[extropy-chat] A Fully Autonomous Robot Builds Its Own Brain and Learns from Scratch

[Eliezer Yudkowsky]

Erik Starck wrote:
> 
> Is this Stephen Thaler-guy for real? I've been reading about his 
> creativity machines before, but his home page looks like your average 
> cult leaders':
> http://www.imagination-engines.com/index.htm
> 
> His idea? Neural nets with noise. Gillette, GE and Boeing are 
> appearantly buying his services. Should I be impressed?

I certainly have not been impressed.  Speaking as an earnest devotee of the 
second law of thermodynamics, I don't believe that creativity is the result 
of injecting entropy into a system.

It's hard to dismiss AI theories unless they're really, really crackpot 
(e.g. Mentifex), because AI theories are so hard to understand.  I still 
don't understand Ben Goertzel's Novamente system, for example, even after 
reading a book about it.  There are claims made on Thaler's site (such as 
application to realtime trading) which I don't think the system described 
should be able to do at all, implying that they're not describing 
everything or that the claims are bogus.  Possibly the former.  The front 
page reads like it was written by a cult *follower*; if you hunt around 
there's some technical stuff that's less breathless and that makes sense 
when considered in isolation, though the enormous overinflated claims 
attached to it are par for the AI course.

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence