On Saturday, May 30, 2020 at 5:17:25 PM UTC+1, Tim Chow wrote:
> On Saturday, May 30, 2020 at 9:20:32 AM UTC-4, bgbl...@googlemail.com wrote:
> > 20 Million games is enough for a quite decent player......
> > if you know what you do. Maybe they should ask one who knows what to do ;)
> Yes, I was rather surprised when I first learned that 20 million games was
> enough training data for a decent bot, because I was naively guessing that
> it would require ten times as much as that. But apparently, even AlphaZero
> chess used only 44 million chess games, so these neural nets are somehow
> able to generalize incredibly well from (what seems to me like) relatively
> little data.
> > But as Tim pointed out: for the average customer to keep happy playing
> > strength might not the right way to go. Does one remember the "Big Bang
> > Board Games" Apple put on their Macs a decade or so ago. Much Bling-Bling
> > and abysmal game play....
> I have posted here before about the terrible level of play of the backgammon
> game on JetBlue. But Stick has a point...if the computer player is *too*
> weak then that can also be unappealing.

FWIW, my intuition was also that the low tens of millions would be far too few for
complex games like chess or backgammon.

You can take the man out of combinatorial game theory, but you can't take combinatorial
game theory out of the man.

Paul

peps...@gmail.com schrieb am Samstag, 3. Juni 2023 um 10:38:45 UTC+2:

> FWIW, my intuition was also that the low tens of millions would be far too few for
> complex games like chess or backgammon.

How many games a human needs ;) When you play with nets that are trained ony a couple of thousand games, you can watch how they graps more and more concepts.