Yes, this is not an AGI. But the hockey-stick takeoff from defeats some players, to defeats an undefeated world-champion, to defeats the version of itself that beat the world champion 100% of the time is nuts. If this happens in other domains, like finance, health, paper clip collection, the word singularity is really well chosen--we can't see past this.
While this is promising, there's a long way to go between this and the other things you mentioned. Go is very well-defined, has an unequivocal objective scoring system that can be run very quickly, and can be simulated in such a way that the system can go through many, many iterations very quickly.
There's no way to train an AI like this for, say, health: We cannot simulate the human body to the level of detail that's required, and we definitely aren't going to be able to do it at the speed required for a system like this for a very long time. Producing a definitive, objective score for a paper clip collection is very difficult if not impossible.
AlphaGo/DeepMind represents a very strong approach to a certain set of well-defined problems, but most of the problems required for a general AI aren't well-defined.
> Do you care to give an example? Are they more or less well defined than find-the-cat-in-the-picture problem?
You mean like go over and feed the neighbor's cat while they're on vacation?
How about instead, being able to clean any arbitrary building?
Go isn't remotely similar to the real world. It's a board game. A challenging one, sure, and AlphaGo is quite a feat, but it's not exactly translatable to open ended tasks with variable environments and ill-specified rules (maybe the neighbor expects you to know to water the plants and feed the goldfish as well).
At this point, there is no evidence that the limiting factor in these cases is AI/software.
The limiting factor with the neighbors cat is the robotics of having a robust body and arm attachment. We know that the scope of current AI can:
1) Identify a request to feed a cat
2) Identify the cat, cat food and cat's bowl from camera data
3) Navigate an open space like a house
Being able to clean an arbitrary building is also more the challenge of building the robot than the AI identifying garbage on a floor or how to sweep something.
It is not clear there are hard theoretical limits on an AI any more. There are economic limits based on the cost of a programmer's attention. There are lots of hardware limits (including processor power).
In my opinion the deepest and most difficult aspect of this example is the notion of 'clean' which will be different across contexts. Abstractions of this kind are not even close to understood in the human semantic system, and in fact are still minimally researched. (I expect much of the progress on this to come from robotics, in fact.)
I remember seeing a demonstration by a deep learning guy of a commercially available robot cleaning a house under remote control. You are seriously underestimating the difficulty of developing software to solve these problems in an integrated way.
This. It is a lot like the business guy thinking it is trivial to program a 'SaaS business' because he has a high level idea in his mind. Like all things programming the devil is in the detail.
The hardware is certainly good enough to assist a person with a disability living in a ranch house with typical household tasks. As demonstrated by human in the loop operation.
We have have rockets that can go to orbit, and we have submersibles that can visit the ocean floor. That does not mean the rocket-submarine problem is solved, doing both together is not the same problem as doing both separately.
The difference is a go AI can play billions of games and a simple 20 line C program can check, for each game, who won.
For "cat in the picture", every picture must have the cat first identified by a person, so the training set is much smaller, and Google can't throw GPUs at the problem.
The absolute value of any Go board position is well-defined, and MCTS provides good computationally tractable approximations that get better as the rest of the system improves but already start better than random.
Check the Nature paper (and I think this is one of the biggest take-aways from AlphaGo Zero):
"Finally, it uses a simpler tree search that relies upon this single neural network to evaluate positions and sample moves, without performing any Monte Carlo rollouts."
In this new version, MCTS is not even used to evaluate a position! Speaking as a Go player, the ability for the neural network to accurately evaluate a position without "reading" ahead is phenomenal (again, read the Nature paper last page for details).
You don't even need to produce an AGI for this kind of intelligence to be frightening.
At some point, a military is going to develop autonomous weapons that are vastly superior to human beings on the battle field, with no risk of losing human lives, and there is going to be a blitzkrieg sort of situation as the relative power of nations shifts dramatically.
If we have two such countries we could have massive drone and cyberwars being fought faster than people even can comprehend what's happening.
Right now most countries insist on maintaining human control over the machinery of death. But that will only last for as long as autonomous death machines don't dominate the battlefield.
It's a fun challenge right now to build a machine that can win in Starcraft, but it's really a hop skip and a jump from there to winning actual wars.
