Timeline for answer to Mistakes in mathematics, false illusions about conjectures by Joel David Hamkins
Current License: CC BY-SA 3.0
Post Revisions
31 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jun 13, 2016 at 10:24 | comment | added | Joel David Hamkins | My answer doesn't mention Go, although the comments do. But indeed, the recent triumph of Alpha-Go (en.wikipedia.org/wiki/AlphaGo) seems to have been based in part on neural learning algorithms, although still involving extensive tree searching. | |
| Apr 14, 2016 at 0:06 | comment | added | Gerry Myerson | Maybe it's time to update the comments on Go. | |
| Jun 27, 2014 at 14:36 | vote | accept | user39115 | ||
| Jun 27, 2014 at 14:36 | |||||
| Jun 27, 2014 at 14:35 | vote | accept | user39115 | ||
| Jun 27, 2014 at 14:35 | |||||
| Jun 27, 2014 at 13:51 | vote | accept | user39115 | ||
| Jun 27, 2014 at 14:35 | |||||
| Jun 27, 2014 at 13:51 | vote | accept | user39115 | ||
| Jun 27, 2014 at 13:51 | |||||
| Jun 27, 2014 at 13:51 | vote | accept | user39115 | ||
| Jun 27, 2014 at 13:51 | |||||
| Jun 9, 2014 at 16:51 | comment | added | Anixx | This is not strictly a mathematical problem, but rather a technological one. One could prove even without computers available that a powerful machine can play chess well using brute force. | |
| Jun 9, 2014 at 16:38 | comment | added | user9072 | From said article: "Go is the only one in which computers don’t stand a chance against humans." | |
| Jun 9, 2014 at 14:39 | comment | added | Joel David Hamkins | Regarding the discussion in this thread, the following article seems relevant. wired.com/2014/05/the-world-of-computer-go | |
| Jun 7, 2014 at 19:01 | comment | added | Douglas Zare | I made two mistakes. First, I said 20 years ago when I should have said something like 15. My comments on the algorithms used, the strengths of go programs, and the meaning of the progress against human players were accurate and informative other than this slip. Second, I responded to quid. I'm done. | |
| Jun 7, 2014 at 15:23 | comment | added | user9072 | But since now you also clarified that you were not trying to be precise we can leave it there and everbody interested can study the link provided by Alexis Beingessner. | |
| Jun 7, 2014 at 15:22 | comment | added | user9072 | @DouglasZare I emphasized the 15 stones not the player being young; 15 stones is pretty extreme. I am not sure why you assume I underestimate something here. Anyway I provided a ranking estiamte too, and I consider something at best 8 kyu strength being described as "can reasonably compete against most amateurs" (what you claimed) as stranger than a 6 dan being described in this way. But I agree that AB's statements in isolation is misleading (but so were yours); but they provided a link from which you seem to quote and clarified that they were not that precise. | |
| Jun 7, 2014 at 14:11 | comment | added | Douglas Zare | ... since they have reached 6d on KGS (which is stronger than 6d in the AGA), and they have won against 9th dan professional players with handicaps of 4 stones multiple times, and with 3 stones at least once. That suggests they would be about 1000th in the world now. The Monte Carlo algorithms scale with parallel computing power, so you can quibble again about the precise definitions of the computer competitors, but if you find one where Alexis Beingessner's statements aren't grossly misleading let me know. | |
| Jun 7, 2014 at 14:05 | comment | added | Douglas Zare | @quid: Of course the definition of the player pool matters if you need the exact percentile, and I wasn't trying to be precise. I probably should have said something like 15 years ago instead of 20. It sounds like you are underestimating the strength of the kids who represented humans against the computers. Many professional go players turn pro at about 14 after years of intense training. The top 12-year-olds are close to professional level. Regardless, it's absurd to call current go programs "mediocre" and say that they "can reasonably compete against most amateurs" ... | |
| Jun 7, 2014 at 10:56 | comment | added | user9072 | Furthermore following a link on wikipedia there is "The strongest Go programs currently[1] play at around 8 - 13 kyu against human opponents" ([1] appears to have been written long ago but slightly less than twenty years ago). I would not consider a strength of 8 - 13 kyu as reasonably paraphrased as "can reasonably compete against most amateurs" (Of course, depending on what notion of amateur you have it might still be in line with your percentile claims. Yet they then would miss the point a bit and not be in line with what I would commonly undertstand under the quoted claims.) [@Douglas ] | |
