?

Log in

No account? Create an account

Previous Entry | Next Entry

anything else is not progress

so i had to (sort of) see for myself, putting all faith in the artificial-intelligence chess software stockfish that is ubiquitous, faith that it virtually always knows the optimal move, because i had to rely on it to play the role of optimal carlsen just as much as it promised to fulfill the part of optimal caruana. and while i'm still not totally convinced that the computer in considering its lines of play for both sides might not be discounting or disregarding a move that a human could uncover to foil one of its optimal lines of play it does seem apparent that even though this software is universally available for free to all takers, it is fully capable of proffering the closest thing to the most optimal moves, proffers them more often than human grandmasters.

so i set up the board with their game 8 at move 24 to avoid fabiano's (mis)play and take the alternate play suggested by ai, and then put it through its many many paces all the way to mate. and while the paths not taken (of worse probability) on either side were numerous, and certainly an infinitesimal amount of doubt remains as to what could have occurred down those roads, i was able to see for myself even with my limited grasp of complex strategy, that caruana, after his first 23 moves, had created a certain win for himself, if only he could recognize it.

it's probably the most intriguing thing for me about this championship match, that we're simultaneously seeing a demonstration of the peak of human skill at a particular endeavor while also being shown the limitations of humans in certain endeavors amenable to machines. obviously i'm aware of deep blue and such but i never took a deep dive (ha! (groan...)) to have it demonstrated to me rather than just told to me.

Comments

( 3 comments — Leave a comment )
nightspore
Nov. 21st, 2018 10:14 pm (UTC)
Part of what humans can do is foresee that in certain situations other humans will play in a way that will eventually lead to a situation that an AI will never lose -- that is in to a forced mate. But when we get there the human intuition that there is a forced mate will be no match for the computer's actually calculating what the forced mate is. tl;dr A grandmaster and a computer working together will usually beat a computer working alone.

Edited at 2018-11-21 10:14 pm (UTC)
jones_casey
Nov. 26th, 2018 04:16 am (UTC)
it would certainly be interesting to see it play out, to see how often the human grandmaster chose to overrule its ally, especially the further the game has progressed. it's hard for me to imagine chess as very collaborative except in the sharing of insights that the ally can comprehend or be persuaded about since at the end of the day one person must make the line of play final decisions.
grashupfer
Nov. 26th, 2018 09:18 pm (UTC)

Ah Game 12 though!!

( 3 comments — Leave a comment )

Profile

blue legacy
jones_casey
cleaning up so well

Latest Month

December 2018
S M T W T F S
      1
2345678
9101112131415
16171819202122
23242526272829
3031     

Tags

Page Summary

Powered by LiveJournal.com
Designed by Witold Riedel