In that case you just nuke the shit out of everybody or create army if autonomous suicide bomber with nukes, biological and chemical weapons of all kinds. Once all humans are extinct the harmony on earth will be restored and everyone will leave happily ever after.
i'm not sure robot soldier is scarier than nukes. generally speaking, if they are just single task robots performing functions in dangerous situations, that seems like an improvement to risking human lives.
The core technique of AlphaGo is using tree search as a "policy improvement operator". Tree search doesn't work on most real-world tasks: the "game state" is too complex, there are too many choices, it's hard to predict the full effect of any choice you might make, and there often isn't even a "win" or "lose" state which would let you stop your self-play.
MCTS means "Monte-Carlo Tree Search". It's the core of the algorithm. The big difference is that it doesn't use rollouts, or random play: it chooses where to expand the tree based only on the neural network.
That's not what Monte Carlo Tree search is. The new version is still one neural network + MCTS. There's no way to store enough information to judge the efficiency of every possible move in a neural network, therefore a second algorithm to simulate outcomes is necessary.
If you read the paper, they do in fact still use monte-Carlo tree search. They just simplify their usage in conjunction with reducing the number of neural networks to 1
Tree search is also used during play. In the paper, they pit the pure neural net against other versions of the algorithm -- it ends up slightly worse than the version that played Fan Hui, at about 3000 ELO.
> AlphaGo Zero does not use “rollouts” - fast, random games used by other Go programs to predict which player will win from the current board position. Instead, it relies on its high quality neural networks to evaluate positions.
This is an interesting question to ask in these "how far away is AGI" discussions:
I was once at a conference where there was a panel full of famous AI luminaries, and most of the luminaries were nodding and agreeing with each other that of course AGI was very far off, except for two famous AI luminaries who stayed quiet and let others take the microphone.
I got up in Q&A and said, “Okay, you’ve all told us that progress won’t be all that fast. But let’s be more concrete and specific. I’d like to know what’s the least impressive accomplishment that you are very confident cannot be done in the next two years.”
There was a silence.
Eventually, two people on the panel ventured replies, spoken in a rather more tentative tone than they’d been using to pronounce that AGI was decades out. They named “A robot puts away the dishes from a dishwasher without breaking them”, and Winograd schemas. Specifically, “I feel quite confident that the Winograd schemas—where we recently had a result that was in the 50, 60% range—in the next two years, we will not get 80, 90% on that regardless of the techniques people use.”
I spent an hour of my life that I'll never get back reading Yudkowski's
overly-long article and I believe I can summarise it thusly:
"We don't know how AGI will arise; we don't know when; we don't know why; we
don't know anything at all about it and we won't know anything about it until
it's too late to do anything anyway; We must act now!!"
The question is- if we don't know anything about this unknowable threat, how can
we protect ourselves against it? In fact, since we're starting from 0
information, anything we do has equal chances of backfiring and bringing forth
AGI as it has of actually preventing it. Yudkowski is calling for random action,
without direction and without reason.
Besides, if Yudkowski is none the wiser about AGI than anyone else, then how is
he so sure that AGI _will_ happen, as he insists it will?
Yudkowski is fumbling around in the dark like everyone else in AI. Except he
(and a few others) has decided that it's a good strategy, under the
circumstances, to raise a hell of a racket. "It's dark!" he yells. "Beware of
the darkness!". Yeah OK, friend. It's dark- we can all tell. Why don't you pipe
down and let us find the damn light?
Sorry but I don't really see Yudkowski's contributions as "fundamental research into AI safety". More like navel-gazing without any practical implications. At best, listening to him is just a waste of time. At worse, AGI is a real imminent threat and having people like him generating useless noise like he does will make it harder for legitimate concerns to be heard, when the time comes.
Yes, I did and it's very bad form to go around asking people if they read the article. Try to remember that different people form different opinions from similar information.
Well, then you should have noticed what the article was about, which was not to detail a research program about AI safety.
Different articles can address different aspects of a problem without being accused of advocating "random action". That's just ridiculous.
>The question is- if we don't know anything about this unknowable threat, how can we protect ourselves against it? In fact, since we're starting from 0 information, anything we do has equal chances of backfiring and bringing forth AGI as it has of actually preventing it. Yudkowski is calling for random action, without direction and without reason.
Are you sure you read the essay? That's literally the question he answers.
