Love the title.
| Life In 19x19 http://lifein19x19.com/ |
|
| AlphaGo- Whatever You Do Is Wrong http://lifein19x19.com/viewtopic.php?f=10&t=13956 |
Page 1 of 2 |
| Author: | wineandgolover [ Wed Jan 11, 2017 1:59 am ] |
| Post subject: | AlphaGo- Whatever You Do Is Wrong |
Hi all, I've been going through the AlphaGo games and have reached a few conclusions. 1. Most of the games are easier to understand than contemporary pro games, and thus are great choices for amateurs to study. 2. When reviewing these games I often get the feeling that AlphaGo is asking questions of its human opponents. But there seems to be no right answer, and the human gets punished, regardless. I have chosen two games and made a video to demonstrate these two points. Please let me know your thoughts. |
|
| Author: | Bill Spight [ Wed Jan 11, 2017 9:09 am ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Love the title.
|
|
| Author: | dankenzon [ Wed Jan 11, 2017 12:45 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Nice comment. At my (limited) level I also notice that: there's a constant asking and feeling that you need to answer. The evident superiority in exchanges remind me what is said about Honinbo Shuei, the master of miai: seems that AlphaGo looks for exchanges where it can't loose. So then you come to ask: may be those questions shouldn't be answered? |
|
| Author: | Gomoto [ Wed Jan 11, 2017 2:12 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Great video! Very good advice: https://youtu.be/6rB2cYOeppQ?t=901 ("The way to win a won game is not to start another fight.") Just applied it successfull to one of my games.  I often tend to put myself under unnecessary pressure with bad decisions. (Trying to fight for live in my opponents territory, instead of aiming for a relaxed win by increasing my own potential when I am already ahead.) |
|
| Author: | Stefany93 [ Thu Jan 12, 2017 12:05 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
I wonder if they made the algorithm public yet? |
|
| Author: | Gotraskhalana [ Thu Jan 12, 2017 12:21 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
One thing that bothers me about the AlphaGo reviews: So ok, AlphaGo clearly plays better than the pros, and AlphaGo occasionally makes uncommon moves. Now everyone focusses on these uncommon moves and new fuseki and joseki are discussed. But what if AlphaGo's dominance comes from all the other moves, simply adding little advantages at every move? It would be perfectly possible for these uncommon moves to be wrong and refuted in a new version of AlphaGo. |
|
| Author: | Stefany93 [ Thu Jan 12, 2017 1:39 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Gotraskhalana wrote: One thing that bothers me about the AlphaGo reviews: So ok, AlphaGo clearly plays better than the pros, and AlphaGo occasionally makes uncommon moves. Now everyone focusses on these uncommon moves and new fuseki and joseki are discussed. But what if AlphaGo's dominance comes from all the other moves, simply adding little advantages at every move? It would be perfectly possible for these uncommon moves to be wrong and refuted in a new version of AlphaGo. Hence why I hate AlphaGo - Go players will turn like chess players or "cyborgs" and will start to play Computer Go, rather than human go. |
|
| Author: | oren [ Thu Jan 12, 2017 2:49 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Gotraskhalana wrote: Now everyone focusses on these uncommon moves and new fuseki and joseki are discussed The main reason they talk about the uncommon moves is that pros have historically ignored it. Since a greater than pro level strength plays it, then pros will of course start focusing on the moves that have been ignored. It's not just fuseki and joseki. AlphaGo analysis also had some interesting midgame points that hit many pros blind spots. |
|
| Author: | SpongeBob [ Thu Jan 12, 2017 3:48 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Nice video, thanks! It makes sense: if Alpha Go is stronger than his opponent, he does not need to pick a risky fight and instead of a tactical move he can play a strategic move, which will be easier to understand. |
|
| Author: | dankenzon [ Thu Jan 12, 2017 3:54 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Stefany93 wrote: I wonder if they made the algorithm public yet? They won't. Because there's no algorith. AlphaGo is a virtual brain. It learned and developed its own style. |
|
| Author: | dankenzon [ Thu Jan 12, 2017 3:56 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
SpongeBob wrote: Nice video, thanks! It makes sense: if Alpha Go is stronger than his opponent, he does not need to pick a risky fight and instead of a tactical move he can play a strategic move, which will be easier to understand. Actually I think it's the other way around: he is so strong because doesn't need to fight, and look for exchanges instead of fights. Also the lack of Ko fights is something to notice. |
|
