[ExI] alpha zero

[spike]

On Behalf Of John Clark
Subject: Re: [ExI] alpha zero

 

 

 

On Thu, Dec 7, 2017 at 9:19 PM, Dave Sill <sparge at gmail.com <mailto:sparge at gmail.com> > wrote:

 

 

​> ​>…The problem is knowing when they're as intelligent as we are. AlphaGO isn't even close. If it wants to argue otherwise I'd happy to engage it.

 

​>…If it isn't even close to being intelligent how can it start from zero knowledge and teach itself to play chess…in less than a day and easily beat ​any human who ever lived?  John K Clark  

 

Ja of course this is impressive, but consider all the ways this could be achieved that would look like it had trained itself from nothing in a day.  An eager press corps could report the program was given nothing but the rules of chess, when in reality it was given the StockFish chess engine with no opening book.  That would constitute being given chess rules only, if the phrase is interpreted broadly.  In fact, that would be a good approach to the problem: StockFish code is highly optimized already, so there is no need to reinvent that wheel.  

 

So imagine you some kind of training which slightly modifies StockFish, improves it, play a match with the original version.  Then you have made a huge contribution to machine learning, to computer chess, to human knowledge, with no intent to deceive.  The press might report it slightly incorrectly, which is what I vaguely suspect happened here.  Note I am not accusing anyone and cheering for the participants.  

 

Regardless of how this is proven out, I think DeepMind has done something cool here.  I just can’t convince myself the wilder claims made by the reporters are all correct.  

 

spike

 

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20171210/0e66176e/attachment.html>