News for hackers. Stuff that matters.

A comment on Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model. Parent comment

Re:Mastering Atari, Go, Chess and Shogi by Planning with a L... by sanxiyn on Thursday November 21, 2019 @12:59AM (#21591784)
It should be pointed out AlphaZero plays better than Deep Blue. I think comparing computational resource usage is important, but direct comparison only makes sense with equivalent performance level.
[ Parent | on HN ]
Re:Mastering Atari, Go, Chess and Shogi by Planning with a L... by mindgam3 on Thursday November 21, 2019 @01:14AM (#21591842)
I did point this out at the top of my original comment.

“I mean, it's cool that computers are getting even better at chess and all“

> direct comparison only makes sense with equivalent performance level

This makes no sense to me. 50% increase in performance can be compared to 50% increase in processing power to evaluate level of brute force-ness.

[ Parent | on HN ]
Re:Mastering Atari, Go, Chess and Shogi by Planning with a L... by summerlight on Friday November 22, 2019 @02:33PM (#21609160)
> This makes no sense to me. 50% increase in performance can be compared to 50% increase in processing power to evaluate level of brute force-ness.

Computational complexity theory taught us that fundamental difficulty of solving specific types of problems does not always linearly scale with the size of the problems. I guess the same logic applies to the quality of the output?

[ Parent | on HN ]
Re:Mastering Atari, Go, Chess and Shogi by Planning with a L... by SonOfLilit on Thursday November 21, 2019 @07:40AM (#21593483)
Yes, but how is "50% better" measured? On what scale?
[ Parent | on HN ]
Re:Mastering Atari, Go, Chess and Shogi by Planning with a L... by mindgam3 on Thursday November 21, 2019 @11:05AM (#21595558)
Easy. Chess uses the Elo rating system.

https://en.m.wikipedia.org/wiki/Elo_rating_system

[ Parent | on HN ]
Re:Mastering Atari, Go, Chess and Shogi by Planning with a L... by SonOfLilit on Wednesday November 27, 2019 @06:42PM (#21653297)
I mean that ELO is not linear and there is no obvious ratio operation for ELO scores. How much is better than 2,600? Is it 5,200? Is it the rating at which you'd have 2:1 odds of winning a match? (Wikipedia says 200 points represent 74% chance of winning, so somewhere below 2,800) There's just no commonly agreed meaning of "50% better Chess player".
[ Parent | on HN ]