Acquisition of chess knowledge in AlphaZero

https://en.chessbase.com/post/acquisition-of-chess-knowledge-in-alphazero

by Bostonian • 5 years ago

225 92 5 years ago

en.chessbase.com

sarosh• 5 years ago

The paper itself is here: https://arxiv.org/abs/2111.09259 with the key conclusion that "Examining the evolution of human concepts using probing showed that many human concepts can be accurately regressed from the AZ network after training, even though AlphaZero has never seen a human game of chess, and there is no objective function promoting human-like play or activations" and "[t]he fact that human concepts can be located even in a superhuman system trained by self-play broadens the range of systems in which we should expect to find human-understandable concepts"

kevinventullo• 5 years ago

The human-understandable concepts are there, but I think finding them is still very hard, right? I.e. it doesn’t sound like the system including AZ was able to identify and articulate the human concepts such as king safety a priori.

iratewizard• 5 years ago

A game like chess is like two competing mathematical models. The human understanding of it is really just noticing patterns that exist in the model. I would hope an ML algorithm finds similar patterns.

Santosh83• 5 years ago

We are quite literally alpha zero ourselves. We developed chess by playing ourselves. I don't get what is particularly surprising that there are not infinite ways to play good chess?

ogogmad• 5 years ago

It's not a priori obvious that a machine can learn to play chess in a similar way to the way a human can.

I suppose if AGI gets achieved then some of this will seem predictable in hindsight. But we don't yet know if it's achievable.

V-25• 5 years ago

Human minds didn't evolve for playing chess. So there was a possibility that there are "hardware" limitations that prevent us from playng it the way it should be (give or take the inevitable error margin).

iechoz6H• 5 years ago

Aren't the shared concepts simply predicated on 'the rules of chess'? No surprise that an algorithm and a 'rational actor' share similar themes when both are constrained by the same rules.

baq• 5 years ago

the concept of material isn't found in the rules and yet was discovered independently during self-play.

posterboy2• 5 years ago

Can this be explained by the fact that the rules of chess evolved accomodating certain preferences that human players developed over the course of time?

Take a much simpler game like paper, rock scissors, where certain strategies exist as well (I hear). This should be much easier to analyze. Can somebody apply alphaZero to rock, paper, scissors, please?

AutumnCurtain2• 5 years ago

AlphaZero’s neural network evaluation function doesn’t have the same level of structure as Stockfish’s evaluation function: the Stockfish function breaks down a position into a range of concepts (for example king safety, mobility, and material) and combines these concepts to reach an overall evaluation of the position. AlphaZero, on the other hand, outputs a value function ranging from -1 (defeat is certain) to +1 (victory is guaranteed) with no explicitly-stated intermediate steps. Although the neural network evaluation function is computing something, it’s not clear what.

I find it fascinating that we can create something that can "learn" in some sense in a way we don't fully understand. Or am I misunderstanding the meaning of this passage?

beaconstudios2• 5 years ago

We do understand how they learn, but the structure of their knowledge is currently hard to discern in the same way that you can't get thoughts or memories out of somebody's head with a MRI scan.

sowbug• 5 years ago

It's interesting to think that if we create a human-level artificial intelligence using ML, we'll probably have no idea how it works.

noxvilleza2• 5 years ago

The way in which I think about it is that "learning" (in humans) often occurs is in the form of heuristics - certain rules or shortcuts which are not guaranteed to be optimal, but do a good job at simplifying (the search-state of) the problem. Instead of trying many simple comparisons between legal moves, we instead use a heuristic to score them.

Non-NN chess engines are really all about doing a good job at balancing these heuristics (like it says above). There's fundamentally an upper bound in how good the overall approach can be, simply because the underlying heuristics are defined and selected by humans: there could be many more heuristics that haven't been considered. This version of "learning" is really just removing human forced constraints which were introduced to help learning, by starting from a position without bias.

