimination has always been a problem but I'd like to believe we see some progress in "classical areas" (for instance sexism). The problem with AI is that it is always pigeonholing. As a result myriads of new classes of minorities which are so small that they neither have a name nor a voice will emerge. For instance, you get rated for creditworthiness and somehow you're not typical re: attributes x,y and z. You get a bad rating but you can't really complain as the algorithm probably didn't take things like sex or race (directly) into account. However, it could be that your FB posts are enjoying above-average amounts of likes from people with low education and this may raise red flags etc...
Imagine when AI undoes all positive discrimination, like for minorities, women etc. because it would try to optimize for results, i.e. bringing its own type of meritocracy, going after biggest gains without political affection (survival of the fittest/strongest).
I saw Garry speak in San Francisco -- I'm convinced he's an intellectual giant, but he was very candid in admitting that for all of his trained knowledge in chess, once you take that to an adjacent challenge (the game of Go), he struggles with the basics. I think the same logic should apply with his understanding of AI.
I do think there is a general phenomenon of "smart people outside of AI saying dumb things about AI" -- I'm guessing it's because of the massive implications artificial intelligence is going to have on every part of society.
It mostly covers his experiences with Deep Blue. Wish it was a longer talk. He generally has pretty positive view on computer and AI.
One interesting bit is when Demis Hassabis questioned if a chess-trained AlphaGo can be stronger at chess against Stockfish. Would be fun if they really try it.
PS: Hassabis once reached master standard at the age of 13 with an Elo rating of 2300 (at the time the second highest rated player in the world Under-14 after Judit Polgár who had a rating of 2335) and captained many of the England junior chess teams.
60 comments
[ 3.4 ms ] story [ 133 ms ] threadhttp://www.wired.com/2017/02/libratus
The ultimate test is playing in full table tournament against players with very uneven skillet. If there is one good player in the table, he will be able to recognize and exploit differences.
EDIT - Jokes of this calibre deserve downvotes.
IMHO the greatest uncertainty regarding AI, and in that sense the greatest risk, is what happens precisely when AI becomes so good that it can be made to have all those features. Intelligence probably is the toughest part of a more global objective that is to emulate whatever the human body and brain can do.
A world where machines can do whatever the human body does is vastly different from ours. It's hard to even imagine it, even though some authors are trying [1]. And some of the possibilities include the end of our biological lineage.
1. http://ageofem.com/
In any case even AI completely indistinguishable from ourselves is likely only to be as much of a threat as our children are. Everything depends on how it is managed. Apocalptic scenarios generally can only happen in a tightly connected, online society with no offline fallback mechanisms for control. If we are that careless/stupid, then perhaps AI taking over our systems (and hence us) is not so bad after all...?
AI needs creative intelligence, ability to learn and adapt the very process by which it learns, as well as links to the physical world enabling it to control and perpetuate its existence. The entire process is in our (human) control. Until the very end.
The other direction is frankly more scary. Augmented human beings and other existing living beings. That is much more prone to go haywire and pose a threat.
1) When AI is able to mimic humans in terms of creativity, passion, emotions, etc
2) When we are able to biologically integrate compute resources into the human body
Future scenarios might be vastly different depending on which of these two technical research projects wins the race.
I believe all those paths eventually lead to the same kind of difficulties in predicting what kind of world they would bring.
Perhaps relevant here is that "computer" was a term used to describe people who did computations.
https://en.wikipedia.org/wiki/Human_computer
The quotes I've heard coming out about Alpha Go (from Go players) seem to be more repectful - that it's actually teaching professional go players new ways of playing.
If people are "pushing back against AI", it's not the progress of technology they're against -- it's the economic consequences. People are worried about joblessness and no financial security for retirement.
It's similar to saying "embrace outsourcing because you get cheaper products" or "embrace H1B because America was built on immigrants and their skills".
You can't just speak of those aspirations as general platitudes without being aware of what the real worries are. People aren't xenophobic -- they just want to keep their livelihoods.
If you don't address the commoner's concerns, you'll be perceived as disingenuous.
or just, quite rightly, ignorant to economic causation.
True but I argue that the current hype about AI replacing jobs has much more to do with media clickbait than any real issue we should be worrying about right now.
If your business is selling ads next to cheap content, it's a great story - scares people into clicks. That to me is a real issue here. How do we prevent the media pandering scare-content like this?
For those old enough remember the "Y2K bug" that was going to have planes falling from the sky on the stoke of midnight 2000 there's a great analysis of the press's role in this at http://www.flatearthnews.net/chapter-one-bug-ate-world ...
