As someone who studied AI in college and am a reasonably good amateur player, I have been following the matches between Lee and AlphaGo.
AlphaGo plays some unusual moves that go clearly against any classically trained Go players. Moves that simply don't quite fit into the current theories of Go playing, and the world's top players are struggling to explain what's the purpose/strategy behind them.
I've been giving it some thought. When I was learning to play Go as a teenager in China, I followed a fairly standard, classical learning path. First I learned the rules, then progressively I learn the more abstract theories and tactics. Many of these theories, as I see them now, draw analogies from the physical world, and are used as tools to hide the underlying complexity (chunking), and enable the players to think at a higher level.
For example, we're taught of considering connected stones as one unit, and give this one unit attributes like dead, alive, strong, weak, projecting influence in the surrounding areas. In other words, much like a standalone army unit.
These abstractions all made a lot of sense, and feels natural, and certainly helps game play -- no player can consider the dozens (sometimes over 100) stones all as individuals and come up with a coherent game play. Chunking is such a natural and useful way of thinking.
But watching AlphaGo, I am not sure that's how it thinks of the game. Maybe it simply doesn't do chunking at all, or maybe it does chunking its own way, not influenced by the physical world as we humans invariably do. AlphaGo's moves are sometimes strange, and couldn't be explained by the way humans chunk the game.
It's both exciting and eerie. It's like another intelligent species opening up a new way of looking at the world (at least for this very specific domain). and much to our surprise, it's a new way that's more powerful than ours.
> It's both exciting and eerie. It's like another intelligent species opening up a new way of looking at the world (at least for this very specific domain). and much to our surprise, it's a new way that's more powerful than ours.
I have been watching Myungwan Kim's commentary for the games - and it seems notable that a few moves he finds very peculiar immediately when they are made, he will later point out to as achieving very good results some 20 moves later. So it also seems quite possible that AlphaGo is actually reading this far ahead, to find those peculiar moves achieve better results than from the more standard approaches.
Whether these constitute a 'new way' or not I think depends highly on whether these kind of moves can fit into some general heuristics useful for considering positions, or whether the ability to make them is limited to intelligence's with extremely high computational power for reading ahead.
> he will later point out to as achieving very good results some 20 moves later
This. It's a fairly common feature of any AI that uses some form of tree search/minimax, and the effect is very pronounced in chess. Even the best human players can only think 6-8 plies into the feature versus ~18 for a computer. What we can (could?) do is apply smarter evaluation functions to the board states resulting from candidate plays and stop considering moves that look problematic earlier in the search (game tree pruning). AI tends to use very simple evaluation functions that can be computed quickly. They do so given that 1) it allows for deeper search, and a weak heuristic evaluated far in the future often beats a strong one evaluated a few plies prior and 2) for some games (like Go) it's really hard to codify the "intuitions" that human players speak of.
Because search based AI considers board states __very__ far in the future, the results are often completely counterintuitive in a game with an established theory of play. Those theories are born of humans, for humans.
The introduction of MCTS some years back was the first leap towards a human level Go AI (incidentally, MCTS is more human-like than exhaustive tree search in that it prunes aggressively by making early judgement calls as to what merits further consideration). AlphaGo's use of deep policy and evaluation networks to score the board is very cool, and the next step in that journey. What's interesting to me is that, unlike chess AI, AlphaGo might actually advance the human theory of Go. It's possible that these "strange moves" will lead to some very interesting insights if DeepMind traces them through the eval and policy networks and manages to back out a more general theory of play.
Yes, but pruning (not considering everything) is as old as game tree search. Previous Go AIs used MCTS as well. What's new in AlphaGo is a more sophisticated approach to scoring game boards - policy networks that help the AI prune even more aggressively, and a value network that's used to "guess" the winner in lieu of searching to endgame. Note that guessing the winner is just a special case of an evaluation function. For any game, if you could consistently search to the end, your evaluation function is always a -1/1 corresponding to lose/win. AlphaGo is still using MCTS - just a more sophisticated form.
I think that Chess machines play perfectly for the next 8 moves, but don't necessarily sense the importance of a Knight Outpost (which may have relevance 20 moves ahead. A proper Knight Outpost will remain a fork threat for the rest of the game).
It is far easier for a Human to beat a Chess Machine at positional play (ex: a backwards pawn shape will probably be a problem at endgame, 30+ moves from now) than to beat a Chess Machine at tactical play (3 moves from now, I can force a fork between two minor pieces)
This was true 10-15 years ago. It is no longer true. Chess engines have positional evaluation algorithms that have been trained using many millions of games, and the weighting parameters for different kinds of positional features have been adjusted accordingly.
Do some reading on Stockfish for example if you doubt the veracity of my statement.
But its just as you say: its weighting parameters and heuristics. When Stockfish recognizes a backwards pawn, it deducts a point value. When Stockfish recognizes "pawn on 6th row", it adds a point value to that pawn.
But that's a heuristic. A trained heuristic using games, but still comes down to what I understand to be a +/- point value (like... +35 centipawns).
In contrast, a chess engine truly knows that if you do X move, it will force a Rook / Minor piece exchange in 8 moves.
When you play positionally vs Stockfish, you're arguing with a heuristic (a heuristic which has been refined over many cycles of machine learning, but a heuristic nonetheless that comes down to "+/- centipawns") . When you play tactically vs Stockfish, it is evaluating positions more than a dozen moves ahead of what is humanly possible.
When you play against Stockfish in endgame tablebase mode, it plays utterly, and provably, perfectly.
Take a pick of what game you want to play against it. IMO, I'd bet on its positional "weakness" (yes, it is still very strong at positional play, but it is the most "heuristical" part of the engine)
It seems that you are trying to create a new word that describe this new way of looking at the world. If human are able to decode the information contained in those unexpected moves, perhaps by creating a new heuristic, that could be viewed as a way of understanding the features the machine use internally, that is reading the machine brain. If human are able to decode that information creating new heuristics we could say that we are in a new state in IA in which learning among different intelligent species should be studied.
It's also possible that the positional evaluation is strong enough that AlphaGo can see the value in a position before the human because of the complexity involved in determining the "value" of a given position.
My experience is with Chess and Chess AI, but in my experience, the more positional knowledge built into the evaluation function, the better the search performs, even if you have to sacrifice some speed for more thorough evaluation. A significant positional weakness may never be discovered within the search horizon of a chess engine because it may take 50 moves for the weakness to create a material loss, so while it's certainly possible that a deep, but carefully pruned search is being utilized, I suspect that some of the Value Network's evaluation is helping to create some of these seemingly odd moves.
For AlphaGo to recognize a position that doesn't achieve a good result for 20 moves, it would often have to search much deeper than those 20 moves (I'm not sure if you're using the term moves to mean ply or both players moving, but if it takes 20 AlphaGo moves for the advantage to materialize, that would be a minimum 40 ply search) to quiesce the search to the point that material exchanges have stopped (again, this is how chess typically does it, I don't know about Go), so the evaluation at the end of the 20 move sequence is arguably more important than a deep search. The sooner you can recognize that a position is good or bad for you, the more time you have to improve the position.
I would imagine it's absolutely thinking that far ahead. That said, it can't possibly search every possible solution, just needs to find an adequate one
There's also the fact that some of the unexpected moves were apparently more about solidifying against a loss than increasing the magnitude of a win. Which has its own kind of eerie implication: since AIs (like all computer programs) do what you say, not what you mean, the "intelligent species" can sometimes work really intelligently towards a goal that wasn't quite what you had in mind. (Gets especially interesting for any AlphaHuman/AlphaCEO/AlphaPresident successors that are given goals more complicated & nuanced than "maximize Go win probability regardless of ending score". BTW, if you haven't already read the Wait But Why series on the future of AI, I recommend it: http://waitbutwhy.com/2015/01/artificial-intelligence-revolu...)
About a year ago I wrote an AI to play the board game "Hive" (shares some similarities with chess). Because I scored all wins equally, it behaved almost exactly like this. It would simply try to minimize my advantage while always keeping open the possibility for it to win, almost like a cat toying with prey. It never actually would make the winning move – however obvious – until it had no other options!
I fixed this behavior by scoring earlier wins higher than later wins. Now it will actually finish games (and win), but almost invariably its edge is very small, no matter how well or poorly I play. Because of the new win scoring, it willingly sacrifices its own advantage if it means securing a win even one turn earlier. (And since scoring is symmetrical, this has the added advantage of working to delay any win it sees for me, thus increasing the possibility of me making a mistake!)
I suppose I could try modifying the scoring rules again, to weight them by positional advantage. A "show off" mode if you like :) And again, with the flip side of working to create the least humiliating losses for itself.
In go, the purpose is to have more territory than the opponent. There is no point in humiliating the opponent by having a big advantage. I think the aim of the strange moves was to increase the confidence of the program in its advance, not to increase the advance.
Sorry, I didn't mean the intent would be to humiliate, just the appearance.
Humans, I think, have the natural instinct to "hedge" themselves in games like go and chess, by creating positional/material advantages now to offset unknowns later. Of course, that advantage becomes useless in the end game, when all that matters is the binary win/lose.
An AI, which may have a deeper/broader view of the game tree than its human opponent (despite evaluating individual position strength in roughly the same manner), may see less of a need to "hedge" now, and instead spend moves creating more of a guaranteed advantage later (as you suggest). And indeed, my experience with my AI is that during the endgame (in which an AI generally knows with certainty the eventual outcome of each of its moves), it tends to retain the smallest advantage possible to win, preferring instead to spend moves to win sooner.
> Humans, I think, have the natural instinct to "hedge" themselves in games like go and chess, by creating positional/material advantages now to offset unknowns later. Of course, that advantage becomes useless in the end game, when all that matters is the binary win/lose.
That's actually an excellent way to win chess games. Keep your eye on the mate while the other person is focusing on position and material.
> I think the aim of the strange moves was to increase the confidence of the program in its advance, not to increase the advance.
Absolutely. Also worth noting that it may be simply unable to distinguish between good and bad moves if both outcomes lead to a win, since it has no conception of the margin of victory being important.
So it might not be that it increased win probability, but that both paths led to 100% win probability and it started playing "stupidly" due to lacking a score-maximizing bias.
But you could indeed humiliate the opponent by actually capturing ALL of his stones.
But that won't happen, if the enemy knows at least the basic concepts ...
Still, if you play well, you cover much ground - while trying to supress the area of the enemy and even crushing him.
But classic go is nice in a way, that it gives weaker opponents a start bonus of some stones - so the game is balanced and domination usually won't happen ...
My brother once played the (then) British Youth Go Champion on a 13x13 board, and lost by around 180 points - literally scoring worse than if he hadn't played at all.
The AI is based on the minimax algorithm [1]. Because of the way Minimax works, the only way for a possible next move to be designated a "win" is if it is a guaranteed win. (The tree is (effectively) fully explored, and the opponent is given the benefit of the doubt in the face of incomplete information.) So, if there are multiple such winning moves, and care is not taken to distinguish the "magnitude" of the win, the AI will choose one arbitrarily.
I suppose that, in Hive, it is more likely that a path to a win is longer rather than shorter. Hence, when my AI was arbitrarily choosing "winning" moves, it statistically chose those that drew the game out.
Usually it's because as it searches the move tree, it finds ways for the opponent to maximize their own winning probability and so has to hedge against that. In minimax games sometimes the evaluator finds a long chain of moves that leads to a win, and once it finds that, doesn't necessarily bother trying to find a shorter one. It can be frustrating to tune that out.
> There's also the fact that some of the unexpected moves were apparently more about solidifying against a loss than increasing the magnitude of a win.
Humans play that way too. Everyone wants to maximize the chance of leading by >=1 stone.
The difference is that AlphaGo is better at calculating a precise value of a position, so that when uncertainly plays in, AlphaGo can play for, say, "1-3 stone lead", while a human can only get confidence in "1-7 stone lead", and thus needs to play excessively aggressively to overcome the uncertainty.
Right. Skynet and Terminator are science fiction, but the slippery, unpredictable reality of how computers actually behave is right in front of your eyes as a programmer every day. Sometimes I wonder if science fiction writers do more harm than good: once they make a movie about some possible future, people feel free to dismiss it as "just science fiction", even if they have easily available empirical evidence that something vaguely like the scenarios described actually kinda has the potential to occur.
Not unlike the Simpsons episode where the military school graduation speech tells them the wars of the future will be fought with robots and that their jobs will be to maintain those robots.
thats .. unlikely. this could only happen if two wealthy and highly developed nations nations want to make a spectacle out of a war.
if you have fully autonomous robots which can fight your war, you'd be able to launch a massive offensive within hours. properly mobilizing defenses and responding to that invasion would take too long, as any command centers would've already been wiped out by the first attack.
I think AlphaGo is playing very natural go! The 5th move shoulder hit that is the subject of so much commentary would fit into the theory of go that players like Takemiya espouse. It has chosen to emphasize influence and speed and has not been afraid to give solid territory early in the games so far. It's very exciting play but not inhuman play, and if professionals are allowed to train with AlphaGo it will surely usher in the next decade's style of play. Don't forget that the game has changed every 10 years for the past 100 years, it should not be surprising that it is continuing to change now!
It didn't look like a Takemiya-style move to me. Takemiya tends to play for a huge moyo in the center. AlphaGo had no such moyo. It wasn't only a strange move; it was also a strange time to play it, and it definitely went against conventional wisdom.
The result of the shoulder hit coordinated with black's bottom formation, and the extension on the 4th line that threatened to cut white's stones off was flexible and could have easily formed an impressive moyo on the bottom. It did not play out that way, but I think that black's strategy was as cosmic as anything Takemiya might have played. His games did not always end with a giant moyo, he was also very flexible. I hope to see written reactions from professional players, and maybe Takemiya will give AlphaGo's style his endorsement :)
Some examples of 5th line early shoulder hits in recent professional play - these situations are not the same as the one seen in today's game, but something like a 5th line shoulder hit is always going to be highly contextual and creative.
Those games are really interesting. In the first two, they are both ladder-breakers played by stronger players; my guess is the weaker players set up the ladders assuming that the stronger players wouldn't play a fifth line shoulder hit to break them, and the stronger player didn't back down. In the third game, the fifth line shoulder hits aren't that surprising; they're reductions against frameworks that were allowed to get big in exchange for growing an opposing framework; they're locally bad moves but the global benefits are clear; you'll note that both players play a fifth line shoulder hit.
The only one I can't parse is the last one. There are a lot of variations where I want to know what black's plan is.
There's an interesting angle to this phrase "intelligent species opening up a new way of looking at the world", which is that we (humans) designed go as a game - a subset of the real world we interact with. Go is "reality" to alphago. The superset of all possible sense data it could have, in principle. Whatever "chunks" AlphaGo uses, if it does use them, all of its policies are built only from subsets of the sense data that is the interactions (self-plays) and inferences from past games. There's nothing outside the game to bring into its decision process. With humans, however, our policies are noisy and are rife with what, for lack of a better term, I would call leaky abstractions.
