[ExI] DeepMind wins Game1 in Go championship Match

[Anders Sandberg]

On 2016-03-10 18:26, William Flynn Wallace wrote:
> Anders:  On page 29 I have some plots from Katja Grace's excellent 
> review, which points out that algorithmic improvements typically 
> produce pretty drastic jumps in capability, unlike the individual 
> learning curve or the gradual collective improvement in chess.
>
> Anders, does this mean that an AI can exhibit insight learning? That 
> is, learning that proceeds through saltatory leaps - in humans we 
> would call those aha moments - insight?  Or did the paragraph above 
> say just that?  For a true insight curve it would have to start out 
> rather flat (meaning no insight yet).

Nah, the page 29 diagram is what algorithms do over time as humans have 
insights and improve them.

But actually, when you run some learning algorithms on some problems, 
you see jumps due to "insights" with flattish plateaus between. My 
favorite example is genetic algorithms like Avida on complex fitness 
landscapes - when they find good tricks they jump.

Now, proper insight learning in AI is mainly something people are 
modelling rather than doing, as far as I know.

-- 
Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20160310/bf78f9a2/attachment.html>