AutumnCurtain2• 5 years ago

Thanks so much for the response! It's really interesting stuff.

porphyra• 5 years ago

Why link to Chessbase instead of to the arxiv paper directly? https://arxiv.org/abs/2111.09259

Chessbase is notoriously sketchy and there is a lawsuit against them for blatantly stealing GPL code.

https://stockfishchess.org/blog/2021/our-lawsuit-against-che...

lacker• 5 years ago

The Chessbase article is a lot more readable than the arxiv paper. Yeah, they did some bad GPL stuff, but that just isn't really relevant to this particular discussion, and the chessbase link is better for the general reader.

throwaway69772• 5 years ago

Personally, I'd rather not support dishonest, but pretty outlets over direct sources.

aaron695• 5 years ago

Why link to Chessbase instead of to the arxiv paper directly?

Because their article is easier to read, your link isn't even directly to the paper.

And unlike the fake news media who gate-keeps information to ensure their monopoly and to stop you Gell-Mann Amnesia them, Chessbase has included a direct link to the PDF and the abstract at the bottom of their article. That's awesome.

Bike-shedding GPL violations over quality and honest articles doesn't interest lots of people. They can read about the OT stuff in proper HN discussions on it.

tedsanders2• 5 years ago

Sounds like when you teach a neural network chess, it ends up learning many of the same concepts that we do.

To me, this is a great illustration that superhuman intelligence is not magic. When a superhuman AI plays chess, the moves usually make sense to an expert. And even if they don't immediately make sense, they usually make sense once the expert plays out some lines to see what happens. Superhuman AIs have crushed humans at chess not by discovering new, wildly counterintuitive openings that dumb humans missed (it actually plays many of the same openings). Rather, it just does what humans do - win pieces, keep your king safe, attack, gain space, restrict your opponent's moves - but better and more consistently.

This paper builds on that concept and finds that not only are a superhuman AI's moves usually understandable by experts, but even some of the superhuman AI's internal representations are understandable by experts (!).

In the 5th century BCE, Greek philosophers thought the Earth was a sphere. Although they were eventually improved upon by Newton's ellipsoid, they still had arrived at the right concept. And here I think we see the same thing: although a superhuman AI 'understands' chess better than human experts, its understanding still builds upon the same concepts.

Thinking broadly, like, suppose we invent some superhuman AGI for $100B in 2100. And then we ask it how to optimize our paperclip production. I don't think its recommendations will be magic. Paperclip production is a fairly well understood problem: you bring in some metal, you reshape it, you send it out. It's probably gonna have much of the same advice we've already figured out: invest in good equipment, take advantage of economies of scale, choose cheap but effective metals, put factories near sources of energy and labor and raw materials, ship out paperclips in a hierarchical network with caches along the way to supply unanticipated demand, etc. Like, an AGI's suggestions may well be better than a management consultant's, but it's not like it's going to wave a magic wand and flip the laws of physics. Rather, I expect it to win within the same frameworks we've developed, but by doing things better and more consistently than humans.

barry-cotter• 5 years ago

Rather, it just does what [less successful people] do - ... - but better and more consistently.

Reminds me of comparisons between Renaissance and other quant funds. If you get enough better it looks like what you're doing is magic until you analyse what was done and see they have better systems and people.

Sometimes, magic is just someone spending more time on something than anyone else might reasonably expect - Teller

belter• 5 years ago

Shame they could not extend their better people and systems to all their customers:

"Renaissance Clients Exit After Firm’s Anemic Run of Results"

https://www.bloomberg.com/news/articles/2021-02-08/renaissan...

noxvilleza2• 5 years ago

When a superhuman AI plays chess, the moves usually make sense to an expert.

It's really interesting to look at very high level players analyzing specific AI moves. Many they understand after a short while, others take longer - and some they don't understand directly, but might have an inkling as to why it's good.

My feeling is that many high level analysts would rather just automatically defer to the AI's analysis rather than challenging it - so there is in some sense a lack of feedback to correct such situations. Even in the 1997 Deep Blue vs Kasparov match, game #1 Deep Blue apparently was a bugged position so it moved randomly (a legal random move) - but because Kasparov believed it was correct he felt the analysis was way beyond him so he conceded.

Galanwe• 5 years ago

When a superhuman AI plays chess, the moves usually make sense to an expert. And even if they don't immediately make sense, they usually make sense once the expert plays out some lines to see what happens.

Do you actually play chess? Half the moves made by an AI at 3500+ ELO (e.g. Stockfish 13+) is completely alien to even a human super GM like Nakamura or Carlsen. It's actually the most reliable and used way to know whether someone is cheating or not, if the moves don't make sense to high level players, then it's very probably someone using an engine.