I think it's more metaphor than leaky abstraction in this case, except to the extent that metaphor is mapping an abstraction of a domain we are trying to understand to an abstraction of one we are better able to understand.
Sometimes optimal solutions don't make sense to the human mind because they're not intuitive.
For instance, I developed a system that used machine learning and linear solver models to spit out a series of actions to take in response to some events. The actions were to be acted on by humans who were experts in the field. In fact, they were the ones from whom we inferred the relevant initial heuristics.
Everyday, I would get a support call from one of the users. They'd be like, 'this output is completely wrong. You have a bug in your code.'
I'd then have to spend several hours walking through each of the actions with them and recording the results. In every case, the machine would produce recommended actions that were optimal. However, they were rarely intuitive.
In the end, it took months of this back and forth until the experts began to trust the machine outputs.
This is the frightening thing about AI - not only can an AI outperform experts, but it often makes decisions that are incomprehensible.
What you said about the expert calling something a bug reminded me of how the commentator in the first game would see a move by alphaGo and say that it was wrong. He did this multiple times for alphaGo but never once questioned the human's move. Yet even with all those "wrong" moves alphaGo won. Didn't watch the second game, so not sure if he kept doing that.
The english-speaking human 9-dan only did this once for AlphaGo yesterday (when AlphaGo made an "overextension" which eventually won the AI the game), but maybe did it approximately 3 or 4 times for Lee (Hmm, that position looks a bit weak. I think AlphaGo will push his advantage here and... oh, look at that. AlphaGo moved here).
Later, he did admit that the "overextension" on the north side of the board was more solid than he originally thought, and called it a good move.
He never explicitly said that a move was "good" or "bad", and always emphasized that as he was talking, his analysis of the game was relatively shallow compared to the players. But in hindsight, whenever he point out an "bad-juju feel" on the part of Lee's move, AlphaGo managed to find a way to attack the position.
Overall, you knew when either player made a good move, because the commentator would stop talking and just stare at the board for minutes, at least until the other commentator (an amateur player) would force a conversation, so that the feed wouldn't be quiet.
The vast, vast majority of the time, the English-speaking 9-dan was predicting the moves of both players, in positions more complicated than I could read. (Oh, but it was obvious both players would move there. There were clearly times when the commentator would veer off into a deep distant conversation with the predicted moves still on the demonstration board, because he KNEW both players were going to play out a sequence of maybe 6 or 7 moves forward).
They really got a world-class commentator on the English live feed. If you got 4 hours to spare, I suggest watching the game live.
Elsewhere in this thread, IvyMike pointed out [1]:
> I sense a change in the announcer's attitude towards AlphaGo. Yesterday there were a few strange moves from AlphaGo that were called mistakes; today, similar moves were called "interesting".
Or, maybe, there could have been bugs in the code.
If I'm an expert in some domain and a computer is telling me to do something completely different ("Trust me--just drive over the river!") I'm certainly going to question the result.
As a competitive speedcuber (Rubik's Cubes) this makes sense. If I watch a fellow cuber solve a cube, I understand their process even if it's a different method than the one I'd use. But a robot solving it? To my brain it looks like random turns until...oh shit it's finished.
Have you ever managed to learn the human Thistlethwaite algo? It basically lets you solve the cube like a robot would. I'm pretty rusty at cubing nw, but I always wanted to learn it.
> AlphaGo plays some unusual moves that go clearly against any classically trained Go players. Moves that simply don't quite fit into the current theories of Go playing, and the world's top players are struggling to explain what's the purpose/strategy behind them.
Could AlphaGO be winning in a way similar to left handed fencers having an advantage over right handers by wrong footing them rather than simply being better? Would giving Lee more chance to see this style give him a chance to catch up?
I'm not a Go player but play other competitive sports. Humans have a herd mentality...as Op mentioned there's certain styles of playing...which has their own strengths and weaknesses. Sometimes people will not examine other styles that may have better strengths and just focus on the exist one. Then comes along someone who 'thinks outside the box' with a new style and revolutionize the playing field.
Think Bruce Lee and the creation of Jeet Kune Do. Before him everyone concentrated on improving one style by following it classically, rather than just thinking of 'how do I defeat someone'.
IMHO Lee is the best at the current style of Go. AlphaGO is the best at playing Go. Maybe humans can devise a better style and defeat AlphaGo, but I'm sure AlphaGo can adapt easily if another style exists.
Swimming. It used to be that swimmers were supposed to be streamlined and avoid bulky muscles. Then a weightlifter decided he wanted to swim. Swimmers today all lift weights.
Programming. It used to be that people built programs in a very top down, heavily planned way. Think waterfall. We now understand that a highly iterative process is more appropriate in most areas of programming.
Expert systems. It used to be that we would develop expert systems (machine translation, competitive games, etc) through building large sets of explicit rules based on what human experts thought would work. Today we start with simple systems, large data sets, and use a variety of machine learning algorithms to let the program figure out its own rules. (One of the giant turning points there was when Google Translate completely demolished all existing translation software.)
Serve-and-volley is pretty much non-existent in modern professional singles tennis. We were always taught to attack the net, and every action was basically laying the groundwork to move forwards and attack.
Nowadays, top players slug it out baseline-to-baseline.
In terms of stance, we were taught to hit from a rotated position where your shoulder faces the net, and a normal vector from your chest points to either the left or right side of the court.
Nowadays, it's much more common to hit from an "open" position, where your body is facing the net, not turned. This would have been considered "unprepared" or poor footwork in my day, but it actually allows for greater reach. It does make it more difficult to hit a hard shot, but that's made up for by racquet technology and generally stronger players.
If you're in the mood for some long form literary tennis journalism about this subject, check out David Foster Wallace's Federer as Religious Experience from 2006.
> AlphaGO is the best at playing Go. Maybe humans can devise a better style and defeat AlphaGo, but I'm sure AlphaGo can adapt easily if another style exists.
Which is a curious point. The gripes about early brute force search algorithms (e.g. Deep Blue?) were that they felt unnature.
However, as the searches get more nuanced and finely grained, is there a point at which a fast machine begins doing fast stupid machine things quickly enough to feel smart?
Are there any chess / Go analogs of the Turing test? Or is a computer players always still recognizable at a high level?
It has been said that a game of Go is a conversation with moves in different areas showing disagreement. The game is also known as 'shou tan' (hand talk). From the commentary, AlphaGO is currently passing the Go Turing Test in almost all cases. There are some moves which some say are uncharacteristic, then later play out well. Or so called mistakes not affecting the outcome of the match. One explanation given was that AlphaGo optimizes for a win, not win by greatest margin, which is a/most valid for human or machine.
Computer players will be recognizable as long as they are designed to win, and not to play the way a human plays.
A Turing test for game players is an interesting idea, it would be useful for designing game players that are good sparring partners rather than brutes that can whipe the floor with you.
Bruce Lee played it very smart and attained a guru status in the West, but there's no evidence he was a world-class fighter, only unsubstantiated claims by his entourage.
As for JKD, people are drawn in by its oriental esotericism, but there's no evidence it is an especially effective fighting style, or that it has something that (kick)boxing does not.
Remember that AlphaGo has spent months developing its own style and theory of the game in a way that no human has ever seen. Its style is sure to have weaknesses, but humans will have a hard time figuring them out on first sight.
Similarly chess computers do better in some positions than others (they love open tactics!) and one of the games that Kasparov won against Deep Blue he won by playing an extreme anti-silicon style that took advantage of computer weaknesses. However Kasparov didn't have to figure out what that style was because there was a lot of knowledge floating around about how to do that.
Therefore I'd expect that Lee Sedol from a year from now could beat AlphaGo from today. And human Go will improve in general from trying to figure out what AlphaGo has discovered.
However that won't help humans going forward. AlphaGo is not done figuring out the game. At its current rate of improvement, AlphaGo a year from now, running on a single PC, should be able to beat the full distributed version of AlphaGo that is playing today. Now the march of progress is not whether computers can beat professionals. It is going to be how small a computing device can be and still beat the best player in the world.
But when the weaknesses it has require looking 20 ply into the game, can anyone exploit those weaknesses? And furthermore, if the computer itself is able to see 20 ply into the game, then it can spot its own weaknesses and you need to look even further, making the question of whether it's really a weakness.
Weaknesses are only relative to capabilities of the opponent to exploit them. If a tank has a weak spot that rockets can hit, but it's being opposed by humans on horseback, is it really a weakness in that context?
The weaknesses that it has will be of the form that it has wrong opinions about certain kinds of positions. In the case of chess, those weaknesses showed up in closed positions where the program owned the center and large amounts of space. In the case of AlphaGo, the weaknesses will be much more subtle, but will be discoverable and exploitable in time.
Additionally AlphaGo has the advantage that it started with a database of human play, so it has some ideas what kinds of positions humans miscalculate.
As for your tank vs horseback analogy, that's flawed at the moment. AlphaGo is probably reasonably close in strength to the human facing him. Improved human knowledge could tip the balance.
However in the future it will become an apt analogy. Computers are going to become so good that knowing the relative weaknesses in their style of play may reduce the handicap you need against them, but won't give you a chance of becoming even with them. That happened close to 20 years ago in chess, and is now only a question of time in Go.
I wonder if AlphaGo has some specialized knowledge to handle ladders, where stones can have an effect at a distance that might only come into play after 20 moves.
20 moves may sound like a huge number of variations but when you prune things early, it can be quite manageable. The alphabeta algorithm in chess does pruning quite a lot.
I would also posit that lefties' advantage basically disappears once you get to a certain level in fencing. Past some point, it's basically all just footwork anyways, and your orientation doesn't change the distance of your target (foil and saber at least, can't comment on epee as they seem to just kinda bounce in place a lot even at Olympic level).
Can confirm. My brother fenced at a club that had a lot of lefties. All the righties got used to it quickly, and had no real disadvantage when playing against lefties.
I could easily see the difference in tournaments with other clubs that were not used to left handed players.
Seems unlikely. Training was partly from human games, and partly from self play; if there's some new, off book heuristics at play, there's no way to know that humans would respond poorly to them. Though I suppose it's possible it would notice that humans do poorly on simply off book moves generally.
Why does this seem unlikely? Humans do poorly with "off book" moves in general in sports and other games; it's why new styles of play or management work really well until others get used to them. Why would it be unlikely in Go?
I think an important point was brought up by the Google engineer in the beginning of the game: Humans usually consider moves that put them ahead by a greater margin and base their strategies on that, while computers don't have that bias.
Building on that, I suspect that if AlphaGo thinks it has a 100% chance of winning with any of several moves, it has no way of distinguishing between them and chooses effectively at random. The longer that goes on - and once it hits 100% chance of winning, it will be that way for the rest of the game - the more chances it has to pick bad moves. As long as the move isn't bad enough to ruin its 100% chance of winning, it can't tell the difference between that and a good move.
(This also applies without a 100% chance of winning, as long as its chances of winning hover near the highest percent it's able to distinguish.)
I doubt the value network ever outputs a literal 100% chance of winning, it would at most be a lot of nines.
Even if it did output an actual 100% chance, AlphaGo would still end up picking moves favored by the policy network, so it would probably just revert to playing like it predicts a human pro would.
Once it gets to enough nines, its monte carlo trees will run out of sample resolution. If it can resolve to three nines, then a 99.93% win branch has a 70% chance of being reported as 99.9% and a 30% chance of being reported as 100%. When all the branches here get rolled up, they report some average around 99.93% but not necessarily exactly it. This propagates upwards in the tree, adding more meaningless digits. Adding the evaluation network in increases the number of decimals, but doesn't really change the effect.
It's similar to how ray tracing renderers start to return weird speckle patterns when the room is dark enough.
And the policy network chooses branches to investigate, not which one to choose. It adds sample resolution to places pros might play, but doesn't add to the estimated probability of winning.
Edit: Actually, since places pros might play have higher sample resolution, they're less random. So worse moves get worse evaluation, and a higher chance of leading the pack. This might actually bias AlphaGo to play some pretty bad moves - but, again, this is all assuming it's going to win anyway.
> I've been giving it some thought. When I was learning to play Go as a teenager in China, I followed a fairly standard, classical learning path. First I learned the rules, then progressively I learn the more abstract theories and tactics. Many of these theories, as I see them now, draw analogies from the physical world, and are used as tools to hide the underlying complexity (chunking), and enable the players to think at a higher level.
The excellent point you're making applies in general to nearly every type of human thinking.
The way we think about other people, our intuitions about probabilities, our predictions about politics, and so on -- all are based on our peculiarly effective, yet woefully approximate, analogy based reasoning.
It shouldn't be surprising in the least when commonly accepted "expert" heuristics are proved wrong by AIs that actually search the space of possibilities with orders of magnitude more depth than we can. What's surprising -- and I think still a mystery -- is how human heuristics are able to perform so well to begin with.
I'm not a Go player, but I saw this same phenomenon as poker bots have surpassed humans in ability. As with AlphaGo, they make plays that fly in the face of years of "expert" wisdom. Of course, as with any revolutionary thinking, some of the new strategies are "obvious" in hindsight, and experts now use them. Others seem to require the computational precision of a computer to be effective in practice, and so can't be co-opted. That is, we can't extract a new human-compatible "heuristic" from them -- the complexity is just irreducible.
They are peculiarly effective only because of lack of comparison. Humans have been the most intelligent species on this planet for millennia, where no other species come even close. We don't know how ineffective those strategies are seen by a more advanced species. Well, until now.
This is a good point. I was coming from the point of view that we've had powerful computers for a while, and yet humans were still dominating them, at least until recently, in games like Go, poker, and many visual and language tasks.
Of course, the counterpoint could be that it's only the case because humans, with their laughable reasoning abilities, are the ones programming those computers.
AlphaGo can’t decide that it’s bored and go skydiving. Humans aren’t merely capable of playing Go. And when they do it, they can also pace around the table, and drink something, all at the same time, on a ridiculously low energy budget. Or they can decide never to learn Go in the first place but to master an equally difficult other discipline. They continuously decide what out of all of this to do at any given moment.
AlphaGo was built by humans, for a purpose selected by humans, out of algorithms designed by humans. It is not a more advanced species. It’s not even a general intelligence.
Your own original point was much better than the one made in response.
I want to thank you for this comment. It's this kind of subtle, low-key, informed speculation that generates good, hard sci-fi concepts, which are absolutely relevant to my WIP novel.
"oh what if the machine suddenly came alive!?" has been done 1000 times. But such concepts like: a computer can detect and act patterns which we cannot, in ways that are almost, if not possibly intelligence, are magnitudes more believable, and therefore, compelling.
Is it about a tyrannical super-AI that maintains power over the human race by strategically releasing butterflies into the wild at specific times and locations?
Actually, it's a Soviet knock-off of a PDP-10. Constructed in 1970's India, the machine has a 12mhz clock rate, a 4M of RAM, and a directive to bring about "World Peace".
Of course, those fools underestimated it. They should have known better...
AlphaGo is essentially built on the work that IBM did on TD-Gammon (a reinforcement learning backgammon player) in the 90s.