At any rate, we do have more than '0 information', and if you make an honest effort to think of what to do you can likely come up with better than 'random actions' for helping (as many have).
>> Are you sure you read the essay? That's literally the question he answers.
My reading of the article is that he keeps calling for action without specifying what that action should be and trying to justify it by saying he can't know what AGI would look like (so he can't really say what we can do to prevent it).
>> if you make an honest effort to think of what to do you can likely come up with better than 'random actions' for helping (as many have).
Sure. If my research gets up one day and starts self-improving at exponential rates I'll make sure to reach for th
... yeah, before reading that link my position was "Wow, that's super neat, but Go is a pretty well-defined game," and after reading it I remembered that my position maybe a year or two ago was "Chess is a well-defined game that's beatable by AI techniques but Go is acknowledged to be much harder and require actual intelligence to play and won't be solved for a long while" and now I'm worried. Thanks for posting that.
Go is still a well defined game within a limited space that doesn't change, and rules that don't change. It's just harder than Chess, but that doesn't make it similar to tons of real world tasks humans are better at.
That's probably true, but that's very much not what people were saying about Go a couple years ago. There were a lot of people talking about how there isn't a straightforward evaluation function of the quality of a given state of the board, how things need to be planned in advance, how there's much more combinatorial explosion than in chess, etc., to the point where it's a qualitatively different game.
For me, as someone who accepted and believed these claims about Go being qualitatively different, realizing that no, it's not qualitatively different (or that maybe it is, but not in a way that impedes state-of-the-art AI research) is increasing my skepticism in other claims that board games in general are qualitatively different from other tasks that AIs might get good at.
(If you didn't buy into these claims, then I commend you on your reasoning skills, carry on.)
About those claims- this is from Russel and Norvig, 3d ed. (from 2003, so a way back):
Go is a deterministic game, but the large branching factor makes it challeging.
The key issues and early literature in computer Go are summarized by Boozy and
Cazenave (2001) and Muller (2002). Up to 1997 there were no competent Go
programs. Now the best programs play most of their moves at the master level;
the only problem is that over the course of a game they usually make at least
one serious blunder that allows a strong opponent to win. Whereas alpha—beta
search reigns in most games, many recent Go programs have adopted Monte Carlo
methods based on the UCT (upper confidence bounds on trees) scheme (Kocsis and
Szepesvari, 2006). The strongest Go program as of 2009 is Golly and Silver's
MoGo (Wang and Golly, 2007; Gelly and Silver, 2008). In August 2008, MoGo scored
a surprising win against top professional Myungwan Kim, albeit with MoGo
receiving a handicap of nine stones (about the equivalent of a queen handicap in
chess). Kim estimated MOGO's strength at 2-3 dan, the low end of advanced
amateur. For this match, MoGo was run on an 800-processor 15 terailop
supercomputer (1000 limes Deep Blue). A few weeks later, MoGo, with only a
five-stone handicap, won against a 6-dan professional. In the 9 x 9 form of Go,
MoGo is at approximately the 1-dan professional level. Rapid advances are likely
as experimentation continues with new forms of Monte Carlo search. The Computer
Go Newsletter, published by the Computer Go Association, describes current
developments.
There's no word about how Go is qualitatively different to other games, but maybe the referenced sources say something along those lines. Personally, I took a Masters course in AI two years ago, before AlphaGo and I remember one professor saying that the last holdout where humans can still beat computers in board games was GO, but I don't quite remember him saying anything about qualititative difference. Still, I can recall hearing about the idea that Go needs intuition or something like that, except I've no idea where I've heard that. I guess it might come from the popular press.
I guess this will sound a bit like the perenial excuse that "if it works, it's not AI" but my opinion about Go is that humans just weren't that good at it, after all. We may have thought that we have something special that makes us particularly good at Go, better than machines- but AlphaGo[Zero] has shown that, in the end, we just have no idea what it means to be really good at it (which, btw, is a damn good explanation of why it took us so long to make AI to beat us at it).
That, to my mind, is a much bigger and much more useful achievement than making a good AI game player. We can learn something from an insight into what we are capable of.
s/2003/2009/, I think, but the point stands. (Also I think I have the second edition at home and now I want to check what it says about Go.)