| Author: | Knotwilg [ Fri Jan 13, 2017 2:16 am ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Great videos! I saw one game, which was discussed by Hay Lee, where the opening seemed equal until Ke Jie made a clear mistake. So in that particular case it didn't seem to be a case of "whatever you do is wrong". I concur with a few comments: 1) for such an overwhelming dominance, I had expected MORE uncommon moves (more moves and more uncommon). The bot plays surprisingly conventional go to my taste. For example, I would have bet on tengen as the eventual opening move. 2) the greatest advantage it seems to have is its estimation of exchanges (furikawari). I've seen swaps where I thought, this is a huge thing to give away, only to see the size of the sacrifice be reduced from all sides. 3) I find alphago easier to understand as well, but only with the hindsight that it wins everything. Cheers |
|
| Author: | Cassandra [ Fri Jan 13, 2017 2:39 am ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Knotwilg wrote: 1) for such an overwhelming dominance, I had expected MORE uncommon moves (more moves and more uncommon). The bot plays surprisingly conventional go to my taste. For example, I would have bet on tengen as the eventual opening move. If your "technical" level is near perfection, you are able to keep it "simple". |
|
| Author: | Bill Spight [ Fri Jan 13, 2017 4:04 am ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Knotwilg wrote: I concur with a few comments: 1) for such an overwhelming dominance, I had expected MORE uncommon moves (more moves and more uncommon). The bot plays surprisingly conventional go to my taste. I think that this reflects its better strategical judgement. AlphaGo's ordinary plays are bigger than they look. Quote: 2) the greatest advantage it seems to have is its estimation of exchanges (furikawari). This is another area where it strikes me that AlphaGo's play is like Go Seigen's.  Quote: 3) I find alphago easier to understand as well, but only with the hindsight that it wins everything. This aspect makes me think that the next generation of pros will be rather stronger than today's. This is the kind of thing that kids just pick up without much cogitation. Their intuition will be better. 
|
|
| Author: | Knotwilg [ Fri Jan 13, 2017 8:46 am ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Bill Spight wrote: This is another area where it strikes me that AlphaGo's play is like Go Seigen's. Isn't it sad the old master fell one year short of living the new new fuseki? |
|
| Author: | Stefany93 [ Fri Jan 13, 2017 3:12 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
dankenzon wrote: Stefany93 wrote: I wonder if they made the algorithm public yet? They won't. Because there's no algorith. AlphaGo is a virtual brain. It learned and developed its own style. Yeah... If that was possible we would be dead already by SkyNet. |
|
| Author: | Gomoto [ Fri Jan 13, 2017 3:28 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
He is right. You are a programmer, you should know it? |
|
| Author: | Monadology [ Fri Jan 13, 2017 4:36 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Stefany93 wrote: dankenzon wrote: Stefany93 wrote: I wonder if they made the algorithm public yet? They won't. Because there's no algorith. AlphaGo is a virtual brain. It learned and developed its own style. Yeah... If that was possible we would be dead already by SkyNet. http://www.youtube.com/watch?v=qv6UVOQ0F44 |
|
| Author: | Stefany93 [ Fri Jan 13, 2017 8:24 pm ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Quote: http://www.youtube.com/watch?v=qv6UVOQ0F44 Neat! (no pun intended)  I also found this one - http://senseis.xmp.net/?AlphaGo Second headline talks about the algorithm, soooooooo... Alpha Go, you are busted Joking aside, since AlphaGo was built on the NEAT principle with neural networks and what not, it proves its superiority to chess - in which chess programs merely store billions of games and just put the move with the highest chance of success. I suppose what they did was feed the programs millions of pro games and its gradually developed a style to beat them, as @dankenzon suggested. |
|
| Author: | mumps [ Sat Jan 14, 2017 3:43 am ] |
| Post subject: | Re: AlphaGo- Whatever You Do Is Wrong |
Stefany93 wrote: I suppose what they did was feed the programs millions of pro games and its gradually developed a style to beat them, as @dankenzon suggested. Actually not, as far as we're aware. Prior to the original Nature paper they only trained it using KGS games. Although they may have added pro games I doubt whether there are even one million of them available - GoGod doesn't even have 100,000. Even if they have a large number it's a good question as to whether the number they have will be significant compared to the total KGS and other online games they may have acquired. Jon |
|
| Page 1 of 2 | All times are UTC - 8 hours [ DST ] |
| Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group http://www.phpbb.com/ |
|