Pretty much the same thing happened with TD-Gammon with it playing unconventional moves, in the longer term humans ended up adopting some of TD-Gammon's tactics once they understood how they played out, it wouldn't be surprising to see the same happen with Go.
From my understanding, computers have also had this affect on chess. The play styles of younger champions has evolved to the point where unpredictability is actually part of the strategy. I'm not a chess expert by any means, but this quote by Viswanathan Anand (former World Chess Champion) describes it.
“Top competitors who once relied on particular styles of play are now forced to mix up their strategies, for fear that powerful analysis engines will be used to reveal fatal weaknesses in favoured openings....Anything unusual that you can produce has quadruple, quintuple the value, precisely because your opponent is likely to do the predictable stuff, which is on a computer” [1]
>powerful analysis engines will be used to reveal fatal weaknesses in favoured openings...
Anand isn't really talking about strategy here, he's just talking about choice of opening. Players with narrow opening repertoires, like Fischer, have always been easier to prepare for than players who play a wide variety of openings.
As far as actual changes to strategy, the most obvious one is that computers tend to value material more highly than humans. So a computer will take a risky pawn if it looks sound, while a human will see that taking the pawn is very complicated and prefer a simpler move.
I wonder if this is similar to how musket battles were fought in the american civil war era, with soldiers lining up across each other in a battlefield and taking turns shooting at each other. I hear they did this because the rifles were very inaccurate so it made sense to use a bunch of them at the same time as an area-effect weapon, in effect like a gigantic shotgun.
Until someone got better weapons and suddenly the "rules" of the battlefield that dictated standing in lines across each other made no sense to follow anymore because the original principles that dictated those rules to be good were not valid anymore.
I like your statement: "It's both exciting and eerie. It's like another intelligent species opening up a new way of looking at the world (at least for this very specific domain). and much to our surprise, it's a new way that's more powerful than ours."
I think this will the theme of our future interactions with AIs. We simply can't imagine in advance how they will see and interact with the world. There will be many surprises.
That quote reminds me of "The Two Faces of Tomorrow" by James P. Hogan. One of the subplots is that humans can communicate because of shared experience. We all must eat, sleep, breathe, seek shelter, etc. Communication with an alien or artificial intelligence may be difficult or even impossible without this shared framework.
>It's like another intelligent species opening up a new way of looking at the world (at least for this very specific domain). and much to our surprise, it's a new way that's more powerful than ours.
It's not like this at all; let's not do this sort of thing. Humans are inveterate myth makers (viz. your description of how people conceive the Go board as army units), and our impositions on the world are easily confused for reality.
In this case, there's no "intelligent species" at work other than humans. We made this, and it is not an intelligence, it is a series of mathematical optimization functions. We have been doing this for decades, and these systems, while sophisticated, are mathematical toys that we have applied. We built and trained this thing to do exactly this.
As a student of AI you know that convolutional neural networks are black boxes and are hard to interpret. A different choice of machine would have yielded more insight about how it is operating (for example, decision trees are easier to interpret). The inscrutability of the system is not a product of its complexity; even a simple neural network is hard to understand.
This, actually, is my primary objection to using CNNs as the basic unit of machine learning - they don't help US learn, they require us to put our faith in machines that are trained to operate in ways that are resistant to inspection. In the future I hope that this research will move more towards models that provide interpretable results, so they ARE actually a tool for improved understanding.
> We made this, and it is not an intelligence, it is a series of mathematical optimization functions
You can say the same about your mind too which is a bunch of optimization nodes. If something is intelligent, does it matter if it's evolved in nature or created by a species who is evolved in nature?
> In the future I hope that this research will move more towards models that provide interpretable results
I think it's not really possible to understand in detail how these networks operate on the level of nodes, because emergent behavior is necessarily more complex than the sum of its parts.
It's a bit precious I think to say that a human is a "bunch of optimization nodes". I can write code to create a CNN, and I can draw a graph of how it operates on a piece of paper. We can't even decode a few million rat neurons the same way.
A CNN is a pure mathematical function - if you want, you could write it down that way. Given a set of inputs, it will always produce the same output. We don't call a linear regression model an "intelligence", a CNN is no different.
Of course I agree that humans are built up of billions of tiny machines like this, but let's appreciate the vast difference in scale.
My exaggeration was intentional to point out that if you scale up NN based systems, we are not that different :) I do appreciate it, but let's not forget that we have finite nodes, so at one point a machine can surpass us with "just mathematical functions".
> A CNN is a pure mathematical function
That's their basic property, but who are we to say that our cell based neural network is superior? Cells are just compositions of atoms and they are defined by quantum mechanics, which is... "just" math and information.
I also think that Go might be a great communication tool between AI and humans. If you look at the commentary from this angle if's fun to think about like this.
As a follow up to your idea, we should explore two paths: first create the most powerful AI, second create subsystems devised to be interpretable. The powerful method could be used to train the interpretable method, that is we need an interpreter to translate from machine AI to human AI, and interpretable systems provide a middle ground.
I think training one function to approximate another function wouldn't help much; we'd inevitably lose the subtleties of the higher-order function and any insights that come with it. If we could train a decision tree to do what a CNN does and then interpret the outcome, why not use decision trees in the first place?
I think the answer must be in figuring out how to decompose the black box of a CNN - it is, after all, just a set of simple algebraic operations at work, and we should be able to get something out of inspection.
I have to imagine Hinton et al. have done work in this regard, but this is far afield for me, so if it exists I don't know it.
Having a machine that gives you feedback in the middle of the game perhaps could be used to describe what is the weak point of a decision tree, and in which situations the method is good. It could detect some situations in which decision trees are good, then use that decision tree to understand what is happening and with that new understanding devise a new method in the middle. We could train a decision tree using new very powerful information about the value of the game in the middle of the game, that is new and powerful.
The same thing happened in chess. Computers play in a very "computerish" way that was initially mocked, but became hugely influential on how humans play chess. Computer analysis opened up new approaches to the game.
> It's like another intelligent species opening up a new way of looking at the world.
And this is just the beginning with AlphaGo. As we keep on training Deep Learning systems for other domains, we'll realise how differently they approach problems and solve them. It'll, in turn, help us in adapting these different perspectives and applying them to solve other problems as well.
I believe that when Google talked last year about DeepMind playing those 70's Atari games, it also surprised the team with some of the tricks that it learned to be more effective in the game. So this is quite interesting stuff.
The analogy I can come up with, based on your post, is of something like addition. We don't know how we add numbers in our heads; but we somehow do it. Some people can do it very, very quickly[1], but won't be able to explain how they did it. On the other hand: a computer doesn't look at digits and numbers; it just looks at bits and shifts them around as appropriate.
Abstraction is the domain we need to research before we understand intelligence in general, the ways our abstraction is determined by nature and more importantly the ways that will become possible when we surpass it.
Can you give an example of an "unusual" move? I'm a (very) novice Go player, and I think it'd be really interesting to see some specific commentary on how the machine is playing the game.
A great breakthrough could be to decode the information contained in the feature space of the nn or the rnn. A topological language in which shapes and chains are explained by analogies with real world situations and actions. Being able to share our vision and communicate our intentions (the weight given to the distinct features and the links among the several layers of the nn - the overall plan) should transform the concept of AI into one of CAI communication between intelligent agents to create a synergistic approach).
Someone somewhere asked why a lot of people in the Go community is taking this in a somewhat hard way, here is my hypothesis:
Go, unlike Chess, has deep mytho attached to it. Throughout the history of many Asian countries it's seen as the ultimate abstract strategy game that deeply relies on players' intuition, personality, worldview. The best players are not described as "smart", they are described as "wise". I think there is even an ancient story about an entire diplomatic exchange being brokered over a single Go game.
Throughout history, Go has become more than just a board game, it has become a medium where the sagacious ones use to reflect their world views, discuss their philosophy, and communicate their beliefs.
So instead of a logic game, it's almost seen and treated as an art form. And now an AI without emotion, philosophy or personality just comes in and brushes all of that aside and turns Go into a simple game of mathematics. It's a little hard to accept for some people.
Now imagine the winning author of the next Hugo Award turns out to be an AI, how unsettling would that be.
Actually this AI kind of confirms these myths, since its basis are not in mathematics but in neural networks. While it could be argued to be just math, so is the brain, but that's besides the point. The point is, even the programmers have no idea what the AI is thinking.
The way it picks moves is very similar to how top professionals do.
Intuition is reduced to memories stored vaguely as neural connections.
I fear this is an overly mystical misinterpretation of neural nets. (Standard feed-forward) neural nets are layers of non-linear feature transformations. They take inputs and at each step transform those inputs into a more compact representation, distilling the inputs into their most important factors (and throwing away unimportant factors).
So the most likely explanation is that policy/value nets in AlphaGo have learned to extract - with cold logic - the key factors that make up what humans believe to be "good" board positions.
It has little to do with voodoo about neural connections and magic emerging from the weights. AlphaGo has most likely managed to identify the important factors of good board positions (by seeing tons of examples of good and bad moves/positions). It only appears to be magical because these factors are most likely very complex and inter-dependent.
This is supported by the AlphaGo paper - they report that AlphaGo without tree search is about as good as the best tree search programs (amateur pro level). So AlphaGo has taken amateur-pro-level board analysis ability, and combined it with tree search, to achieve top-player performance..
> So AlphaGo has taken amateur-pro-level board analysis ability, and combined it with tree search
I don't think that's an entirely accurate way of putting it, because a player who reaches that level is also doing a little tree searching. Maybe if you found some human who managed to reach "amateur pro" level by playing purely on snap, instinctual decisions without any logic or exploration of variants at all, yes, then you could say their ability to evaluate positions is as good as AlphaGo's.
But I would guess that you are right anyway that we can deduce that its ability to evaluate a board position really is below that of better professionals, and its huge strength is due to the tree search (which of course involves a second "policy" net to pick moves to explore).
In go, the tree searching is named "reading". Even beginners need to read very deep trees (more than 7 moves) to anticipate the result of a Semeai (capturing race).
> distilling the inputs into their most important factors (and throwing away unimportant factors).
Isn't that pretty much the definition of intuition? Combining a bunch of things in some unknown and nonlinear way to result in a 'feeling' about the situation?
Pretty much. The nature of evaluating a Go position, is that it has some notions of similarity like islands, "aliveness", and local features that can be looked at liberty-wise. But in order to fully understand how it all pieces together, one must have seen other positions to intuit the kind of future playstyle that will result in the game. Anyhow, it's perfect for a stochaistic system that just recognizes patterns. As long as you can provide the tree search (ie. the partial position evaluation) you can basically just let it rip.
You can still beat Go using "simpler" math. If you have enough compute power, you can always just minimax the whole game tree. Neural networks aren't un-mathematical; they're just a slightly more complicated technique for discovering an approximation to a function that does what you want (even if you're not sure what that function looks like internally).
So AlphaGo's reign will last only until QuantumGo arrives on the scene? It would be sort of ironic to have spent decades developing practical AIs with classical computing to have them swept aside only when they started to really deliver results...
I mean it's not clear in anyway "where" is the intuition, it just sort of magically emerges from the connections and the weights in a way that no one can really grasp. If nothing else, just due to the sheer number of neurones.
Is this just something you're imagining, or did Google's developers explicitly say, "AlphaGo is beyond our understanding. We have absolutely no idea how it makes its decisions"?
AlphaGo developers don't understand how it works in the same way you wouldn't understand how the program you've written to find prime numbers actually found a big prime number. The sequence of operations is known, but numbers are too big to be comprehended.
Which kind of makes me think. Would machine learning succeed even mildly at recognizing primes? Would we be able to decode the final weights after weeks of learning, and find a sieve program encoded as data?
I suppose in the era of Bobby Fischer, chess was proxy for superpower one-upmanship. That's long gone, but Chess as a game is still doing fine and I expect that it will be the same for Go.
We still have chess tournaments, super-star grandmasters and circus freaks (people who can play blindfolded against multiple opponents). And, yes, computers can easily smoke all but elite players.
In chess the top engines are rated hundreds of ELO points above Magnus Carlsen (top human). No top ranked human vs computer match has been publicised in over 5 years because humans are thoroughly trounced. There are cyborg matches which are interesting. Human + Computer vs Human + Computer because gameplay techniques are considered different. Humans still depend more on higher level goal strategy and less on ruthless positional efficiency (which is probably why they get beat midgame).
What is mind boggling is that 6 months ago no go engine was scratching the surface of professional level go. It took the engines getting a 4-5 stone handicap to be competitive at the lowest level of professional levels.
It looks like this one algorithm has blown through the professional ranks in about 3 months. And a 5-0 victory here would be like 2006 vs 1996 (or even 1993) chess in 3 months.
I guess it's because Go isn't that well known in the West but I find it a bit surprising (though not really) that this isn't getting more press. When Kasparov lost it was news but not really surprising. If it wasn't Kasparov that a computer beat, it would have been the next champion. The writing was on the wall for a long time. It was just a question of when exactly.
As for Go, I guess I would never had made a long bet against computers. But as recently as just over a decade ago, computers lost to merely competent players and people working on Go programs were pretty much saying that they didn't even know what the path forward looked like. Things improved a lot with Monte Carlo but even that stalled out. Admittedly, I don't follow this area closely, but these wins pretty much came out of nowhere.
Go may not be well known in the west, but it was extremely well known in AI circles. Even before DeepBlue, Go was considered the holy grail of competitive game AI.
Oh, I'm well-aware of that and obviously this is big news on sites like this one. I was mostly remarking that this is pretty low on general news radar screens.
I see. What one means by "computer" is always a moving target. Deep Blue was a specially prepared supercomputer from 20 years ago. Apparently, today's top chess software running on a half-decent off-the-shelf machine could crush Deep Blue.
I played Chess as a kid. I watched the local tournament shrink from the big town hall, to the side room in the same building, to a local school hall, to a classroom in that school. I really do think the game is dying - perhaps something that was happening already, but Kasparov losing to Deep Blue seemed to really catalyse it.
(Not saying this is a bad thing. Evolution in games is natural, and I think it's amazing how much innovation is going on right now (particularly enabled by Kickstarter) - you'd think that board game design would have been worked out decades or centuries ago, but in the same way that incandescent bulb development accelerated massively when competition arrived, it feels like game design has got so much better when forced to compete with computer games. If there are other activities that people find more fun than Chess, that's all to the good)
I don't believe there is a relation between the improvement of computer chess and the reduction in the availability of in-person chess.
I also played chess as a kid and the allure of both local and national tournaments was that you could play with a multitude of different players, as opposed to the same 4 or 5 habitual chess players in your family/school/circle of friends.
But now, with the internet, at any second you can play with different people from all over the world, different strengths, styles and whatnot.
Hence now, instead of looking for the local chess club in the weekends we can play, any time of the day, any day of the week, anywhere.
Sites like the excellent lichess [1] are even free (in this case, free both as in beer but also as in speech) and, at any moment there there are 9 thousand, 10 thousand players enjoying this magnificent game.
Because it is far more complex.
But apparently self learning KI has advanced enough.
And yes, chess has lost some reputation. And I guess it will be similar with go. I mean there is a University just for go.
But learning something where you know you can become the best, is something different, than learning something knowing computers will be allways better than you ... so I guess they are having a hard time right now ..
Yeah, can confirm. I am not a Go player, and didn't know much about it. But top Go players are really well respected as some of the highest talents in the society, almost like being feared. It is smart people's game, eventually.
And the fact, that, prior to the deep learning revolution, Go is the only board game that human cannot be beaten, add even more myth and charm to the game and players alike.
Now, it comes to the time, that Go can be modeled by computers, and hundred years of human study is topped by computer in less than a year's time. All those myths around it will be gone. That is the biggest bummer I guess.
How well would the computer fare with a slightly different game -- something like go, but with differing rules? Would a smart human learn faster (in real time, or alternatively with comparable energy use) than an artificial reinforced deep learning system?
And who can make the most interesting new go-like game?
Perhaps this could be tested with chess or checkers, even.
On a more realistic side note... Professional Go players devote decades in training ever since their youth, giving up normal educations and lots of other more lucrative opportunities for their lives. It's very easy to imagine their frustrations now that their life-time devotion actually means nothing in front of the AI.
It's an upright denial to the way of life they so chose and devoted.
IMHO Google should donate the prize towards Go education and Go organizations instead of some random charities.
Isn't this a good thing? Why are high IQ people devoting their entire lives to a game? Maybe this will make them shift their priorities to solving problems that only really smart humans (like them) can solve.
At the root of it, they earn a living by being entertainment. This can be applied to any of the arts or sports. Why are smart people making movies, writing fiction, making music? I think these are the sorts of things that make life worth living.
Abstract strategy games require highly domain specific skills. These skills do not transfer to other endeavors. The world champion Go player might just end up as, had he not played Go, a mid-level lawyer or manager. Who knows. Source: https://books.google.com/books?id=nCMWxjkTAvEC&pg=PA130&lpg=...
Didn't happen in Chess, won't happen in Go. Entertaining a few million people is too lucrative. Everyone wants to cheer for their country in an international competition, so it's always going to have large prize pools.
Uh, talent for Go doesn't translate automatically into talent for math, physics, finance or other branches of science. Even if they are, being the top Go player is probably more attractive than being a meh quant or programmer.
This could equally apply to the bankers -- and the software engineers who enable them -- who crashed the economy in 2008. Go and chess players have contributed much more to the world than these psychopaths.
Perhaps Go playing is on it's way to being one of the first white collar jobs to be lost to AI.
I don't think people will pay to watch Go Bots square off, but I think this example of "obsolete education" is a great reminder that it's not just the assembly line jobs on the chopping block.
Google, should they win, is donating their money to Go charities, STEM education and UNICEF.
So they're doing what you want them to (I can't find a summary of how they're allocating the money across each category). Personally I think the work UNICEF is doing to help women in developing countries is more important than Go charities, but I guess their choices should satisfy everyone.
I wouldn't worry about a quick shift like that. You can look to the Chess world, there are still plenty of masters and grandmasters earning their bread. There's still lots of interest in the human vs. human aspect of the game. In lectures, some GMs make good use of those widely available AIs for analysis, too.
I follow neither Go nor martial arts, but there seem to be some interesting parallels here with some emotional reactions to what appears to be the relative weakness of karate or kung fu versus grappling in UFC. The mystical aura of these martial arts as traditionally practised for hundreds of years suddenly falls away in the face of what often seems like brute force.
In fairness, UFC fairly severely limits what can be done. To begin with, they use gloves, and things like eye gouges and finger locks are illegal.
Because strategies like "ripping out someone's intestines" are illegal, boxing and wrestling have an unfair advantage because they don't have to worry about that stuff to begin with. For more realistic fighting situations, see: https://en.wikipedia.org/wiki/Lei_tai
The eyes are small targets. The only time you can reliably gouge them is when you're already winning a grapple, at which point it's superfluous. Finger locks are similar, with the additional disadvantage of being less likely to end the fight even if successfully applied. Admittedly, UFC rules now also take entertainment into consideration, but early UFC rules mostly just banned things that risked permanent injury for very little tactical benefit.
it was high time that these ancient martial arts were shaken from their comfortable reverie. what were once based on actual fights centuries ago had devolved into pedantic adherence to ritual and form. it actually started way before the UFC: bruce lee shook the kung fu establishment with jeet kune do in the 1960s. helio and carlos gracie shook up jiujitsu by incorporating real world situations. you are actually now seeing wing chun make a comeback. look at conor mcgregor's unorthodox style and you will see phenomenal angles that come from kung fu and tae kwon do (karate), spinning leg kicks that are actually landing, switch stances, etc.
i believe all these ancient arts are making a big comeback. the roots are still there, they just got concealed over the centures.
To me the book was always about Go taken to the extreme. The Game is the Empire and visa-versa. Imagine if Japan had been more dominate in previous war, then conquered space - then you had the Player of Games.
Art is, mostly, domain of knowledge not yet claimed by science. When a technique turns art into science, it means that humans are ready to tackle even harder problems. Think about medicine before and after microscopes. Holy art turned into boring science: huge win for human race.
If an AI won the next Hugo award, I would be rejoiced. It wouldn't mean the end of literature at all; it would mean that humans are ready to produce an even higher form of literature.
We have all kinds of visual art made by computers and AI - prom painting from photos to abstract art to 3D renders.
We have computers writing poems and haiku.
The only thing that's missing is the conceptual creation, which, let's be honest, most human artists struggle at as well. So writing and interesting story is not yet in the AI's domain.
Art is more personal though. There is no single path to "winning" in art, and "good art" tends to mean different things to different people.
I'm sure soon (if not now) AI can easily create art that regurgitates popular trends in the past, and perhaps some artists may find a way to use AI / other algorithmic techniques in a way that complements their personal vision. But AI is a long way off from replicating the quirks of human nature, the unique personalities and personal visions of humans. Until that happens, I can't see terribly interesting art emerging from AI alone.
None of this stuff is that good yet. I see no fundamental limit, but let's not pretend that machine-generated music or poetry is as good as the best human stuff, yet.
I disagree. Art is mostly not a "problem" to be "solved" by science. Art is not graded in a scale of difficulty, from "easy" to "harder" art that mankind has to gradually reach.
Literature is not a lower form of art that we must strive to automate so that we can dedicate ourselves to more "complex" forms.
Maybe. That's definitely not how I read it. Example:
> If an AI won the next Hugo award, I would be rejoiced. It wouldn't mean the end of literature at all; it would mean that humans are ready to produce an even higher form of literature.
To me this seems to be claiming that what we have now is a form of "lower" literature, to be tackled by AI so that humans can produce "an even higher form of literature". But, of course, literature isn't graded in a scale of "low" to "high". (Well, there is lowbrow and highbrow, but that's something else).
The mention of medicine as "holy art turned into boring science" (already somewhat dubious) also seems to point to the idea that it is art that's being "solved". But I admit I might have misread it.
By the way, I don't rule out that art can be produced by an AI (whatever that means). I subscribe to the notion that art is in the eye of the beholder, so if humans can find meaning in something produced by a non-human, that's probably valid art!
> To me this seems to be claiming that what we have now is a form of "lower" literature, to be tackled by AI so that humans can produce "an even higher form of literature". But, of course, literature isn't graded in a scale of "low" to "high". (Well, there is lowbrow and highbrow, but that's something else).
Being "low" or "high" is all dynamic. We already have a good example: the advertisement industry. When a way of advertising your product first came out, it is fresh and captures people eyes. As more and more advertisers follow suit, it became bad ad, and advertisers are forced to find new ways to attract people. Basically the criteria for good ads changes all the time, but that doesn't kill the ads industry.
Now imagine if AIs can write sci-fis that are "good" according to today's criteria. That would mean there will be loads of "good" sci-fis in the market, and people soon get tired of it. Now sci-fi authors have to come up with more creative ways of writing good sci-fis.
So AIs being able to produce literature means more variations and faster iteration in literature style, much like the ads industry today. I don't know whether this is a good or bad thing, but it is certainly far away from the death of literature.
In general, I don't have a problem with your opinion for all human endeavors. I readily accept that many of them can be optimized and automated, indeed freeing humankind to pursue worthier goals.
I'm specifically objecting to your notion of art.
The advertisement industry is not a good analogy. It can indeed be improved, possibly by automated means. In contrast, the progression from "good" to "better" art doesn't work like that -- if it even exists at all! What is your measure of quality, anyway? Complexity? But sometimes minimalism is preferred in art. Maybe how many people like it? It doesn't work either; a lot of people like stuff that is not enjoyed by the majority.
When is art "better"? How can it be "improved"?
PS: the Sci-Fi market is already flooded by below-average human writers, so we don't need an AI to picture this nightmare scenario of good SF writers struggling to sell their books :P
I could see an AI with huge access to data and a database of existing publications finding potential correlations in disparate datasets, generating hypothesis and the experiments required to test them, and forming conclusions based on the results of those experiments. At least Go/Chess/Sports/Art have value in both the action and observation of the act, but when human-driven research can't keep up with AI-driven research, scientist might need to rethink how to explore new forms of science.
> Now imagine the winning author of the next Hugo Award turns out to be an AI, how unsettling would that be.
I've been thinking precisely about that. I think a book written by a machine will make the NYT bestseller list within our lifetimes (I would give it a 75% chance within 10 years, but that's just a gut feeling).
But aren't the winners of literary awards sort of subjective? I mean, chess and Go are based on beating a direct opponent based on a set of known rules. But getting an award for the best book that is picked by a judge means that you have to hit all the right notes in all the right places for that specific judge or panel. I'm not saying it won't happen, just that it's more subjective
Literary awards are definitely subjective, but I don't think that matters. I'm just saying I think an AI will write a book that a significant portion of the population thinks is very good. The first "good" AI book will likely be from some formulaic genre (think 50 Shades of Gray) that most people think is trash, but enough people will like it to legitimately propel it to the top of the sales charts.
I also think my kids will live long enough to see an animated movie that is conceived of, written, scored, and animated by an AI.
Music is a much easier problem (tight structure, lots of existing data to quickly analyze), and the animation bit is already being pretty thoroughly explored by procedural generation in games. You'd still need a "director" to pick shots, but most of the other pieces are nearly in place. We have algorithms that can create new environments, and design and animate new characters.
But generating a coherent narrative, and good writing to "implement" that narrative. These are huge problems which - as far as I'm aware - would require major breakthroughs to achieve. Machine translation is still utter garbage, and that's fairly straightforward work. We're nowhere near an AI which actually understands language.
the first AI to replicate the success of: an oscar-winning movie; a pulitzer prize novel; a tony award winning musical-- will all at once devalue the IP value of all creative content worldwide by a measurable degree. it will be like watching a global stock market crash in terms of valuation. but, as in chess, and as in go, humans will probably learn from this AI and emulate it as well.
I'm not sure what will happen. Creative works aren't generally fungible. If I order tickets to go see Jack White play, I'm not going to be tempted by Nickelback tickets that are half the price.
It's the same fungibility argument though. The market for sheet music for Willie Nelson's catalog isn't affected by the availability of other music, is it? If I want to play On the Road Again, there's only one IP owner for that.
Depends if you want _that_ music that you remember, or are looking for _some_ music that you'll like though. Once you're going into it with no prior values, price popularity and expectation of interest will dominate your choices, at which point computer generated options will be totally viable.
Imagine the best book you've ever read. Entrancing, enlightening, cathartic. You reach the end, and it's ... perfect. Oh hey, a sequel. Wow, the sequel is just as good as the first book. It expands upon it without diminishing the original -- you feel better, more complete for having read it. Wait, is that a third book in the series? Wow, it's even better than the first two! A fourth -- well, maybe you should go to work now, it's Monday, but the book is so good. Calling in sick once won't hurt anything.
Imagine a perfect series of books, published without end, each better than the last, a new one coming out weekly ... daily ... hourly ...
What you describe is a push situation: the books come out and you have to try to keep up to speed with their release. It could however, also be a pull: whenever you feel like reading an amazing book, you just ask the AI to generate one for you, optionally continuing the last story you read.
Not one, but the best book that you need to read in this particular moment. Full with all the advice that you were seeking, with the right amount of new things that you learn and familiar knowledge that you reinforce. The protagonist casually comments things very related to the open issues in your work, and helps you see the particular issue you're having with your boss from another perspective. With just the right amount of common content so you can comment with your peers at work (perhaps your office pal is reading the story of a side-character in your book - the watercooler conversation is great, he gives you new insights for the reading of this evening - and now you both agree on the discussion thread of last week).
Hey, what's that? It seems that the new upgrade is now able to create scenes in Unity with the scenarios that are covered by your next novel. Great! Also there's this interactive package where your work items can be not only an input but also an output and turns your work into a game. By the way your girlfriend has entered your book, let's switch to some of the shared scenes... let's put on our VR glasses... good. Now I only need someone to feed and clean me.
You could also potentially specify constraints on a book and have it generated for you, e.g. generate me a book about a gay dutch vampire in the 1800s. Could open up a whole new concept of hyper speciaised books tailored for individuals particular desires and preferences.
The practical problem with this is that, as I understand it, the deep learning system needs a pretty large data set to work with to infer rules from. You can do this with go because there is a constraint on legal moves and a deterministic win condition, but given how vast the number of potential novels is (If we count the space of all ten thousand word collections of grammatically acceptable sentences) the existing number of novels may no be enough to infer a pattern. (Though possibly you could split the problem up by separately doing the natural language processing and abstractin out the plot)
Ignoring the training problem, apply it to movies: I'd like to see this movie, but starring these actors, directed by this director, with a soundtrack by this composer/band.
This is only a bad thing if you feel compelled or addicted to reading these books. There is nothing wrong with always having a better book ready until you obsess about it.
> And now an AI without emotion, philosophy or personality just comes in and brushes all of that aside and turns Go into a simple game of mathematics
Well, I am of the opinion that mathematics is the language that subsumes all other kinds of languages and line of thoughts. In the end we shall be able to describe every idea or thought in purely mathematical form.
I believe that mathematics is just a tall model of implications built from atoms and relations. To say that math underlies everything is to say that we can model things. To say that math is not in something is to say that no set of atoms and relations can account for that something's behavior at a rate better than chance.
Even though we do not have a complete understanding of exactly how the networks work, what is the function that they are minimizing but what we do know is that it is a mathematical function i.e. it has been mathematically modeled. So, I think it is safe to say that it does turn it into a game of mathematics.
Yes the function themselves are pure mathematical functions, but the way to derive them relies on human intuition (not mathematical formula). It's like saying: I know how to do a 1 times 3, I just do 1 + 1 + 1, but I can't tell you why, then I can't say that 1 times 3 is just mathematics.
"As a casual player of Go myself, some of the moves that AlphaGo made were crazy. Even one of the 9th Dan analysts said something along the lines of 'That move has never been made in the history of Go, and its brillant.' and 'Professional Go players will be learning and copying that move as a part of the Go canon now'."