And here I think we see the same thing: although a superhuman AI 'understands' chess better than human experts, its understanding is still leverages the same concepts.

That is very far from being a truth that lots of chess experts would agree on. Most chess players at high level reason with similar, well established concepts, such as tempo, positions, baits, color dominance, etc.

Super AIs don't use these concepts, and it's obvious when you see their moves. They often break tempo abruptly, disregard color dominance, enter what looks like dangerous positions,etc.

sudosysgen• 5 years ago

You can find out why an engine is making a move though. Hikaru and Magnus won't immediately understand why a move was played but they eventually will as they analyze the game.

hackinthebochs• 5 years ago

Stockfish prior to adopting a neural network evaluation engine made a lot of questionable-looking moves. But it makes many more natural looking moves after being upgraded to use the NN evaluation engine. Broadly speaking, NN based engines make more human looking moves compared to rule-based evaluations[1], with the human-expert accuracy going up as the human rating increases.

[1]https://www.cs.toronto.edu/~ashton/pubs/maia-kdd2020.pdf (scroll down for the graphs)

z5h2• 5 years ago

To add to Galanwe's point:

Wikipedia says (about AlphaGo):

"AlphaGo's playing style strongly favours greater probability of winning by fewer points over lesser probability of winning by more points. Its strategy of maximizing its probability of winning is distinct from what human players tend to do which is to maximize territorial gains, and explains some of its odd-looking moves."

In other words optimizing the likelihood of winning looks strange to experts who use other heuristics to assess positions.

jimmytucson• 5 years ago

Half the moves made by an AI at 3500+ ELO (e.g. Stockfish 13+) is completely alien to even a human super GM

The analysis in the paper draws the exact opposite conclusion. Alpha zero’s play agrees remarkably with humans despite not seeing a human game (e.g. it found the Berlin defense to the Ruy Lopez).

Also your point is trivially not true for other engines because they use books from human-played games.

Engines don’t win by playing novel moves all the time that humans can’t comprehend. They win by playing the most accurate move every time (never making a mistake).

BurningFrog• 5 years ago

Sure, but when I watch Carlsen or Nakamura okay, their moves often look alien to me.

tedsanders2• 5 years ago

Do you actually play chess?

Yes. I have a chess set on my kitchen counter from playing blindfold last night. I led my high school to the state championship. I am rated ~2000 in blitz and crazyhouse on Lichess. I watched 10+ analysis videos of AlphaZero when it came out.

If you have a criticism of my claim, just state the criticism. Even if I was a grandmaster, I would not be immune from being wrong.

And to be clear, my claim is that human experts usually (not always) understand the pros and cons of a move made by AlphaZero. My claim is not that human experts usually know why one move, with its set of pros and cons, was selected over another move, with a different set of pros and cons.

And of course, there are exceptions. Super long endgame sequences, for example. Even with a lifetime I doubt any human can understand these sequences beyond the tautological "it's winning because it wins": https://en.chessbase.com/post/just-one-of-17-823-400-766-pos...

Scarblac• 5 years ago

Some moves are hard to understand, but it's more like one or two per game. The AI is after all a lot better than Nakamura and Co.

The usual way to detect cheating is detecting incredible accuracy over a whole game, not individual engine like moves.

xondono• 5 years ago

Do you actually play chess? Half the moves made by an AI at 3500+ ELO (e.g. Stockfish 13+) is completely alien to even a human super GM like Nakamura or Carlsen

No they're not. GMs get thrown out on certain moves, because we tend to play chess strategically, that is, we make a plan and try to find ways to make it work. AIs look like they have a plan and then immediately change course when they compute it's beneficial to do so.

As GP says, it takes a minute or two, but those same GMs will (Nakamura does this constantly in some of his streams) reevaluate the strategy, play some lines to see what is tripping the AI and then figure it out. It's that rapid change of strategy from, as an example, extremely defensive to wildly aggressive, which makes them suspicious that they're playing a computer.

And here I think we see the same thing: although a superhuman AI 'understands' chess better than human experts, its understanding is still leverages the same concepts.

