Amen to Don Dailey. He would be so proud.
From: Computer-go [mailto:computer-go-boun...@computer-go.org] On Behalf Of Jim O'Flaherty Sent: Thursday, March 10, 2016 6:49 PM To: computer-go@computer-go.org Subject: Re: [Computer-go] Finding Alphago's Weaknesses I think we are going to see a case of human professionals having drifted into a local optima in at least three areas: 1) Early training around openings is so ingrained in their acquiring their skill (optimal neural plasticity window), there has been very little new discovery around the first third of the game with almost all professionals relying fairly strongly on the already time tested josekis - AIs can use reading to explore closer and closer to the start of a game using less and less automatic patterns thereby confusing humans who have memorized those patterns 2) The middle of the board is so high in reading complexity, there has been little investment to figure out how to leverage it until mid game as it has been more expedient to focus on the corners and edges - AIs are going to get faster, better and deeper at reading through and then intentionally generating complexity 3) As a human's cognition tires, the probability of reading errors rises non-linearly which increases the probability of late mid-game and end game errors - I think AlphaGo has already progressed pretty far in the end game I'd consider these the three primary general vulnerabilities of human Go playing against any future AI. Given AlphaGo's training mechanism is actually search space exploration engine, it will slowly but surely explore and converge on more optimal play in all three of these domains significantly faster and cheaper than directly investing in and expending human cognition efforts; i.e. professionals studying to do the knowledge expansion and verification. And I think they will continue to optimize AlphaGo's algorithms in both human and self-play. The window where humans are going to be able to fish out a win against AlphaGo is rapidly closing...and it may have already closed. Other thoughts... I think we are going to see some fascinating "discoveries" of errors in existing very old josekis. At some point, I think we will even see one or two new ones discovered by AIs or by humans exploiting AIs. We are going to see some new center oriented fighting based on vastly more complex move sequences which will result in an substantial increase in resignations at the professional level against each other. Said a slightly different way...even if Lee Sedol figures how how to get a lead in a game during the opening, AlphaGo will just continue to elevate the board complexity with each move until it is just beyond its opponent's reading ability while staying well within it's own reading ability constraints. IOW, complexity is now an AIs advantage. AlphaGo doesn't have the human frailty of being nervous of a possible future mistake and then altering its priorities by pushing winning by a higher margin as a buffer against said future reading complexity mistake. IOW, AlphaGo is regulated by it's algorithm's prioritizing the probability of win higher than the amount of margin by which it could buffer for a win. What seems like a weakness is turning out to be one hell of a strength. Add to the fact that this kind of behavior by AlphaGo is denying it's opponent critical information about the state of the game which is more readily available in human-vs-human games; i.e. AlphaGo's will continue to converge towards calmer and calmer play in the face of chaotic play. And the calmer it becomes, the less "weakness surface area" it will have for a human to exploit in attempting a win. This is utterly fascinating to get to witness. I sure wish Don Daily was still here to get to enjoy this. On Thu, Mar 10, 2016 at 2:52 PM, Thomas Wolf <tw...@brocku.ca <mailto:tw...@brocku.ca> > wrote: With at most 2x361 or so different end scores but 10^{XXX} possible different games, there are at least in the opening many moves with the same optimal outcome. The difference between these moves is not the guaranteed score (they are all optimal) but the difficulty to play optimal after that move. And the human and computer strengths are rather different. On Thu, 10 Mar 2016, uurtamo . wrote: If that's the case, then they should be able to give opinions on best first moves, best first two move combos, and best first three move combos. That'd be interesting to see. (Top 10 or so of each). s. On Mar 10, 2016 12:37 PM, "Sorin Gherman" <sor...@gmail.com <mailto:sor...@gmail.com> > wrote: From reading their article, AlphaGo makes no difference at all between start, middle and endgame. Just like any other position, the empty (or almost empty, or almost full) board is just another game position in which it chooses (one of) the most promising moves in order to maximize her chance of winning. On Mar 10, 2016 12:31 PM, "uurtamo ." <uurt...@gmail.com <mailto:uurt...@gmail.com> > wrote: Quick question - how, mechanically, is the opening being handled by alpha go and other recent very strong programs? Giant hand-entered or game-learned joseki books? Thanks, steve On Mar 10, 2016 12:23 PM, "Thomas Wolf" <tw...@brocku.ca <mailto:tw...@brocku.ca> > wrote: My 2 cent: Recent strong computer programs never loose by a few points. They are either crashed before the end game starts (because when being clearly behind they play more desperate and weaker moves because they mainly get negative feadback from their search with mostly loosing branches and risky play gives them the only winning sequences in their search) or they win by resignation or win by a few points. In other words, if a human player playing AlphaGo does not have a large advantage already in the middle game, then AlphaGo will win whether it looks like it or not (even to a 9p player like Michael Redmond was surprised last night about the sudden gain of a number of points by AlphaGo in the center in the end game: 4:42:10, 4:43:00, 4:43:28 in the video https://gogameguru.com/alphago-2/) In the middle and end game the reduced number of possible moves and the precise and fast counting ability of computer programs are superior. In the game commentary of the 1st game it was mentioned that Lee Sedol considers the opening not to be his strongest part of the game. But with AlphaGo playing top pro level even in the opening, a large advantage after the middle game might simply be impossible to reach for a human. About finding weakness: In the absense of games of AlphaGo to study it might be interesting to get a general idea by checking out the games where 7d Zen lost on KGS recently. Thomas On Thu, 10 Mar 2016, wing wrote: One question is whether Lee Sedol knows about these weaknesses. Another question is whether he will exploit those weaknesses. Lee has a very simple style of play that seems less ko-oriented than other players, and this may play into the hands of Alpha. Michael Wing I was surprised the Lee Sedol didn't take the game a bit further to probe AlphaGo and see how it responded to [...complex kos, complex ko fights, complex sekis, complex semeais, ..., multiple connection problems, complex life and death problems] as ammunition for his next game. I think he was so astonished at being put into a losing position, he wasn't mentally prepared to put himself in a student's role again, especially to an AI...which had clearly played much weaker games just 6 months ago. I'm hopeful Lee Sedol's team has been some meta-strategy sessions where, if he finds himself in a losing position in game two, he turns it into exploring a set of experiments to tease out some of the weaknesses to be better exploited in the remaining games. On Thu, Mar 10, 2016 at 8:16 AM, Robert Jasiek <jas...@snafu.de <mailto:jas...@snafu.de> > wrote: > On 10.03.2016 00:45, Hideki Kato wrote: > > > such as solving complex semeai's and double-ko's, aren't solved yet. > > To find out Alphago's weaknesses, there can be, in particular, > > - this match > - careful analysis of its games > - Alphago playing on artificial problem positions incl. complex kos, > complex ko fights, complex sekis, complex semeais, complex endgames, > multiple connection problems, complex life and death problems (such as > Igo Hatsu Yoron 120) etc., and then theoretical analysis of such play > - semantic verification of the program code and interface > - theoretical study of the used theory and the generated dynamic data > (structures) > > -- > robert jasiek > _______________________________________________ > Computer-go mailing list > Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> > http://computer-go.org/mailman/listinfo/computer-go [1] Links: ------ [1] http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> http://computer-go.org/mailman/listinfo/computer-go _______________________________________________ Computer-go mailing list Computer-go@computer-go.org <mailto:Computer-go@computer-go.org> http://computer-go.org/mailman/listinfo/computer-go
_______________________________________________ Computer-go mailing list Computer-go@computer-go.org http://computer-go.org/mailman/listinfo/computer-go