Yeah, I've been noticing the pronouns thing. In chess challenges I always got the impression that the AI's play style was like a chain chomp. Limited, but ruthless within its limits, and definitely 'mechanical'. In these games the commentators are treating AlphaGo like a person.
If people didn't know AlphaGo was a machine, and simply played anonymously online against masters, I wonder how they would interpret AlphaGo's personality?
"Go, unlike Chess, has deep mytho attached to it."
While I'm not able to comment on the length/depth of history of chess vs. go, the above statement seems foolish. Chess also has a lot of mythology and mystique attached to it. Champion chess players (perhaps more so a decade or two or three ago) are also treated with a respect that is not casual.
I liken Go to a real-time strategy game. Essentially a game of StarCraft 2 can have an infinite set of 'moves'. When an player wins at StarCraft 2, you can argue that he is wise too.
Yuioup said that the winning player can be called wise, not the game it self. I doubt there are any three millenia old Go players around, but I can be wrong.
> Now imagine the winning author of the next Hugo Award turns out to be an AI, how unsettling would that be.
Considering some of the recent Hugo winners, it turns out the story doesn't actually have to be good, so yeah, a computer-written story winning is probably closer than we think.
Let's compare Go and Chess. We all know that Go is more complex that Chess, but how much more?
There's 10^50 atoms in the planet Earth. That's a lot.
Let's put a chess board in each of them. We'll count each possible permutation of each of the chess boards as a separate position. That's a lot, right? There's 10^50 atoms, and 10^40 positions in each chess board so that gives us 10^90 total positions.
That's a lot of positions, but we're not quite there yet.
What we do now is we shrink this planet Earth full of chess board atoms down to the size of an atom itself, and make a whole universe out of these atoms.
So each atom in the universe is a planet Earth, and each atom in this planet Earth is a separate chess board. There's 10^80 atoms in the universe, and 10^90 positions in each of these atoms.
That makes 10^170 positions in total, which is the same as a single Go board.
I am not sure that calculating the raw number of positions is a good indication of complexity at a given point. What if most positions are obviously junk in go while they are more difficult to assess in chess? Not saying this is the case in this particular example but thats a possibility in theory.
To illustrate your point: you can just add rows to a game of Nim (https://en.wikipedia.org/wiki/Nim) to get a truly enormous state space, without changing the simple winning strategy.
> What if most positions are obviously junk in go while they are more difficult to assess in chess?
I wouldn't go with most (because I don't know about that), but many of these boards would also be either impossible to achieve (in a normal game) or illegal.
This doesn't seem to be the main reason why Go is harder than chess for computers. It was noted that even in 9x9 Go, with a comparable branching factor to Chess, traditional Go programs are still no stronger than on big boards. The main difficulty for Go is that it is much harder to evaluate board positions. So in Chess the depth of the search can be significantly reduced by using a reasonable evaluation function, whereas in Go no such function seems to exist.
Well, I guess it was more true before the advent of Monte Carlo Tree Search. Even so, note that even in the case of MoGoTW in 2011, it played blind Go (this helps the computer), and out of 4 games, won two games against a 9p player, and lost 1 game to a 5p player. Though it is perhaps better than MoGo's performance on 19x19, it still isn't very good, doesn't seem much better than MoGo on 13x13, and performs much worse than computer Chess, despite a similar branching factor.
The branching factor of 9x9 Go isn't 75. 75 could be the factor in early game, but the average factor is somewhere between 40 and 50, versus 35 in chess. State-space complexity is also considerably higher in Chess than in 9x9 Go.
Not sure what you meant regarding MCTS, I never said anything about MCTS not being able to beat pros.
This evaluation function does exist, and it's better than the super-simple chess evaluation function.
See, a chess program needs to find a lot of valid moves (see Deep Blue which won because it had stupid but extremely fast HW move generators), evaluate the moves and do a very deep search, up to 14, out of the very few alternatives. Russian chess programmers were better those times. They came up with AVL trees e.g. But hardware won.
In Go it's completely different. A move generator makes no sense at all, and a depth search of 14 neither. There are not a few alternatives, there are too many.
What you need is a good overall pattern matching of areas of interest and an evaluation of those areas. And we saw that this feature outplayed Lee Sedol. Sedol couldn't quite follow in the recalculation of the areas.
Same as in chess AlphaGo learned the easy thing, that the center is more important than the corners, something Lee forgot during the game.
But it's not a deep search, it's a very broad search, and very complicated evaluation function. A neural net is perfect for this function.
> whereas in Go no such function seems to exist.
It does exist. It's the neural net. It's a simple pattern recognizer, which learns over time more and more.
On the other hand, 2^565 is already slightly larger than 10^170. In other words, a couple of hydrogen atoms as quantum bits can perfectly well encode every possible position.
This game was largely played extremely well by both sides. There were a a few peculiar-seeming moves made by AlphaGo that the commentator found very atypical. These moves ended up playing a very important role in the end game.
I should also say that it's somewhat clear that Sedol made one suboptimal move, and AlphaGo capitalized on it. Interestingly, the English commentator made the same mistake as he was predicting lines of play. This involved play in the center of the board, in a very complicated position. Prior to this set of moves, the game was almost a tie. Afterwards, it was very heavily in AlphaGo's favor.
Seeing the way chess computers have evolved, this won't be far into the future. There are several tournaments just for chess computers, and it is very hard even for grandmasters to follow every more. Somehow, after a 20-move sequence that appears to accomplish nothing, one side is slightly up, and the rest of the game is decided.
There's been a big annual Go computer cup for a while now, it's been fun to read about: http://jsb.cs.uec.ac.jp/~igo/eng/index.html Of interest this year is Facebook's new bot is competing, which employs the same strategy as AlphaGo (Deep nets with MCTS), but from what I remember they've said (they've been a lot more open about its performance, you could even play a version on KGS, though I did really like AlphaGo's paper) it's quite a bit weaker. However I suspect it's also running on a lot less computer power, and that it will still be a while before you can get AlphaGo or better performance from a mere modest gaming rig let alone a standard laptop.
One benefit of programs like AlphaGo eventually becoming available to individuals is that we'll see a rise in some incredibly strong young players that leave their ancestors in the dust, as I think has happened with Chess via Magnus. For the amount of training that can only be done by playing more and more games, being able to do that against a computer will be a lot more efficient.
> Seeing the way chess computers have evolved, this won't be far into the future.
With chess, there were two breakthroughs. First, there was Deep Blue, which threw massive hardware resources at the problem and achieved world champion level play.
That was interesting, of course, but didn't really do anything for human chess, because most humans did not have access to the necessary hardware.
The second breakthrough was when the developers of chess programs that ran on commodity desktop computers improved their algorithms to the point that they could play at (and far beyond) Deep Blue's level even though they were only able to search about 1/100th as many positions per second.
That was when humans started being able to really use computers to help the humans understand chess.
The breakthroughs in chess algorithms on commodity computers had little, if anything, to do with the Deep Blue breakthrough. The two are just too different.
Can AlphaGo be made available on hardware that top human Go players have access to, or is AlphaGo to Go as Deep Blue was to chess?
Hey, that wetware is ten time as powerful as what AlphaGo has to work with. Give or take a few orders of magnitude. And given that Lee's brain only uses a portion of that on Go.
> Hey, that wetware is ten time as powerful as what AlphaGo has to work with.
I don't think that's actually true. The hardware that AlphaGo is on is probably a lot more powerful than the one that is available in a single human brain, the big difference is in the software.
See the difference between the very best chess programs of a decade ago versus the ones now.
Seems like somewhere in-between. They created a novel approach that is scalable and improves as you throw more hardware at it. And Google is throwing a lot of hardware at it based on their past matches with hundreds of CPUs and GPUs. I think the fact that they have been so mum about what hardware they're using suggests it's quite extreme, but hopefully they release more details soon.
Its somewhat interesting to think about the differences in marketing between IBM and Google; IBM was marketing hardware and HPC with deep blue, but Google is marketing AI when so much of their advances in AlphaGo are enabled by distributed systems and HPC running billions of games training deep neural networks. It feels a little smoke and mirrors which is probably why they won't release much until after they get enough marketing value from this tournament :)
Fan Hui has already learned from AlphaGo. He's been playing matches against her regularly, and (perhaps as a result of that) won _all_ his games in the last European championship.
Myungwan Kim said he feels like it plays like a she. Personally I think its informal gender was determined by the nigiri of the first match, as it's common to refer to black as he and white as she absent of player names. And I'm expecting to see at least one really cute AlphaGo-tan drawing any day now.
Just that in that time AlphaGo will have played millions of more matches against itself, learning at a much faster rate than him. Not sure, he might still beat the machine though. That needs to be seen.
Impossible for anyone of us to know. The Google people might know, but it's in there interest to not present AlphaGo as clearly superior because an even game attracts more views.
I watched the live stream commentated by Myungwan Kim so I know what the parent is referring to (later in the game, Myungwan Kim referred back to that possibly being a mistake), although not being a Go player myself most of it was over my head.
It was fairly early in the game and about 4-5 lines down from the top and towards the center, center-left. Apparently, Lee Sedol played a little conservatively and "didn't take a ko" (?) when he could have.
Hope that helps you figure out what they're referring to, or maybe someone else can chime in, but that's what I remember.
I find it very interesting that to a layperson, the idea of a computer being able to beat a human at a logic game is pretty much expected and uninteresting.
You try and share this story with a non-technical person and they will likely say "Well, duh..it's a computer".
I'd say if anything the average person's perception of what AI can do in the opposite direction. In the mid twentieth century the idea of a near future involving computers/robots that thought and interacted much like humans but never made mistakes was pretty mainstream. People have rather dialled back their expectations of feasible computers since then, to the point the average layman thinks a hardcoded easter egg humorous response in Siri is impressive because although talking to Siri is just like a more error-prone alternative to using the keypad, the response sort of seems like how a human would handle the question.
The average person isn't impressed with computers winning at Go because they vastly underestimate the complexity and open-endedness of Go and wonder why it's really all that much more complex than chasing Pac Man through a maze like their computers were doing quite happily, and even with apparent personality, in the 1980s.
That's why task based notions for general intelligence are rubbish.
Here is what humans can do: When presented with pretty much any task, specified however poorly, they can design hardware and algorithms that can beat themselve at that task.
That's a decent measure of intelligence. A decent measure of creativity is coming up with tasks that make the intelligence part as interesting as possible.
Optimizing compilers do this already and they aren't 'intelligent'. Iirc alpha go is based on the same algo that could beat all atari games sight unseen.. satisfying the 'novel situation' requirement you set forth..
Optimizing compilers do this already and they aren't 'intelligent'. Iirc alpha go is an extension of the same algo that could beat all atari games sight unseen..
That's a point of concern for using the Turing Test to assess the intelligence of a computer, but that's only because the computer has to also be trained to expose human-like features that are considered to be unintelligent.
But on one hand, in order to have a shot at passing the Turing Test, a computer has to first be able to understand and speak human language. And that's pretty damn hard actually.
Yes we can deploy statistical methods, deep learning or what have you for classification, feature extraction and so on, but natural language is context rich and ambiguous. NLP isn't a conscious process for us either, but our brains are capable to disambiguate effortlessly. And a big part of that is also the rich (human) experience we gain while growing up. Being able to speak natural language is the one big trait that separates us from animals. It's why we can cheat death, because our children can learn from the acquired knowledge of all of their ancestors.
This is different from playing Chess at least. With chess you can deploy smart algorithms, but in the end it's still a raw search for good moves in the space of all possible moves, while giving up on branches with a bad score. Raw search is what computers have always been good at. That's not really possible with NLP, because at least with chess your vocabulary is very limited and for calculating good moves you don't need extensive knowledge about the world we live in.
Computers will surely be able to do that in the future and that will be a major milestone towards "true AI", but for now computers cannot do it.
On the Turing Test, some features considered as being unintelligent, like the tendency for lying, or being sensitive to insults are actually evolutionary traits, that have arguably helped humans to survive. So while emulating some human traits will be counter productive, a "true AI" will be concerned with survival and as a consequence will end up doing whatever it takes.
So while passing the Turing Test may not be enough, not passing the Turing Test is a sign that the computer is unintelligent.
I know this is a semi-joke, but it's worth discussing. While such people are less smart than average or have a deficiency in their education or both, unless you're talking about mentally handicapped people with a medical diagnosis, all normal humans have a conscience, are capable of reasoning, can understand complex symbols, can entertain complex thought, are very good pattern matching machines, can speak natural language and can learn and acquire new skills, which makes them intelligent.
I know the kind of people you're talking about. I have a family member like that. She's not that smart, she failed her baccalaureate, she's semi-illiterate and she probably suffers from ADHD, though a lot has to do with her upbringing. But getting answers like that from such a person means that you're not asking the right questions or the incentive to answer is not there. Give a cash reward to a cash-strapped person and you will never get an answer like "im so bored this test sux".
Yes, but you also have a database of answers from people, so even if you ask things like "what color is the sky" you're still getting the right answer.
it jokes with you, it gives vulgar answers sometimes, etc.
it's NOT good enough to fool someone, but what if you just made it sound like it's a person with a disability? Is that still beating the Turing test? Is talking to a "five year old" still beating the Turing test?
Or does it have to be a 100 IQ adult person fluent in English? In which case that's just improving the bot a bit. To make it a funny and charming bot it would require even more effort. But it's just slowly improving the state of the art with some kind of techniques like reinforcement learning or neural networks or whatnot.
When a bot actually beats the Turing test it wouldn't be big news because it would just be a slightly better jabberwacky.
> So while passing the Turing Test may not be enough, not passing
> the Turing Test is a sign that the computer is unintelligent.
You could have an intelligence that's just not smart in the human
sense. Consider running into an alien intelligence evolved from our
equivalent of octopuses, you ask it a questions but it only
communicates via color changes on its body.
Similarly you can conceive of an AI that's smart, self-aware and
intelligent just hasn't been developed to talk to humans.
The Turing test is a fine test to figure out if your AI is
conversational with humans, but the OP I was replying to was
suggesting it as a general AI intelligence test, it's not meant for
that, and will give you both false positives & negatives.
I encountered this many times over the past couple of days haha. I then have to explain that previous AIs, like those in chess largely used brute force computation to simulate each move, while these AIs "actually learn, similar to our brain". Probably not the most scientific explanation, but I feel your sentiment.
They are using Monte Carlo Methods for looking around, more like a tiny sampling, it's an incomprehensibly large space. They have both value and policy (deep) neural nets. Train one to get the sense of good/bad individual board states (value) and one to get the sense of good/bad trajectories (policy, evaluates chains of moves).
Chess has a fairly straight forward ranking of pieces, fairly well established ranking of piece power. A knight is more or less always a knight.
Go, you kinda make up your "pieces" from scratch and there isn't any single common ranking. This makes the problem of even evaluating the board difficult (value), which is a sub step of making your higher plans (policy).
The output of the two dNN help cut off, or prevent the sampling of the bad game states and encourage looking into the fruitful areas.
Well, it's a bit more complicated than that. Neither chess engines nor go engines use brute-force (that is, exhaustive search). Although go does have a much higher branching factor and that does affect the search algorithm used, the biggest challenge is being able to write good evaluation and move-guessing heuristics.
