| Jul-02-17 | | mnsek: gg human |
|
Feb-23-22
 | | An Englishman: Good Evening: Deep Blue looks like it could not find a plan in the major piece game after 21...Qxf6 and played much too passively. |
|
| Feb-23-22 | | Jamboree: This brings up an interesting question:
When was the last time that a human player defeated a computer program playing at full strength? In the '60s and '70s, chess-playing computers were laughably weak. Starting in the '80s, the best ones could play at Expert or Master strength. In the '90s they started getting very good, and by the '00s the top programs were all GM-level. By the 2010's they were nearly unbeatable, and by now (2022) the top ten computer programs are all rated 3400+ and are essentially unbeatable (by humans). At some point we crossed a threshold where computers became impossible for even the top humans to ever defeat. But when was that moment? It seems to have passed without anyone taking notice. Was it Kasparov beating Deep Blue in Game 1 of their 1997 rematch? Or...when? Surely no human these days could beat a 3500-rated Stockfish or Komodo or Fritz. I'm not talking about situations where the computer is intentionally "dumbed down" to make it an even match, or situations were there is a glitch, or a program playing on insufficient hardware, or ultra-fast time controls -- I'm talking about a top-20 program playing at full strength on the latest hardware under classical tournament conditions. When was the last time any of these lost to a human? |
|
| Feb-23-22 | | SkySports: <Jamboree: This brings up an interesting question: When was the last time that a human player defeated a computer program playing at full strength> If you don't consider (as I understand) rapid and blitz games, I'd say this victory of Ponomariov vs Fritz in 2005: Ponomariov vs Fritz, 2005 |
|
| Feb-23-22 | | goodevans: <An Englishman: Good Evening: Deep Blue looks like it could not find a plan in the major piece game...> Interesting how our tendency to anthropomorphise stretches to computers. These beasts aren't even capable of recognising a simple move transposition (see the discussion in Vitiugov vs J van Foreest, 2020) so clearly they don't 'plan' anything. Anything they come up with that resembles a plan is a product of what we overlay onto their calculations. |
|
Feb-23-22
 | | Troller: <When was the last time any of these lost to a human?> Depending on how gullible one is, there is also:
Adam Celander |
|
Feb-23-22
 | | LRLeighton: This doesn't entirely fit Jamboree's criteria, but in 2005, Arno Nickel, who is a correspondence GM, crushed Hydra, the world's strongest chess program at the time, in a correspondence match 2.5-0.5. This result was largely ignored, but Nickel had argued that the strength of computers was not in their brute-force calculation, but rather in their speed and endurance. A human couldn't possibly keep up with a super-computer in a classical game because a human wouldn't have time to calculate all of the variations. But Nickel's argument was that in a correspondence game, a strong human would have time to find the correct moves (Botvinnik had believed that the rise of computers would force humans to play slower, not faster, but apparently nobody else got the message). And at least in 2005, Nickel's argument appeared to be valid. |
|
Feb-23-22
| | LRLeighton: To be fair, Nickel was allowed to play as a centaur, specifically, he could use chess software that he could purchase over the counter. So he bought a $60 package of Fritz, which at the time, was weaker than Nickel was. So it's unclear how much Fritz helped him with his analyses. |
|
|
|
|
Kibitzing