> my opinion about Go is that humans just weren't that good at it, after all. We may have thought that we have something special that makes us particularly good at Go, better than machines- but AlphaGo[Zero] has shown that, in the end, we just have no idea what it means to be really good at it (which, btw, is a damn good explanation of why it took us so long to make AI to beat us at it).
> the last holdout where humans can still beat computers in board games was GO
False, because nobody ever bothered to study modern boardgames rigorously.
Modern boardgames have small decision trees but very difficult evaluation functions. (Exactly opposite from computational games like Go.)
Modern boardgames can probably be solved by pure brute force calculation of all branches of the tree, but nobody knows if things like neural networks are any good for playing them.
In AI, "board games" generally means classical board games (nim, chess, backgammon, go etc) and "card games" means classical card games (bridge, poker, etc). Russel & Norvig also discuss some less well-known games, like kriegspiel (wargame) if memory serves, but those are all classical at least in the sense that they are, well, quite old.
I've seen some AI research in more modern board games actually. I've read a couple of papers discussing the use of Monte Carlo Tree Search to solve creature combat in Magic: the Gathering and my own degree and Master's dissertation were about M:tG (my Master's was in AI and my degree dissertation was an AI system also).
I don't know that much about modern board games, besides collectible card games, but for CCGs in particular, the game trees are not small. I once calculated the time complexity of traversing a full M:tG game tree as O(b^m * n^m) = 2.272461391808129337799800881135e+5564 (where b the branching factor, m the average number of moves in a game and n the number of possible deck permutations for a 60 card deck taking into account cards included multiple times). And mine was probably a very conservative estimate.
Also, to my knowledge, Neural nets have not been used for magic-playing AI (or any other CCG playing AI). What has been used is MCTS, on its own, without terrible success. The best AI I've seen incorporates some domain knowledge, in the form of card-specific strategies (how to play a given card).
There are some difficulties in using ANNs to make an M:tG AI. Primarily, the fact that a truly competent player should be able to pick up a card it's never seen before and play it correctly (or decide whether to include it in a deck, if the goal is to also address deck-building). For this, the AI player will need to have at least some understanding of M:tG's language (ability text). It is my understanding that other modern games have equal requirements to understand some game context outside of the main rules, which complicates the traditional tactic of generating all possible moves, pruning some and choosing the best.
In any case what I meant to say is that people in AI have indeed considered other games besides the classical ones- but when we talk about "games" in AI we do mean the classics.
> but when we talk about "games" in AI we do mean the classics
Only because of inertia. There's nothing inherently special about "classics". Eventually somebody will branch out once Go and poker are mined out of paper and article opportunity.
Once we do then maybe some new, interesting algorithms will be found.
In principle, every game can be solved by storing all possible game states in a database. Where brute-force storing is impractical due to size concerns, compression tricks have to be used.
E.g., Go is a simple game because at the end, every one of the fixed number of board spaces is either +1, -1 or 0. Add them up and you know if you won. This means that every move is either "correct" or "incorrect"; the problem of classifying multidimensional objects into two classes is a problem that we're pretty good at now, and things like neural networks get the job done.
A slightly more complex game like Agricola has no "correct" and "incorrect" moves because it's not zero-sum; you can make an "incorrect" move and still win as long as your opponent is forced to make a relatively more "incorrect" move.
Not sure how much of a difference that makes, but what's certain is that by (effectively) solving Go we've only scratched the surface. It's not the end of research, only the beginning.
Sure. Research in game playing AI doesn't end with Go, or any other game. We may see more research in modern board games, now that we're slowly running out of the classics.
I think you're underestimating the amount of work and determination it took to get to where we are today, though (I mean your comment about "inertia"). Classic board games have the advantage of a long history and of being well understood (the uncertainty about optimal strategies in Go notwithstanding). Additionally, for at least some of them like chess, there are rich databases of entire games that can be used outright, without the AI player having to generate-and-test them in the process of training or playing.
The same is not true for modern games. On the one hand, modern board games like Agricola (or, dunno, Settlers or Carcassonne etc) don't have such an extensive and multi-national following as the classics so it's much harder to find a lot of data to train on (which is obviously important for machine-learning AI players). I had that problem when considering an M:tG AI trained with machine learning: I would have liked to find play-by-play data on professional games but there just isn't any (or where there is it's not enough, or it's not in any standardised format).
Finally, classic board games have cultural significance that modern board games dont' quite match, despite the huge popularity of CCGs like M:tG or Pokemon, or Eurogame hits like Settlers. Go, chess and backgammon in particular have tremendous historical significance in their respective areas of the world- chess in Eastern Europe, backgammon in the Middle East, Go in SE Asia. People go to special academies to learn them, master players are widely recognised etc. You don't get that level of interest with modern board games- so there's less research interest for them, also.
People in game playing AI have been trying for a very long time to crack some games like Go and, recently, poker (not quite cracked yet). They didn't sit around twiddling their thumbs all those years, neither did they choose classical board games over modern ones just because they didn't have the imagination to think of the latter. In AI research, as in all research, you have to make progress before you can make more progress.
> Go is acknowledged to be much harder and require actual intelligence to play
No, Go is a much less intelligent[1] game. It has a huge decision tree and requires massive amounts of computation to play, but walking trees and counting is exactly what computers do well and what humans do poorly.
[1] 'Intelligence' here means exactly that which differentiates humans from calculators: the ability to infer new rules from old ones.
The smoke is when things like the same simulated robot that learned to run around like a mentally challenged person also learns to simulate throwing and can read very basic language.
It will seem quite stupid and inept at first. So people will dismiss it. But when they have a system with general inputs and outputs that can acquire multiple different skills, that will be an AGI, and we can grow it's skills and knowledge passed human level.
The hockey stick is lying horizontally though instead of vertically. If it took 3 days to go from 0 to beating the top player in the world, I wouldn't have expected it to take 21 days to beat next version. I guess something happens at the top levels of Go that make training much harder.
On another note, I didn't look at the details closely but it seems AlphaGo Zero needed much less compute training time than Alpha Go Master. Could getting rid of any human inputs really make it that much more efficient? That implies it will be able to have an impact in many different areas, which is a bit scary...
(Updated - it took 3 days to beat the top player in the world.)
This type of curve is what I would expect out of machine learning. At first there is rapid improvement as it learns the easy lessons. The rate then slows down as further incremental improvements become less impact.
What is, perhaps, surprising is that human play happens to be relatively close to the asymptote. Although this could be explained by Alphago being the first system to beat humans. If its peek performance were orders of magnitude higher than humans, a weaker program would have already beaten us.
The horizontal hockey stick makes sense to me in terms of learning. Each increased layer of of understanding a complex system could mean a potentially exponentially increasing difficulty.
I'm sure it's naive to jump to sci-fi conclusions just yet, but I admit it's equal parts fascinating and terrifying. The general message of the posts is that human knowledge is cute but not required to find new insights. Define the measure of success and momma AI will find the answer. At this point, the path to AGI is about who first defines its goals right and that seems... doable? Even scarier: We think the holy grail of AI is simulating a human being. The AI of the future might chuckle at that notion.
Wait for Alpha StarCraft for some real panic. So far RL based method has limited success outside of simple games(Not to say Go is simple, but rather the presentation and control parts of the format).
I'd like to see a StarCraft player AI that wins using a mere 1/10th of the effective actions per minute (EPM) of world class players.
To me it seems beating another player while using fewer actions indicates superior skill, understanding and/or intellect.
Not sure I agree with this fully. Certainly many actions used in a typical SC game are redundant, but there are reasons for it. Lag for one. If there's a possibility of lag or dropped packets, spamming a command will help nullify this problem.
The other is the entire reason for high APM, the stop/start problem. Pro players keep high APM so that when they actually need high EPM their muscle memory is already at full tilt. If you slow down your APM during lulls in the action it becomes harder to suddenly increase it when a fight happens.
Certainly that's an entirely human condition that a machine wouldn't need to worry about. But I'm not sure it means lack of skill.
You expressed my exact thoughts and I was about to link to the same insightful article. I guess my comment could've been shortened as a silent upvote, but I commented anyway.
Games are a joke compare with real life. The number of variables and rules is well defined in games, while in real life it is not. That is why AGI is not coming anytime soon.
https://intelligence.org/2017/10/13/fire-alarm/
Yes, this is not an AGI. But the hockey-stick takeoff from defeats some players, to defeats an undefeated world-champion, to defeats the version of itself that beat the world champion 100% of the time is nuts. If this happens in other domains, like finance, health, paper clip collection, the word singularity is really well chosen--we can't see past this.