Go is a lot more psychological and emotional than playing tic-tac-toe. Its a strategy game, which undecided outcomes, where ones style and vision changes how you play and see the game.
The computer was able to overcome the computational difficulty humans compensated with abstractions and strategic concepts. We still use those to understand whats going on in the game, but the computer is oblivious to them and only uses a tactical view of it.
To me, "creative" is not the right word. Maybe try "intuitive".
As Radim says, of course, the intuition is only relevant when logic fails. However, no computer, including AlphaGo, has sufficient processing power to take the logic-only approach to the, uh, logical extreme (more board states than atoms in the universe, etc etc).
So both humans and computers play by a combination of logic and intuition. Surely Lee Sedol must be incredibly good at both. Perhaps AlphaGo is better than Lee Sedol at the logic part (or perhaps not, but just suppose for the sake of argument).
In this light, what is interesting is that AlphaGo is sufficiently good at intuition, a domain we might have considered uniquely human[0], to complement its ferocious logical power.
[0] I find it foolish to consider anything uniquely human, but a humanist essentialist might make such a claim.
The other interesting bit is the combination, the intuition guiding (and evaluating the results of) the logical searching.
I'm still holding out for the discovery of a proof of winning strategy like other combinatorial games. That to me is the logical extreme, not magic evaluation of all states. There are two interesting books on analyzing Go with combinatorial game theory, they found a neat system of scoring in the endgame and can create example boards with one best move that can be found by mechanically applying their system but stumps 9p players (who use whatever systems).
It's a creative game only to the degree your logic (reading) is not strong enough :)
I mean, deep down, Go is a perfect-information-zero-sum-discrete-2-player game.
I think a more successful pitch is showing that the strategies and skills coming from Go (human or robot) can be useful outside the game, as I argue here: http://rare-technologies.com/go_games_life/
What I really liked about those games so far, and Michael Redmond commentary, is that AlphaGo not only beat Lee Sedol twice, but also Redmond. He is playing the same style as Sedol, he constantly predicts Sedol's moves and he is as surprised and does the same miscalculations as Sedol. He really needs some time to find out when he made a mistake, the same mistake Sedol was eventually doing. This is high class commentary. Even if they have no Computer screen to clear the screen after some variations. He remembers all the stones and immediately clears his own moves, amazing. I'm not sure if a better device would actually help.
> He remembers all the stones and immediately clears his own moves, amazing.
Chess grandmasters can often play simultaneously and blindfolded against 20 different opponents (and win all but one or two games). This is nowhere near the extent of Redmond's memory strength. :)
I sense a change in the announcer's attitude towards AlphaGo. Yesterday there were a few strange moves from AlphaGo that were called mistakes; today, similar moves were called "interesting".
fun fact: when TD-Gammon hit the backgammon scene in the 90's, it didn't just defeat the top level human pros, it shattered the whole metagame and changed how humans play the game. It could be that human vs human go will look very different in the future due to what AlphaGo has learned and can teach us.
AlphaGo maximises the probability of winning, and not the margin by which it does. So those "mistakes" yesterday turned out to be fortifying moves because AlphaGo was confident of a win. And similarly today the weird moves were interesting because they perhaps indicated that AlphaGo thought it was ahead.
Yeah, that explanation from the DeepMind team member today put a whole new spin on some of the 'odd' late game moves. It doesn't 'care' about about margins so it will shore up its odds of a win in preference to increasing the margin if it wins.
I had a feeling after yesterday's win that people might be tempted to call it a close win by AlphaGo and predict that Lee could overtake it with a bit better play, but that we'd find no matter how good Lee played, AlphaGo would adjust and continue to come out just a bit more on top. It's actually probably really hard to tell how much better AlphaGo is because it probably plays quite conservatively overall and hides a lot of potential strength.
To address this, has anyone considered pitting AlphaGo against a team of 9p players consulting with each other, and perhaps taking charge of different conflicts on the board?
I can't find the source now, but I remember reading a while ago that a team of go players is, perhaps contrary to intuition, not significantly better than their strongest individual alone.
There's just no useful way to "pool" human thinking power and redistribute it to where it's needed the most – all the players will simultaneously consider the same branches. At best you reduce the risk of making a silly mistake, but sitting pros are already pretty good at that.
> all the players will simultaneously consider the same branches
A computer-assisted pool of humans might work. Feed each human a board state advanced by 1/2/3/N moves down the decision tree in some direction, and have them evaluate that particular sub-tree. It's a map-reduce problem!
For some reason this made me think of the Focused in A Deepness in the Sky - where real general AI isn't possible [at least where we are] so human minds are harnessed to solve problems in a deeply unpleasant way.
Is there a non-dystopian way we can do this? The immediate problems seem to be bandwidth of communication and ability to quickly generate shared culture and jargon.
Are Bridgewater Capital's employees Ray Dalio's focused?
AFAIK Mechanical Turk is just getting random people to do stuff for you - the concept of Focus is something else entirely and genuinely quite terrifying.
If you haven't read A Fire Upon the Deep I'd recommend reading that first. It's not vital but there are some subtle links between the two that are rather cool (not the shared character, more the shared "physics" and tech).
That's actually a really interesting idea. At this point you're basically replacing the "policy" neural network in AlphaGo with biological human neural networks.
Is there a meaningful response that each one could give that would be optimal? I figure this would only stand a chance of working if the pool of humans were cloned from a 9 dan :-)
Each one needn't give a 9 dan response. Each person gives one response to the best of their ability, and scores the situation on a scale of like 1–100. Collect all the responses and distribute the new board situations over the same – or a different – group of people. The score is mostly used to value the quality of previous moves. If a branch leads from a situation most people rated as 60 to a situation where most people rate 10 it's not a good branch.
Of course, Go has way too many eligible possible moves for random people to vote on, but a large enough group of top pros might be able to do well just by voting.
Four to five expert chess players suggested moves for the world team. I feel that for any reasonable non-expert it comes down to "choose between these suggested moves" rather than "pick a move". It's also not really random human beings, since the selection of participants is self-selected and therefore much more likely to contain very good chess players.
This is mostly semantics, but anyone who doesn't read your link might get the wrong idea, so I felt like clarifying it a bit.
Fair points. Thanks for pointing this out. I used "random" to really mean "not-top-pros." Moves suggested by experts being voted on by a large sum of people are kind of similar to how democracy and capitalism work.
Part of the reason go is a difficult computational problem is that local conflicts can have global relevance. If it was the kind of thing that could be decomposed easily to different humans then it would be an easier game for a computer to crack.
Yes, particularly because of the sente/gote dynamic. However, it takes time to read individual situations through, and this might be chunked and distributed. The problem, as usual, is efficiently communicating analysis in a useful way. There would probably need to be a hierarchy where "specialists" report to the captain about the battle, and then the captain has to prioritize and decide on the order of action for each region. But the minute detailed analysis for each region would be entirely delegated. Note that leiutenents would be responsible for understanding their region's relationship to the whole, but only the captain is truly responsible for understanding it all.
At the same time, only now are people seeing how its hard for people to play Go against computers. Until AlphaGo I didn't even think computers were close.
Actually computers were already in the top 10 percentile or so. I mean top go playing bots ranked something like 5dan amateur which is very hard to achieve, so it could already defeat most human players.
Go is much harder than chess. The tree of available moves is massive and difficult to prune, as it's hard to get a reasonable heuristic about the "strength" of a given Go position.
I wonder how this will affect future human play. About 30 years ago my brother and I started playing a simple African stone game Kala. We each won about half the games until I coded up a brute force search to play against. Given a game tree to the end, the program made the weirdest looking opening move, when playing first. I started making that move and forever after won.
The situation with Go is different. (I wrote the Go program Honninbo Warrior in the 1970s, so I am a Go player and used to be a Go programmer.) Still, I bet the AlphaGo, and future versions, will strongly impact human play.
Maybe it was my imagination, but it sometimes looked like Lee Sedol looked happy + interested even late in the two games when he knew he was losing.
I'm totally uninformed about Go, but by now it seems that unless you're clearly in the lead by the end of the midgame, AlphaGo is going to win, simply because at that point its Monte Carlo Tree Search is going to our-compute you in examining all the tactical variations in the endgame. Lee Sedol really seemed to be under a lot of time pressure at the end.
EDIT: clarified to what I originally meant: "end of midgame"
It really depends on how good AlphaGo is at avoiding mistakes. If you get critical points on the board, it's possible to create a resilient structure that is defensible against comebacks. But when I play, there are times when I can make an overwhelming comeback because my opponent missed one critical move. I've likewise had it done to me. This is of course probably less common at higher and higher levels of play.
I'm reminded of the time when Deep Blue made a buggy move that totally threw Kasparov off because it was so counterintuitive. But it turned out to be a real mistake. If Kasparov had kept his composure, who knows what could have happened?
Of course, as been stated here, AlphaGo doesn't have the issue of having to keep composure, while Lee does have that issue and pressure. But who's to say that AlphaGo won't make a similar significant mistake, and that Lee won't be able to capitalize on it to make a comeback in that particular game?
edit: admittedly, as pshc as noted, the search space probably grows smaller as the game continues.
edit 2: suddenly, his comment disappeared for some reason?
In the mid-game – and hell, even in the early end-game – the search space is still too big to practically exhaust the useful branches with MCTS. Keep in mind "mid-game" in Go means there's still over 100 moves left in the game – well longer than a full chess game.
I'm not saying you're wrong, I'm just saying I don't agree with your motivation.
I don't think so. Pros are already pretty good at the endgame. Yeah, they make blunders now and then, so it's possible that AlphaGo would gain a few points in the endgame in some games, but not enough to overcome a significant lead and not every time.
In any case, I personally find it more interesting to see what we can learn from AlphaGo about the opening and the midgame.
My writing was sloppy, let me rephrase it: in a pro match, whatever is the estimated score when the game goes into the endgame, that will likely be the final score.
Where I wrote that a pro will blunder "now and then" I should have written "rarely" -- I don't have data to back this up but I'd guess once in every n games for some n > 30.
The next person that will beat alphaGo may not be a top go player.
In particular, I'm wondering if a computer scientist with access to the alphaGo source code and all the weights of the network could trick alphaGo in order to win games automatically (cf. the papers that show a neural net can be tricked to classify a plane as any other class).
If a human with the knowledge of the source code and the weights can do this, it is scary. Imagine a similar algorithm runs your car. An attacker that knows the source code and the weights may trick the algorithm to send your car in a wall!
You cannot do something unpredictable in a game with strict rules and borders. You had to trick this algorithm the billion times it trained for this game.
I had this conversation earlier with a friend wondering if any of Sedol's Korean pro buddies have noticed any systematic biases that could be exploited. I think it would be possible to make the neural net relatively useless by playing strange sequences that hack the weights, but you're still left with a monte-carlo-tree-search bot which alone depending on its implementation is between 2d and 6d amateur (so on the far weaker side of the professional scale). Whether you could make those strange sequences to trick the neural net while also not dying to MCTS, I'm skeptical.
I'm kind of hoping for an unconventional opening in game #3. Come on tengen or 5,5... Additionally I think that might be one way of weakening the bot in that it will find its net for suggesting candidate moves less useful and lean more on unguided MCTS, but that's just a wild guess.
Sure, almost any HnG fan will play them at least once. ;) But there's so few of them in general I would think any training data derived from them would be low-value.
If AlphaGo wins all 5 matches, what do you think DeepMind will do with it? My intuition is that they won't continue development, and instead focus on other applications.
Ke Jie believes he is slightly better than AlphaGo given the play he has seen, although he thinks AlphaGo will outpower him in a few months.
The belief that every good player that has so far played against AlphaGo would win highlights a common misunderstanding of the nature of the human mind. There seems to be in fact a very small difference between the aptitudes of the average players and those of the top players, such that it is very easy for a machine to go from average to superhuman. Getting to average is by far the harder part.
> There seems to be in fact a very small difference between the aptitudes of the average players and those of the top players
This is an interesting observation across many domains that involve "intelligence". It's actually pretty easy to see why, if you reconsider the scales for "good" vs "bad": The "bad" end of the scale is not at "the village idiot". The "bad" end of the scale is at "a rock". So, yes, once you get from "rock" to "village idiot", "superhuman" is a comparatively tiny step away.
I think people in the Go community wants to see AlphaGo play against the top ranked player. Otherwise I hope they'd at the very least release it as an AI for people to play against.
Lee is widely considered the greatest player of the last decade, winning 18 world titles, but has been surpassed in the last few years by younger players, with this unofficial ELO rating system ranking him at No.4: http://www.goratings.org/. Go is really a young man's game right now.
The probability that any human player will learn to beat state of the art chess engines is zero (at least until we have humans with biologically or electronically augmented brains). There was a very small time window in which an expert player could beat chess engines by 'adapting their style'. Do you have reason to believe Go will be any different?
They'll keep using the technology. They probably won't want to play Go again - they'll have little to gain and everything to lose - and I suspect most of the possible audience will lose interest.
Putting on a conspiracy hat for pretend, the whole thing is clearly all a big PR scheme to get Google the brand better name recognition and improve their position in Asian markets.
That's not a conspiracy theory - I think everyone agrees this is as much a publicity stunt as a true test. Unless you're suggesting they bribed Sodol to lose?
As the only 9p player in the western world, Michael Redmond is of course very impressive.
Chris Garlock, on the other hand, doesn't add that much value to the broadcast. Maybe somebody will start a "Left Commentator" meme, just like Left Shark.
As I understand it, the value network takes the place of the heuristic for scoring a given board layout, and the policy network takes the place of the heuristic for ordering moves from most to least promising.
When searching the game tree, at each ply the most promising N moves are examined (as determined by the policy network) and leaves of the game tree are scored by the value network.
I don't think that's fair at all to Chris. As someone who has never watched GO before, he kept the topic often to very beginner level questions, and often kept Michael on topic a little bit. I would say Chris added a lot of value to my experience, although this is the first time watching either of them.
I mean it's classic sports broadcasting. You have the expert and the guys who gives the expert someone to back and forth with. E-sports has been using that formula for ever.
Michael Redmond. I thought he was excellent too, and I really liked his willingness to jump into hypothetical situations and play it out. Really, really good commentator.
I was also super impressed. I know him as マイケルレドモンド (Michael Redmond in Katakana), and it was the first time I saw him speak English. He's fluent in Japanese and frequently commentates on matches in Japan. This is a video of him commentating on the Meijin title matches.
What I find fascinating - and I guess this really highlights that I have no idea whatsoever how AlphaGo works - is that at the start of game 2, AlphaGo plays P4, then Lee Sedol plays D16. To a layman, this looks like it would be a very, very common opening. Moreover, it's symmetrical - I'm not sure how that affects things, but my naive intuition is that it makes the game state less complex.
Nonetheless, AlphaGo takes a minute and a half to play its next move. Can anyone explain what on earth is going on during those 90 seconds?
My guess is it's doing the same thing it does every move: look at the state of the board and calculate the next move.