| Jun 7, 2014 at 10:39 | comment | added | user9072 | @DouglasZare Wouldn't we also have to decide what an amateur is? Everybody who knowns the rules? Everybody who plays on a regular basis? Everybody that competes regularly in amateur tournaments? This hugely changes the percentile wise strength of an amateur player. So what is your notion of an amateur? I asksince for a notion of amateur I consider reasonable here, your claim regarding programms 20 years ago seems false to me and this seems in line with claims on page above that in 1994 the best (or one of the best) programms repeatedly lost against a youth player giving fifteen stones. | |
| Jun 4, 2014 at 16:41 | comment | added | Douglas Zare | @Alexis Beingessner: What percentile do you think corresponds to "can reasonably compete against most amateurs?" 50th? 99.99th? The normal meaning would be somewhere between 50th and 90th, which describes go programs from 20 years ago. More accurate is to say programs now "routinely beat almost all amateurs." Brute force is an accurate description of chess algorithms, which primarily improved from C player in 1975 to grandmaster around 1990 due to computing speed, not good algorithms. "Guess and check" doesn't describe the breakthroughs in go algorithms. | |
| Jun 4, 2014 at 15:32 | comment | added | Alexis Beingessner | @DouglasZare "Computers have reached the top amateur levels and perform at the level of some professional players." seems to match my claim of "can reasonably compete against most amateurs". Regardless, I was just poking a bit fun at Monte-Carlo techniques in the same vein as the answer (which summarizes mini-max and other related search techniques as "brute-force"). | |
| Jun 4, 2014 at 14:48 | comment | added | Douglas Zare | @Alexis Beingessner: Go programs have gotten much stronger in the last 5 years. It's a misrepresentation of the progress in go to say that the best computer programs "can reasonably compete against most amateurs" and that the algorithms are "guess and check." Computers have reached the top amateur levels and perform at the level of some professional players. One of the key ideas since 2006 was to randomize go, to predict which plays are strong on average if you turn go into a game of skill and chance. The analogue in chess would be IM or weak GM, roughly 1000th in the world. | |
| Jun 4, 2014 at 0:13 | comment | added | Simon Lyons | John McCarthy, the inventor of LISP, has a good quote about this: “Computer chess has developed much as genetics might have if the geneticists had concentrated their efforts starting in 1910 on breeding racing Drosophila. We would have some science, but mainly we would have very fast fruit flies." | |
| Jun 3, 2014 at 19:59 | comment | added | zeb | @Alexis the recent breakthrough is a bit more than just "guess and check". A big part of it comes from studying the explore/exploit trade off (applied to deciding which branches of the game tree you should think hardest about). | |
| Jun 3, 2014 at 18:21 | comment | added | Alexis Beingessner | @DavidFernandezBreton Computers are now simply "mediocre" at Go (Wikipedia lists a timeline of progress) . To my knowledge, the best algorithms can reasonably compete against most amateurs now. Hillariously, the AI breakthrough that has allowed this is the progression from "brute force search" to "guess and check". | |
| Jun 3, 2014 at 17:18 | comment | added | David Fernandez-Breton | However, it seems that computers are still very bad at go... | |
| Jun 3, 2014 at 16:02 | comment | added | Joel David Hamkins | @StefanKohl I agree with that, but that issue is orthogonal to the point of the metaphor. If you follow the link, you'll see that Donald Michie was saying that research on computer chess was fundamental, like Mendel's genetic experiments on peas or the work on drosphila. He says, "The use of chess now as a preliminary to the knowledge engineering and cognitive engineering of the future is exactly similar, in my opinion, to the work on drosophila. It should be encouraged in a very intense way, for these reasons." The subsequent computer chess developments reveal that we was probably wrong. | |
| Jun 3, 2014 at 13:17 | vote | accept | user39115 | ||
| Jun 3, 2014 at 13:17 | |||||
| Jun 3, 2014 at 13:17 | vote | accept | user39115 | ||
| Jun 3, 2014 at 13:17 | |||||
| Jun 3, 2014 at 13:13 | comment | added | Stefan Kohl♦ | Well -- even what insects do requires considerably more intelligence than playing chess. -- For example so far nobody can construct a collection of robots who act like an ant population and collaborate in building an anthill, or like a hive which collects nectar and builds honeycombs, etc.. | |
| Jun 3, 2014 at 12:53 | vote | accept | user39115 | ||
| Jun 3, 2014 at 12:53 | |||||
| S Jun 3, 2014 at 12:45 | history | answered | Joel David Hamkins | CC BY-SA 3.0 | |
| S Jun 3, 2014 at 12:45 | history | made wiki | Post Made Community Wiki by Joel David Hamkins |