When did the Computer Surpass Man?

[Dirt]

The nine games against Ponomariov and Adams were way above championship level, which I think outweighs the few others being somewhat less.

BayesElo output for all the GM games with Hydra I found for 2004-5:

Code: Select all

Rank Name                  Elo    +    - games score oppo. draws
   1 Hydra                2872  159   74    18   86%  2687   28%
   2 Veselin Topalov      2857  383  392     1   50%  2872  100%
   3 Alexander Khalifman  2857  383  392     1   50%  2872  100%
   4 Rustam Kasimdzhanov  2782  271  395     2   25%  2872   50%
   5 Sergey Karjakin      2738  341  917     1    0%  2872    0%
   6 Evgeny Vladimirov    2685  195  393     4   13%  2872   25%
   7 Adams,Mi             2629  164  387     6    8%  2872   17%
   8 Ruslan Ponomariov    2612  204  865     3    0%  2872    0%

Using a prior of 1, and an offset selected to make the Elo of Hydra's opponents match their weighted average (2687). Kasparov was around 2810 at the time.

What is the minus column? What is the minus 74 for hydra? Could that give hydra a 2798?

It means that BayesElo calculates that Hydra, according to these games, has a 95% chance of having a true Elo between 2798 and 3031. The uncertainty is probably really larger because some of the statistical assumptions are not strictly met. For instance, Adams first game with Hydra probably affected his later games, so the games were not fully independent of each other. Statistics never prove anything, but at some point we, as humans, will make up our minds. For me Deep Blue didn't play enough games to convince me, but Hydra did. Your mileage may vary.

[S.Taylor]

Anything from
when a machine first beat a chess expert in a public game, to when a machine wins all games against a world champion.

To narrow that down a little, I'd say anything between when a computer first beats an IM, to when a computer never loses a game to any human, would show that machine has surpassed man.

In the second case, that would easily translate something like when David Levy was beaten (was it a match actually?), till, not yet overtaken.

I go in for the "not yet overtaken".

When no human can any longer win even a single game, then machine would have overtaken man, for sure.

Until then, humans are still beating machines, and one might still be able to argue that humans still know how to do it.

[james uselton]

Terry, you said "Unless you're a very strong player you can't assess Deep Blue's play versus Kasparov's."

Are you saying you can assess Deep Blue's play versus Kasparov? I checked your Canadien chess rating--- 1750. I got a real kick from your post a while back about Beating GMs mano a mano.

Terry, you seem to know a little bit about a lot of things---unfortunately, chess aint one of 'em.