A person might memorize standard openings and play them by rote. The computer _could_ do that, but it would just be a computational short cut. In a sense, the machine is re-discovering the classic opening.
90 seconds per move was normal for AlphaGo, wasn't it? I didn't pay attention to how its time per move varies, but it's a question I would love to ask the team (although, maybe it's explained in the paper?)
You have to realise that it's not using an opening book, it's computing the best response from scratch.
The thing I find amazing about this is how soon this has happened. We all were expecting this to eventually happen but if you asked anyone who played go and was across the computer go scene when it would happen, say a year ago, they would say it was "10 years out". AlphaGo is one incredible feat of engineering.
Does anybody know how many CPUs and GPUs they're using this time? It was 1200+ and 700+ in October against Fan Hui. It would be interesting to know if AlphaGo became better only because the extra learning or also because of extra hardware. I googled for that and didn't find anything but I could have missed the right source.
What's even more exciting is that there weren't direct mistakes by Lee Sedol in this game, like there were in Game 1. So does that mean that AlphaGo is just playing on a level beyond him?
That might be the only strategy to learn from AlphaGo: If she makes an unexpected bold move, like she did in the center a couple of times, she's onto something and you certainly miscalculated that area. Don't gamble, defend that.
I've had this thought watching this play out over the past few months. You have this deeply mystical, zen-like game of ancient China which represents the philosophy of the East and it's pitted against this pure capitalist, cold and calculating (literally) machine.
You can hold out for a few thousand years, but eventually the uncontrollable and amoral technological imperative will catch on and crush you.
It's kind of poetic and sad. It feels like technology will render everything un-sacred eventually.
On the other hand one could say that it's a pretty shallow understanding of Zen, given that Buddhism assumes that everything in the mind - being a part of nature - is fully conditioned (deterministic), as opposed to Christian view of a human as a holy, God-like creature, distinctly different than the rest of nature.
Lee Sedol seemed to be doing well before he went into extra time (as far as I could follow from the commentators). How is it ensured that this is a fair game given the time constraints? I'm guessing adding more computing power to the AlphaGo program should definitely help it in this regard.
The human's strength is intuition and insight. They can look at a move and have a good understanding of strengths and weaknesses of positions by "feel" developed by long practice of the game. More time doesn't really help this much.
Another part of the game is "reading" -- playing out scenarios of response and counter-response to evaluate how strong a move is. The computer excels at this, because it can play out as many moves as its computation time allows and remember all the results with complete accuracy. Humans are slower and prone to mistakes when they do lots of reading.
So adding clock time lets the computer increase its advantage over the human in reading depth, but doesn't so much increase the human's advantage of intuition and insight.
I find it interesting that AlphaGo seems to take a long-ish time to play the move every other pro would play instantly. These pauses probably help balance any time oddities. I think Sedol managed his time well, and even AlphaGo went into byoyomi near the end. Also the fact that it's on even time is more than fair, since the standard until now has been "go on, add more computing power, take more time to play a move, you'll still lose to a pro". As for AlphaGo, I kind of remember reading that doubling the computing resources at this point gave an increase in 60 ELO points. (So if they solidly win against all the Sedol matches they may need to double once or twice or find enough software optimizations to take down Ke Jie using standard time control, but it's not out of reach..)
Well, I think the last time computers played the world chess champion (in 2006), they didn't allow the computer to think on the human time!
And you can always give the computer less time than the human, but this just shows that it's stronger than you and you need to handicap it to have a chance.
Increasing the human time is not an option, since no one wants to watch 8 hour games, and fatigue could also come into play, so the only real option for a winning chance is reducing computer time/power
Correspondence go, similarly, would see fewer errors. Holding a world championship would take quite a bit longer than in chess, though (rough guess: 300-ish half-moves per game versus 100-ish half moves, the latter, I guess, with a bit more variation). That could be problematic, as a world championship in chess already takes years (curiously, some championships finished before the one started a year earlier did)
Just as arbitrary. Unless you're a self-sustaining vegan, you're costing a lot of energy even to just get you the few thousands of kcal your metabolism consumes.
Unless you're not counting support systems, in case it becomes very complicated to calculate and decide exactly which energy expenses are for support systems and which are directly integral for function.
We have two computational substrates, human brains, and CPU/GPU clusters. Forget what it takes to support them, just consider what they consume while computing, that is, the energy consumed while they are playing the game.
Lee Sedol is vastly more efficient than the entire AlphaGo cluster. However, while AlphaGo gains a predictable amount of power as its computing power is increased, it's not clear that one could do the same with humans. Our Go players optimize individual play, not multi-brain distributed play. What would the match look like if we trained up a bunch of humans to play Go as a team, and pitted AlphaGo against a team of humans that consume the same number of joules over the course of the match as it does?
Let’s not forget that aside from being vastly more energy efficient as a Go player, Lee Sedol is additionally capable of taking on a virtually unlimited list of other, equally machine-challenging tasks – while AlphaGo can only do one thing. In fact, Lee can lift himself off the chair to a standing position, pace around the table, lift a glass to his mouth, keep it there while emptying some of it, and think about his next move – all at the same time. (And on the same energy budget.) And far beyond all that, he decides whether to do these things – or something else instead.
I admit my first thought on fairness did go in the same direction of limiting energy budgets. But after reflecting on it just long enough to realise the above, I am finding myself surprisingly uninterested. It now seems to me that nothing particularly insightful would be revealed: limiting energy budget is no less arbitrary than limiting time unless the artificial opponent is expected to be capable of a range of things comparable to that expectable of an average human. Or if expectations are much lower, the artificial opponent would need to contend with drastically tighter limits to approach “fairness” – though at this time it would be guesswork how much tighter they ought to be. Either way, it is glaringly obvious that no computer would come within miles of competing.
So ultimately the fact that Go has been “broken” (in a particular sense) at all is far more interesting to me than whether the machine is competitive with the human in any more general sense. “It’s not” as the universal answer is boring.
And to digress a bit from there: From that perspective, this was ultimately a very human achievement. It was humans who chose Go as a problem to attack and it was them who picked MCTS and deep learning as the way to go. (Uh, no pun intended.) That’s not just reassuring. It’s also a framing we should keep in mind as computers become more entangled with the physical world and more autonomous.
Irrelevant in this case, there was the latency of the operator reading the move on the screen and physically picking up a stone and placing it on the board.
I may be glad no one took my bet offer of me paying $19 if AlphaGo won 3/5 vs them paying $1 otherwise... I had a prediction at 90% confidence that nothing would show up before the end of this year that would be capable of beating the top players (though since I first heard about MCTS's success the idea of coupling it with deep learning seemed obvious, so I had an unfortunately non-recorded prediction that if a company ever bothered to devote about 8-12 months of research and manpower into combining those two algorithms with a very custom supercomputer or tons of GPUs then they would have something that could beat the best), then AlphaGo was announced. But the top pros weren't too impressed with its defeat of Fan Hui, and Ke Jie estimated something like "less than 5%" chance of it beating Sedol so I updated to 5% for this match of it winning 3/5...
Tonight's game was beautiful. Last night's was a fighting game way too high level for me to really grasp (I have no idea how to play like that, all those straight and thin groups would make me nervous). I'm expecting Sedol to win Friday since I imagine he's going to have a great study session today, but I'm no longer confident he'll win the last two.. Still rooting for him though. :) (I also want to see AlphaGo play Ke Jie (ed: sounds like from the other submission on Ke's thoughts that may happen if Sedol is soundly defeated), and for kicks play Fan Hui again and see whether it now crushes weaker pros or is strangely biased to adopt a style just slightly stronger than who it's facing.)
Re last paragraph: "just slightly stronger" is expected. AlphaGo is designed to maximise its probability of winning, not its margin of victory. You can expect solid not-very-flashy plays that definitely maintain an advantage, rather than even slightly risky plays that probably increase the advantage.
Poker remains unsolved (except for 2-player "Heads-up" limit hold'em[1]). Poker has many different variations and a player's strategy can change significantly based on many factors. The University of Alberta Computer Poker Research Group has already used counterfactual regret minimization in their Cepheus AI. I wonder if they could work with DeepMind to apply some of the techniques used in AlphaGo to poker.
As a programmer and a go player, I knew this day would come, but I'm a bit disappointed that this is how it happened, for two reasons:
1. As the game of go progresses, the number of reasonable moves decreases, so that as the game progresses, players on average play closer and closer to optimally. By the end of the game, even weak amateurs can calculate the optimal move. Logically, I would guess that stronger players are able to play optimally earlier than weak ones. Lee Sedol is known for his strong middle and endgame, often falling behind early on and making it up late in the game. He is so strong at this that he has driven an entire generation of go players to developing very strong endgame. But AlphaGo, running Monte Carlo simulations, almost certainly can brute force the game earlier than Lee Sedol can. Lee Sedol is playing AlphaGo on its own turf. A player known for their opening prowess, such as Kobayashi Koichi in his heyday, might have had an advantage that Lee Sedol doesn't. (Note: I'm not strong enough to analyze Lee Sedol or Kobayashi Koichi's play styles; I'm repeating what I've heard from professionals.)
2. I hoped that when an AI beat a pro at go, it would be with a more adaptive algorithm, one not specifically designed to play go. If my understanding of AlphaGo is correct, it's basically just Monte Carlo: the advances made were primarily in improving the scoring function to be more accurate earlier, and the tree pruning function, both of which are go-specific. It's not really a new way of thinking about go (at least, since Monte Carlo was first applied to go). It's just an old way optimized. The AI can't, for example, explain its moves, or apply what it learned from learning go to another game. It's certainly a milestone in Go AI, and I don't want to downplay what an achievement this is for the AlphaGo developers, but I also don't think this is the progress toward a more generalized AI that I hoped would be the first to beat a professional.
> I hoped that when an AI beat a pro at go, it would be with a more adaptive algorithm, one not specifically designed to play go.
The particular algorithm used by AlphaGo is of course specific to Go (the neural network inputs have a number of hand-crafted features), but the overall structure of the algorithm - MCTS, deep neural nets, reinforcement learning - is very general. So there's two ways to look at it. One is that what you wanted has actually transpired.
The other is that what you asked for is completely unreasonable. I think it highly unlikely that an algorithm not specialised to Go will ever be able to beat all specialist Go playing programs.
AlphaGo can't explain the outputs of its two NNs, but it can still explain its moves by showing which variations it thinks are likely.
> ...the overall structure of the algorithm - MCTS, deep neural nets, reinforcement learning - is very general.
It is general in the sense that humans can apply those algorithms to different problems (and have been doing so for decades). It isn't general in the sense that we can't apply AlphaGo to other problems unmodified. AlphaGo can't even play chess badly. It is not really even a step toward strong AI. (Note that "strong AI" is a term with a specific meaning. [1])
> The other is that what you asked for is completely unreasonable.
That's tantamount to saying strong AI is unreasonable.
> That's tantamount to saying strong AI is unreasonable.
No, what I said is that it's unreasonable to expect that strong AI would play better Go than whatever the contemporary state-of-the-art Go AI is. But stated like that, I'm not sure I can agree with my statement. Strong AI could design and implement its own specialised Go AI. How would you count that?!
Yeah, that's an interesting case. My initial reaction is that I'd think of it as a tool that the AI was using. If a human used such a tool I'd consider it cheating at the game. But a self-modifying strong AI could integrate the specialized go AI into itself. If that is not considered cheating, should it be considered cheating for a human player to integrate tools into their physiology? Today it's pacemakers, why not a specialized go chip with a neural interface tomorrow? And this is assuming the strong AI even has a concept of a self separate from the software it controls; that separation might not even make sense.
I think we might not be able to answer these questions until a strong AI emerges.
So given that this victory seems to be happening a decade or so before experts predicted, how likely are we to see similar acceleration in reaching other AI milestones? (Especially given that AlphaGo is using the same algorithm that won the Atari games, so it has the potential to be very general in its application)
You probably saw this, but linking anyway: https://news.ycombinator.com/item?id=10983539 The general point is that it's more evidence improvement can come in discontinuous leaps, it doesn't have to be some smooth (even if accelerating) incremental process. So timeline predictions should probably be wide, with closer-to-present lower bounds (especially when successful generalizable techniques become public). I don't think this initial view would have changed much this week unless Sedol just totally crushed the bot, perhaps suggesting there's still something more.
Personally I think the approach of combining deep learning with MCTS to beat Go was obvious to anyone sort of familiar with each thing, and with a good funded team could be done in a year or less, but a lot of 'experts' were ignorant of one or more of the areas. The uncertainty should have revolved around when some group would get around to writing the software and scaling up with powerful hardware. Implementation details, the theory work was already known.
My own lesson from this is that if all that is needed is tough engineering work (but not really new theory) that work could literally arrive in a week instead of the x+delay time it might take from scratch, because some group could already have been in the process for about x time that isn't public knowledge. AlphaGo kind of came out of nowhere; there were early signs with papers on using deep learning techniques, but I don't remember any public commitments to much. That just indicates companies are still quite capable of doing secret projects. If they had a secret new theory, too, it could be even more amazing. I know of one startup in particular that's been securely at the top of its niche because internally they have secret CS research unknown to academia.
The OpenAI initiative may be useful from the perspective that if they've shared what probably shouldn't be shared, at least we can suspect something is imminent and try to plan a last minute stand of getting it right first, vs someone like Google doing all the research in secret and then bam, unleashing UFAI.
Hugely agree with your response. Just to add slightly to that, I think the fact that Facebook had a quite similar Go AI in the works (just without self play reinforcement learning) in the works is an indicator of how clear this research direction was. The complexity of Google's solution (2 neural nets and a fast evaluation function, plus other small details?) really indicates to me a lot of manual engineering and iteration went into this. So it is not really an indicator of cool new theoretical breakthroughs, but an indication that applied engineering to make use of known techniques can achieve great things.
I predict 5-0 for DeepMind. Now, Lee has a broken self-confidence to battle (crucial for a human player), something that will not and can not trouble the DeepMind team.
Someone here said an interesting thing. Perhaps the next AI challenge would be to see whether AI running on weaker machines can beat AI of yesterday on stronger machines. And this test can be automated to find even better algorithms. Like can Rybka running on an iPhone today beat Fritz running on a distributed supercomputer? Or thinking for 2 seconds rather than 2 minutes, on the same computer?
There is something unnerving about a computer that can answer in 0.01 seconds and still have the move be better than any human would come up with in an hour. At that point a robot playing simulatenous bullet chess would wipe the floor with a row of grandmasters, beating them all without exception.
I don't get the mystery of this. This algorithm is complex. SURE! But deep learning is very fast training / repeatition of a game (or some other goal) while saving the good or bad results. Predict user moves. Find good positions/patterns. Or did i miss some here?
I think this is really fascinating but also scary. Imagine you are the best in the world in something. That is your thing and no one else can do it better than you.
Then suddenly a computer comes along and takes that title from you. But it takes it in such a way that you are never in your life able to re-take it because of how the AI works.
A game will likely just be the first field. My girlfriend is working in translation and interpretation which is another area already in the crosshair of neural networks. AIs will step by step become more efficient than people and that is terrifying.