> "Encouraged by these stories, some governments had spent fortunes in public money (and secured no better result than those who spent next to nothing). Journalists reported that the British government had spent £396 million on Y2K protection. They also reported that it had spent £430 million. And that it had spent £788 million. The American government had spent far more, they said - $100 billion, or $200 billion, or $320 billion, or $600 billion, or $858 billion, depending on which journalist you were reading. Anyway, it was a lot. Beyond that, the private sector had spawned a mini-industry of companies selling millennium bug kits, while publishers turned out bug books and bug videos, and estate agents sold bug-resistant homes, and a few families sold their houses and fled to remote cabins in order to give themselves a chance to survive the coming bug-related chaos. But this was not a story."
> "The sun rose on January 1 2000 like the lights coming on at an orgy. Everybody who had been so busy - the journalists, the governments, the bug-related businesses and the computer experts - all picked themselves up, hoped nobody was looking and quietly tip-toed away."
Hype? For sure. But the current advances in autonomous vehicles will have a very real impact on the working life of the common truck driver (3.5 million in US alone), I think we all agree on that.
The purpose of a system is what it does (POSIWID). Feeling purposeful and not feeling purpose is just utility function giving feedback.
Is passion anything else than utility function configured so that the machine does not wander aimlessly and idle? If the machine has lots of freedom to choose its actions and it chooses to avoid 'side quests' and prefers one issue over all others, does it have passion?
How about curiosity? Consider phenomenon that is hard to predict and the machine can't recognize it. Is it curious if it has been programmed to investigate, poke around and learn. When boosting algorithm increases the weight for incorrectly classified instances, isn't that a primitive form of curiosity?
Kasparov also says that AI does not make us obsolete. I assume that obsolete means that if meat unit is removed from it's environment, others (society) don't miss it's contribution. What does 'us' mean? I interpret that as 'AI does not make all of use obsolete.' This is most likely true since for foreseeable future. If not for any other reason than comparative advantage. The oversupply of meat machines has already made many of them undesirable (overpopulation).
No, BBC. Gary Kasparov is not an AI authority. Please don't treat him like one.
Uh, no. You're simply favouring people who have dedicated their lives to studying and researching something, over people who had a spare month and thought "You know what? I'd love to write an article over something I know absolutely nothing about".
I love to read about amateurs and autodidacts as much as the next person on HN, but the simple truth is that most people talking about a field, who have not majored in it, are not going to be either of those people. Autodidacts who put the effort in are rare these days, what is significantly less rare are people who write articles about a subject, who haven't read even the basic texts on said subject.
This is true even for specialists who are majored in another field, such as: Stephen Hawking, Stephen Wolfram, Elon Musk (Although I admit it is dubious to think of a CEO as a specialist in this sense), and now Garry Kasparov. Rationalwiki has a list dedicated to people who are Nobel prize winners in on field, talking absolute bullshit about another field: http://rationalwiki.org/wiki/Nobel_disease (That is, bullshit in the sense of nonsense, not in the insult sense).
Actually, I would probably go so far as to say that this attitude is strictly against autodidacticism, because where autodidacts favour knowledge about a subject, the other favours arguments from authority and complacence of thought.
Do you not think the people that have dedicated their lives to studying AI are likely to be more optimistic about its uses than the population at large?
I dislike the 'opinions as news' trend on the whole, but there is room in public debates for non-experts in a specific field. The knowledge required to develop the technology is different from the knowledge needed to be able to engage in debates about its uses.
AI has the power to take jobs away from humans. This is not a controversial view, mainly because it's already happened (such as in car manufacturing). In my opinion, there's not enough debate about the implications of the increasing sophistication of AI (not enough debate amongst the general population, there's plenty in tech circles).
Of course, discrimination has always been a problem but I'd like to believe we see some progress in "classical areas" (for instance sexism). The problem with AI is that it is always pigeonholing. As a result myriads of new classes of minorities which are so small that they neither have a name nor a voice will emerge. For instance, you get rated for creditworthiness and somehow you're not typical re: attributes x,y and z. You get a bad rating but you can't really complain as the algorithm probably didn't take things like sex or race (directly) into account. However, it could be that your FB posts are enjoying above-average amounts of likes from people with low education and this may raise red flags etc...
I do think there is a general phenomenon of "smart people outside of AI saying dumb things about AI" -- I'm guessing it's because of the massive implications artificial intelligence is going to have on every part of society.
It mostly covers his experiences with Deep Blue. Wish it was a longer talk. He generally has pretty positive view on computer and AI.
One interesting bit is when Demis Hassabis questioned if a chess-trained AlphaGo can be stronger at chess against Stockfish. Would be fun if they really try it.
PS: Hassabis once reached master standard at the age of 13 with an Elo rating of 2300 (at the time the second highest rated player in the world Under-14 after Judit Polgár who had a rating of 2335) and captained many of the England junior chess teams.