In that I'd agree with you that GP is wrong. For the AI there's no concepts, it's evaluating the chance of success, the perceived "weights" are irrelevant, you could map any number of concepts and get a correlation between them.

bluecalm• 5 years ago

I am sorry but you're completely wrong. Modern engine's moves are usually easy to understand if you can play around with the engine running. The engine is just way more precise. Usually the variations needed to understand the moves aren't particularly long either. For me (2300 ELO) it's usually 3-6 ply deep when I go "ahhhh obviously this is better".

The method you're referring to is based on detecting how often a human picks the top move out of 2-3 similarly rated ones and most of the time understanding which one of the 56.5%, 57.1% and 57.3% moves is the best is difficult but understanding why 55% move is better than 51% one is usually pretty easy. I am using AlphaZero/Lc0 evals here btw as I find them more intuitive than traditional engines ones.

I think your opinion might have been formed many years ago. There was indeed a time where engines like traditional Stockfish (before NNUE) were already much stronger than humans but their strategic play made little sense. It turned out that they were in fact weak strategically and modern crop of engines showed it. Try running Lc0 with modern network vs say Stockfish 7 and you will see games which could be described as humiliating defeats one after another.

It turned out that we are actually very good at figuring chess out. Not very precise, slow and inconsistent but we've got most of the high level concepts right.

It doesn't mean engines don't come up with occasional move that is just too deep to get without long analysis and too counterintuitive to consider during play. It's just rare. Most of the moves are intuitive, and make a lot of sense to a human.

29athrowaway• 5 years ago

In the first match of the Lee Sedol / Alpha Go jubango, many moves were initially interpreted as slow, or simply mistakes.

Human players try to maximize score difference, Alpha Go tries to maximize the probability of winning. This means, between playing a risky move that makes more territory, and a safe move that makes less territory, Alpha Go prefers the safer move.

To experts, those move came across as "slow", inefficient or suboptimal. But they turn out to be the superior moves.

AceyMan• 5 years ago

The difference in strategy makes sense when you consider humans know we are likely to make some variable (suboptimal) choices and so our strategy includes building a buffer to guard against the inevitable gaff.

GAI/NN doesn't 'fear' variability and so it's OK to ride a razor thin margin all the way to the victory.

[edit: rephrase]

csee2• 5 years ago

But could there be a class of games whose solutions are so complex that we can't understand them, yet an AGI would?

We designed chess to be easily understood by humans. There's a cap on the conceptual complexity that fits within our capabilities. So it's not surprising that we can understand AI solutions that merely asymptote to that fixed cap.

ajuc2• 5 years ago

Chess are very limited compared to the whole universe, AGI could for example invent self-replicating fusion-driven paperclip factories, there's nothing in the universe rules that makes this impossible we simply don't know how to do it (yet).

zwegner• 5 years ago

Or even without using any such new technology, it could at least take advantage of the fact that almost all human communication is mediated by computers, and coerce virtually the entire human race into helping it. I think the GP wasn't being nearly creative enough when thinking about the possibilities/dangers of AGI.

notahacker2• 5 years ago

Sure, and God could smite all unbelievers, turn water into wine and bring about the Rapture: there are no limits to what purely hypothetical omniscient beings can do.

But based on we actually know about the sort of AI systems humans actually know how to build and obtain useful results from, we'd end up with a computer which absorbed time series of paperclip sales figures, supply chain costs and marketing campaign data which recommends small paperclips for the Kazakh market, and there's no reason to believe a model trained to optimise paperclip production would know anything at all about how to generate fusion power. This article suggests we might even be able to analyse why the algorithm prioritised that obscure market after applying a lot more thought to the relevant data ourselves. (We can also reason that the AI model is more likely to be wrong about the small paperclips for the Kazakh market, because the permutations for the Kazakh market aren't as perfectly encapsulated in the underlying model and simulated by training on data as the permutations of chess)

snarf21• 5 years ago

Serious question: I haven't done the research but I've always felt that in abstracts like Chess and Go that a larger part of the skill is in "reading the board". Humans are leveraging our pattern matching but we can get tired or miss even one little thing. The AI never misses, ever. I've always wondered how a human would do against an AI if the human was given a list of the top 50 potential moves in a randomized order to consider before making their move. Would the human be on more level playing field in terms of reading?