Does anyone know what DeepMind's software stack looks like? Just based on past work of some of the people working there, I'm guessing most of the code is C++ with some Lua. Anyone know for sure?
Does anyone notice the lack of ko[1] in the games? In all 7 public games (5 with Fan and 2 with Lee) there isn't any ko. This is unusual. If we still can't see ko fights in the following 3 games...I would suspect that AlphaGo isn't able to handle ko well enough yet, and Google asked Lee and Fan to not initialize ko fights in the games.
Isn't it more likely that the opposite is true, and Ko is the kind of thing where it is possible for a computer to play flawlessly?
I am very very very far from a Go expert, but Ko did always seem to be "first person to make a mistake looses". If that's true, I'm sure Sedol deliberately stayed away from it as an area where the computer is likely to be particularly strong.
I can't believe they would be told not to, or that Lee would abide by it in this match!
Now, I'm not good enough to evaluate if this seems likely, but I could understand that AlphaGo (especially if it was winning) avoided Kos, since they are high variance situations.
What I find fascinating about this is that the system was programmed by people who were presumably not as good at Go as Lee Sedol.
So if the first comment in this thread (about how it's a completely non-human approach) is true, it's really interesting that humans can enable computers to come up with non-human ways of solving complex problems.
Seems like a big part of this story, if I'm not being completely dumb here.
These matches are not really fair: the AI team can "prepare" and examine the human's previous games, find weaknesses, aso, while the human doesn't really have anything to guide his/her preparations.
I don't think that AlphaGo was trained more on Lee Sedol's games, than on others' games. The team said that they can't find computer weaknesses until AlphaGo plays against top caliber.
Did Lee Sedol have access to a dataset of AlphaGo games in preparation for this match series? I wonder if it would help him if he could study the computers moves and strategies in other matches.
Most people other than the researchers and hackers, really did not understand what AI was capable of doing. The very idea of AI seemed too abstract to comprehend(I consider myself guilty).
But AlphaGo showed us what AI is really capable of doing in an eerie sort of way and I think interest in AI will soon become mainstream which is a good thing for the development of AI.
Now it's at least easier to comprehend the context of all those doomsday warnings about AI destroying humanity which I never took seriously.
AI enthusiast and amateur player here: Michael Redmond made a great point yesterday, if the algorithm is only interested in maximizing probability of win and ignoring margin of victory, shouldn't there be some override for weak moves played when the lead is sufficient? AlphaGo played some weak moves when it perceived it was sufficiently ahead yesterday in the end game. A truly intelligent opponent will play strong moves even when sufficiently ahead, no?
I think the idea is that AlphaGo decided that the "strong" moves, while it may have increased its lead, would have been more risky than moves that just fortified its current lead.
I think the point is the seemingly stronger move (giving a bigger margin) can have follow ups that can lead to a lower chance of winning for the AI evaluation
Is there a complete recording of the commentary? They had one for game 1. The current live stream only goes back two hours and doesn't include the beginning of the game.
I tried to watch commentary on the DeepMind channel for game #1 and found it to be very un-informative: I recommend the AGA channel (although it's been linked a few times): https://www.youtube.com/watch?v=6ZugVil2v4w, still waiting on the game 2 static video.
Last night I found that on the DeepMind channel they were explaining the game a lot for beginners. This was a bit tedious for me but then this is a historic event so there'll be noobs tuning in so I can understand that, and I appreciate that the TV people are thinking of them.
Also, the AGA channel you just seem to have a split screen showing two people with headphones on talking over Skype? (Granted, one of them is Myungwan Kim, 9p, a super likeable guy - but Michael Redmond, 9p, is also super likeable.) The DeepMind channel switches between the game board and the large commentary board, with occasional shots of Lee Sedol and occasional shots of the DeepMind terminal.
Michael Redmond was reading out lots of variations and repeatedly trying to count the board and evaluate the position. I think he was trying to be as informative as he could. Occasionally he would get lost in thought as he calculated owing to the complexity. You could call it many things but uninformative is not one of them.
> Last night I found that on the DeepMind channel they were explaining the game a lot for beginners.
Yeah, I think I caught some 15 minutes of this and stopped watching, maybe it got better later? I would have expected them to the introduction for beginners before the start of the real game, and they ended up skipping commentary on what seemed like pretty key choice in the early game to do so.
> Also, the AGA channel you just seem to have a split screen showing two people with headphones on talking over Skype? (Granted, one of them is Myungwan Kim, 9p, a super likeable guy - but Michael Redmond, 9p, is also super likeable.)
There were even some technical difficulties with the skype feed on a few occasions and I wouldn't call it great quality, but on the other hand I rather prefer the KGS virtual board - on which it's quicker to show the possible alternatives and easier to point out the key stones.
On this I do agree with you :) I found it annoying that there was not more analysis happening in the fuseki. I was like, "is it not important to be evaluating these moves?" I guess because the DeepMind crew didn't it meant that beyond pointing out that Lee Sedol tried to break out of opening patterns quickly and that the order of moves in the top-right joseki were considered sub-optimal for AlphaGo there was not much that urgently needed commenting on.
I switched to KGS at intervals to see what people were saying on the mirrored board kibitz. And the DeepMind commentary did begin to irritate at times but I feel they had to cater for noobs. Once the middle game got going it was fine though. Agreed on the KGS virtual board.
I was wondering why they didn't keep a running total of the score after each move, and show it somewhere on the screen. Isn't that pretty easy to automate?
> I was wondering why they didn't keep a running total of the score after each move, and show it somewhere on the screen. Isn't that pretty easy to automate?
Until you are close to late-game the scoring of many of the positions is quite flexible (depending on sente, and how certain local situations are resolved) so although you would may have a semi-reasonable approximations for total points, the games have been close enough that a +/- 5 point error in estimating will make it impossible to tell who is actually ahead.
That said, it would be very curious if we could get a virtual-board with AlphaGo's evaluation function used to score each position.
There's a good reason for that: there is no game on Friday. Next game is Saturday, and they may or may not broadcast that. Andrew is ready, but he needs a pro or otherwise really skilled player to provide high-level commentary. Hyungwan who did it the last two games will not be available.
They put a version of the broadcast up after it is finished, if last time is anything to go by. Took a little while because the stream is still going, with the press conference.
I always wanted to learn to play Go and one of the reasons was because it was the only game where computers hadn't defeated human - well, it is no longer the case and I kind of lost motivation to learn it.
I wonder what would be interesting games (intellectual sports) where computers have yet to defeat humans that you would probably be interested in learning?
Isn't it kind of interesting that Google is pushing the lead for these projects? It reminds me when IBM took on the gusto of developing Chess AI when they had strong technical superiority. It's almost as if Google is taking the mantle from IBM to develop these renaissance projects.
Wasn't reading the whole thread, but was it possible for Lee Sedol to play against the final AlphaGo before? Although AlphaGo seems to be a huge achievement I would find the lack of training before a bit unfair as AlphaGo was probably able to play lots of Games from Sedol before.
What happens if the Go master tries to deceive the oponent? As in purposefully play a counter-intuitive position, or even "try to lose"? Will the AI's response be confused as it is expecting rational moves from its oponent?
No. There is a scoring heuristic that AI programs use to determine if they are winning or not - ie: +2 stone ahead, -2 stones behind, etc. (Note: I don't actually know what the real heuristic is, I'm making it up). A computer will have no real reaction to a "try to lose" or counter intuitive move; it would merely recalculate the heuristic score and doing exactly what it has been doing all along - finding the next move that maximizes that score.
Well, the nice thing with go is the handicap system. I wonder how many stone handicap the human champion needs to beat alpha go, and watch that number increase over time.
I wonder if chess could use a handicap system to keep things interesting.
I would like to see another experiment where Lee is aided by a computer and plays against AlphaGo and see who wins...some believe that human intuition working with a mediocre computer is much more powerful than a supercomputer by itself.
A question: Should we take it as "a computer beating a human" or "developers beating a Go player"?
I had this discussion with my friends and we have opposite opinions.
It should be "a computer beating the human" in my opinion. It's just like parents bathing in the success of their offspring. Yes, they can be proud, but no, they didn't achieve the "win" themselves.
If the developers behind AlphaGo did the same calculations with pen and paper as they sat across from the human player, would that be them achieving the win themselves? If so, why does having an automated piece of paper change that?
"It's just like parents bathing in the success of their offspring."
There is no offspring here. There is a set of calculations, written by some people.
If a parent watches their child win a game, would that be them achieving the win themselves? If so, why does having nine months of creation time and ten years of training time change that?
There is no child here. There is a deterministic set of calculations. The developers created a deterministic set of calculations, and then executed those calculations.
AlphaGo is not a child. AlphaGo is a set of deterministic calculations, defined and created by some people. Your question about humans and their children is irrelevant. There is no child here.
In this flawed analogy, the "parents" are the developers, and the "child" is a set of calculations. If someone wrote down a set of calculations, and then executed those calculations and won, is it right to say that the person who wrote down those calculations won? I suggest that it is.
I'm pretty sure that if I wrote a child from scratch (AGCTTAACGGUAA ... etc), understood the underlying mechanisms connecting proteins to wining a baseball tournament ... I should get some credit for the win :)
There is no insight in making a child (can be done totally drunk and half passed out), although there is some in education.
What is interesting to me is that the computer makes clear mistakes when its on the lead. Since it might find the chances to win equally among different scoring results, it often picks a weaker one.
This has a powerful consequence: we have not seen AlphaGo pushed to the limit, he is lowering the distances as if it were playing a teaching game.
Lee Sedol I think came to this conclusion, and the only human strategy left is to take a lead big enough to maintain the rest of the game. And that might be the last strategy to play to show the computer is already unbeatable, because it will be pushed to its limits to win a game and it might overcome humans.
I don't see how you can term them as "clear mistakes" when the game is playing at a higher level than any of us meatbags. In this case (according to DeepMind) it's no different to a racing driver backing off the pace in the last few laps if they have a big lead - it's better to guarantee a win than to win by a large margin.
You might be confusing correctness with relevance. You can pass at the last move and lose 1 point, but if your lead is 5, its the same win-result but not the same count-result. Well, that would be a mistake by all accounts, even if AlphaGo made it.
Some of the moves AlphaGo played both in this game and the previous one are very definite mistakes, but they were irrelevant to the difference in the game.
> Well, that would be a mistake by all accounts, even if AlphaGo made it.
It is only a mistake to your human biases. AlphaGo literally has no conception of the margin of a win. It doesn't care either way how many points it wins by, as long as the win percentage is maximized.
Exactly. It's actually a pretty common behaviour for tree-searching game playing algorithms. Unless they're set up to explicitly give weight to faster wins or higher margin wins, they get into an unassailable position and then just kind of dick around. If every move's a win then it doesn't matter which move you make, right?
I think that his point is that it would be interesting to see AlphaGo pushed to its limit and playing a point-maximizing game. Which would be pretty terrifying.
>What is interesting to me is that the computer makes clear mistakes when its on the lead. Since it might find the chances to win equally among different scoring results, it often picks a weaker one.
When they interviewed the devs briefly, they said this is because AlphaGo doesn't really consider the score other whan winning, so it will pick a move it think is an 81% chance to win by one stone over an 80% chance to win by 10 stones. When it's ahead these moves can look like mistakes but a better way of describing them would be hedging.
do you think lee sedol should change his goal from trying to win all remaining three games to winning just one? in other words, sacrifice the next two games to learn about alphago and then try to win the final game.
Very impressive. Since there is a ton of hype about this and many media stories (at least NYTimes, with no citation at all) saying that this came 'a decade early', I think its worth looking over Yann LeCun retrospective on research in this area (https://www.facebook.com/yann.lecun/posts/10153340479982143). Clearly he was saying all this to preface the results of Facebook research in comparison to Google's, but I still think it is a very good overview of the history and shows the ideas did not come about suddenly. Quoting a few key things since the whole things is very long:
"The idea of using ConvNet for Go playing goes back a long time. Back in 1994, Nicol Schraudolph and his collaborators published a paper at NIPS that combined ConvNets with reinforcement learning to play Go. But the techniques weren't as well understood as they are now, and the computers of the time limited the size and complexity of the ConvNet that could be trained. More recently Chris Maddison, a PhD student at the University of Toronto, published a paper with researchers at Google and DeepMind at ICLR 2015 showing that a large ConvNet trained with a database of recorded games could do a pretty good job at predicting moves. The work published at ICML from Amos Storkey's group at University of Edinburgh also shows similar results. Many researchers started to believe that perhaps deep learning and ConvNets could really make an impact on computer Go.
...
Clearly, the quality of the tactics could be improved by combining a ConvNet with the kind of tree search methods that had made the success of the best current Go bots. Over the last 5 years, computer Go made a lot of progress through Monte Carlo Tree Search. MCTS is a kind of “randomized” version of the tree search methods that are used in computer chess programs. MCTS was first proposed by a team of French researchers from INRIA. It was soon picked up by many of the best computer Go teams and quickly became the standard method around which the top Go bots were built. But building an MCTS-based Go bots requires quite a bit of input from expert Go players. That's where deep learning comes in.
...
A good next step is to combine ConvNets and MCTS with reinforcement learning, as pioneered by Nicol Schraudolph's work. The advantage of using reinforcement learning is that the machine can train itself by playing many games against copies of itself. This idea goes back to Gerry Tesauro's “NeuroGammon,” a computer backgammon player that combined neural nets and reinforcement learning that beat the backgammon world champion in the early 1990s. We know that several teams across the world are actively working on such systems. Ours is still in development.
From what I gather, if you have a computer powerful enough, you can solve any game by simply applying Game Theory, as long as you can assign a numerical value to the possible outcomes.
AlphaGo plays some unusual moves that go clearly against any classically trained Go players. Moves that simply don't quite fit into the current theories of Go playing, and the world's top players are struggling to explain what's the purpose/strategy behind them.
I've been giving it some thought. When I was learning to play Go as a teenager in China, I followed a fairly standard, classical learning path. First I learned the rules, then progressively I learn the more abstract theories and tactics. Many of these theories, as I see them now, draw analogies from the physical world, and are used as tools to hide the underlying complexity (chunking), and enable the players to think at a higher level.
For example, we're taught of considering connected stones as one unit, and give this one unit attributes like dead, alive, strong, weak, projecting influence in the surrounding areas. In other words, much like a standalone army unit.
These abstractions all made a lot of sense, and feels natural, and certainly helps game play -- no player can consider the dozens (sometimes over 100) stones all as individuals and come up with a coherent game play. Chunking is such a natural and useful way of thinking.
But watching AlphaGo, I am not sure that's how it thinks of the game. Maybe it simply doesn't do chunking at all, or maybe it does chunking its own way, not influenced by the physical world as we humans invariably do. AlphaGo's moves are sometimes strange, and couldn't be explained by the way humans chunk the game.
It's both exciting and eerie. It's like another intelligent species opening up a new way of looking at the world (at least for this very specific domain). and much to our surprise, it's a new way that's more powerful than ours.