[ExI] alpha zero

[spike]

-----Original Message-----
From: extropy-chat [mailto:extropy-chat-bounces at lists.extropy.org] On Behalf Of Alejandro Dubrovsky

>...BTW, while this is, to me, the biggest newest in the history of chess, it follows naturally from their AlphaGo Zero paper published on Nature a whole month and a half ago: https://www.nature.com/articles/nature24270...

Ja I agree.  At top level human chess, no one just shows up one day beating everybody.  They have a history.  One of the oddities of our time is having guys play their computers at home and get really good, then show up at a tournament already in really good form.  But this will be a first (as far as I know) for computer chess where a program without a past wins it all.  If it does what they describe, it is a couple hundred Elo above this year's champion (Houdini 6.03).  A gap that size at that level doesn't happen.  I don't think it ever has in the history of TCEC.

>...I don't think there's anything sinister in the whole affair, except for the usual DeepMind showboating...

If they did what they say they did, it ain't showboating, it's just telling it like it really is.  That will be so cool, assuming...

>... (and the usual question of why they denied StockFish their opening book and only gave it 1GB of hashtable memory)...

...assuming they didn't do something to StockFish to artificially weaken it.  I don't have StockFish, so I can't determine that question myself.

>...I expect people to indepedently replicate this, albeit at a slower pace due to limited resources, within the next couple of years...

I expect this will be proven true or false in the next couple weeks with first reports showing up anytime now.  StockFish owners will take those games and see if SF really does what the paper reports, given these settings.

I can't shake the suspicion that something is amiss.  Evidence to follow.

spike