PeterisP• 5 years ago

There never are 50 reasonable possible moves in Chess, but with top 3 it would be an advantage.

Perhaps it's illustrative to consider the chess variant "hand and brain" played by teams of two people, one ("brain") picking the piece to be moved, and the other ("hand") choosing where to move it; the relationship is complex, but in general having a strong "brain" limit your possible moves to just a single piece would improve the chances of a weaker "hand" above what they could do on their own.

ucha• 5 years ago

After watching some of the games AlphaZero played against Stockfish, I have to say I don't really agree that the moves make sense to a grandmaster... AlphaZero has this uncanny ability to sacrifice many pawns to restrict the movement of the opponent's pieces, something that you might see in advanced play but nowhere near the extent you see it with AlphaZero. Many more of the moves it plays are hard to make sense of compared to a classic engine such as Stockfish.

See GothamChess' very entertaining analysis: https://www.youtube.com/watch?v=8dT6CR9_6l4

themeiguoren• 5 years ago

The fallacy here is that we can only measure whether AZ is playing human strategies, by comparing against our own known space of strategies. Sure, we human have developed good heuristics the AI has also found, but we can’t know if it has better/more subtle concepts that we haven’t discovered, exactly because we haven’t discovered them yet. Who is to say “mobility” is even a coherent top level concept compared to some unpronounceable abstraction the AI has figured out?

a_shovel2• 5 years ago

The sci-fi trope that aliens or superintelligent AIs will have thought processes completely different and incomprehensible to humans is probably not true.

If the aliens live in our universe, evolved on a planet with limited resources, need food, breathe, etc., then they'll likely have a lot in common with Earth animals psychologically. If the AI is tasked with doing something that humans also do, then its tactics will be at least recognizable to an expert in that field. Moreso in well-understood systems such as chess.

Is it really plausible that an advanced chess AI's tactics would have no concept of threats or king safety? The same tactics apply to it as do to us.

rtpg• 5 years ago

I don't want to be too dismissive of the work going into this, but Chess isn't that huge of a search space. Games can quickly end in many scenarios from certain points.

Does having an evaluation system that takes into account king safety only as a consequence of having taken the entire game state into account, mean that it understands king safety? Can you extract these tactics from the AI? Can it explain the moves?

I would be interested in seeing work here that can provide some comprehensibility to the correlations, as that's how we can get to actual novel intuition. "The computer is good with this neural network due to pattern matching" is less actionable than "the computer is good and found this kind of pattern, and we think it relates to this".

EDIT: I think that if anything Chess is an interesting example of humans emulating computers. People sitting around thinking about variations, trying to snuff out what is good/bad. Granted, the fact that you're playing against other humans is pretty important at a higher level, but it's centuries of people sitting down and running through the search space, using our approximations for what are good and bad.

csee2• 5 years ago

I don't share this view. They would have had a lot in common at a point in time in their history, but who's to say the consequences of millions of years of additional evolution post-energy abundance, let alone the consequences of embryo selection and genetic engineering? We've had a few tens of thousand years of evolution with relative (but still not really) caloric abundance and look how much bigger our brains got. Add nuclear fusion and an extra 50 million years to that nascent process.

I think chess strategy is a flawed way to be thinking about this because we designed the rules and the solution space to be comprehensible to us. It's not surprising that we can understand AI strategy. That simply follows from the solution space itself being small and comprehensible.

TheRealNGenius• 5 years ago

When a fly makes first contact, do you care to reason with it in terms it can understand, as to why you are now swatting it out of existence?

posterboy2• 5 years ago

Yes, I wave and take a few sloppy shots before I get up in arms. Some take flight (ha!) but some of the more agressive beasts seem to send the message that I should take what's coming for me.

Swarms of agressive insects are also known to be quite defensive and alert the troops.

Take the human out of the loop and consider mating rituals. It amazes me to no end how simple bugs manage to procreate ...

This does of course not meat the common definition of communication, if it involves pheromones and such like. Nevertheless, when another philosopher says that Pain is impossible tondefine, I will show them pain right there.

plutonorm• 5 years ago

"Sugar good. Planet is sugar. Sorry bye". But translated into firing of neurons within the fly brain. Of course there is that concept of other and self that the fly does not have. But this might be more successful with monkeys. Your other point about motivation to describe their purpose still stands - but the point about there being too limited similarity between conceptions of the world to convey meaning between species - that doesn't work for me. They are situated in the same universe, and it's very likely that there isn't a qualitative difference between their intelligence and our own. Intelligence must be a universal law, something that falls out of the correct interpretation of thermodynamics in open systems.

Interdimensional aliens might have a more varied purpose I suppose, but then their actions wouldn't be intelligible to us at all. They might decide to convert half all life on earth into the opposite chiral form, because reasons. Or rather because there is some unknown mapping between chirality in this plane to the goals within another plain. The chirality of these biological forms within these dimensions being but the lower part of higher dimensional things which are subject to the motivating forces of higher dimensions. Although it's much more likely that any changes would not be coherent at all in our perspective of reality. The purposeful actions of other beings within other perspective frames of totality would more likely look like random thermal/quantum fluctuations within our perspective. There I solved quantum randomness for you.

astrobe_• 5 years ago

then they'll likely have a lot in common with Earth animals psychologically

The term "psychologically" here is interesting because it skips the "biologically" part. What if they don't have neurons like us or AZ? Could one even talk about psychology?

a_shovel2• 5 years ago

Well, that's kind of the point. Sci-fi authors will talk about how Super-AIs or aliens might think completely differently from us, but then impose lots of implicit similarities that would make that implausible.

Sure, non-sapient fungus-like extraterrestrial life-forms will definitely not think like us, even if they have neurons. But "aliens" in most stories implicitly refers to mobile intelligent beings capable of cooperation, communication, developing technology to a level more advanced than our own, and are animal-like and thus need to breathe, eat, drink, maintain homeostasis, etc. Often they'll even have the same body plan as us.

Also, things like the prisoner's dilemma, the tragedy of the commons, and that lying may be beneficial in some circumstances aren't mere artifacts of human psychology, but are straightforward results of game theory.

If a creature is that similar to a human physically, can it really be totally incomprehensible psychologically?

V-25• 5 years ago

Chess is a closed system. Alien beings could be "playing checkers", so to speak. A biology completely different to anything existing on Earth (say, inteligent life living in the astmophere of a hot gas giant planet) would likely lead an utterly incompatible psychology. It's kind of like in the famous Wittgenstein quote: "if a lion could speak, we could not understand him". Well, lion is an Earth animal, but you see the point.

Tenoke• 5 years ago

From recorded data, we can see that everyone seemed to play e4 in the 1500s. Over centuries, moves like d4, Nf3 or c4 emerged as credible and fashionable alternatives.

The graph they have seems to suggest the opposite - nobody playing e4, just d4 early on (and AZ also gravitating most towards d4).

njarboe• 5 years ago

I think it pretty cool how AlphaZero does not converge to a single best opening move and its variation of opening moves is higher than humans.

Seeing people analyze AlphaZero games on YouTube, it plays a more open and aggressive game of chess and goes for the win instead of the very defensive play of most of the top players today. Some people hope that if people study how AlphaZero plays chess (the best chess player in the world) to improve their game it might make chess more interesting as most games at the high level now end in draws.

Y_Y2• 5 years ago

It's a somewhat counterintuitive result from game theory that many of the best strategies we know to various non-trivial games are "mixed strategies". This refers to a strategy (presumably) like the one by AlphaZero here, where moves are chosen from a probabilistic distribution, rather than just picking "the best one". The sub-optimal move choice is counterbalanced by the unpredictability.

jstx12• 5 years ago

Mixed strategies in game theory don’t involve making sub-optimal decisions for the sake of unpredictability. At equilibrium every sub-strategy of the mix is a best response (you have multiple best responses).

computer• 5 years ago

In games with hidden information, yes. In open games? There is an optimal path from start to victory, for one of the players. No need to deviate randomly.

phreeza2• 5 years ago

What I want to know is if it is possible to do the opposite: extract novel, useful and human-understandable heuristics from a chess engine?

karol• 5 years ago

Yes, it has been done with go and Deep Mind. The program found novel patterns of playing that were not known. I don't see why this wouldn't happen with chess.

phreeza2• 5 years ago

I was thinking more along the lines of automated discovery of such heuristics.

Acquisition of chess knowledge in AlphaZero

Related Stories