----- Original Message ----

From: Matt Mahoney <[EMAIL PROTECTED]>

I don't claim that compression is simple.  It is not.  Text compression is
AI-complete.  The general problem is not even computable.

...I claim that compression can be used to measure intelligence.  I explain in 
more detail at http://cs.fit.edu/~mmahoney/compression/rationale.html

-- Matt Mahoney, [EMAIL PROTECTED]
-------------------------------------------

It will take me a while to read your paper.  However, I want to say that I am 
skeptical that you would be able to use compression to even measure 
intelligence.

I do think it might be worthwhile to come up with basic elements of 
intelligence, and these could include correlations of productive output from 
different algorithms or something like that.  But, from there you have to 
continue to build the  system.  It would be necessary to show how those 
elements can be combined to produce higher (or better) intelligence, and the 
Shannon/Hutter enthusiasts (along with everyone else) simply have not done 
this.  (I think the contemporary advancements in AI are probably due to faster 
memory access and parallelism as much as any achievement in AI software.)  But 
this means that you are advancing a purely speculative theory without any 
evidence to support it.

Right now I am working on my own religious journey (but mine is seriously 
religious interestingly enough) writing a polynomial time SAT program. Now 
let's say that this SAT theory actually worked and was followed by a theory 
that showed that it could be used both to advance AI and to compress data.  You 
might have a -I told you so- moment.  But I might then have a -so what- moment. 
 (I say that in a competitive but cordial way.)  Of course intelligence will 
involve some kind of compression method!  But so what?  It will also involve 
some kind of speculative method.  Does that mean that we can use speculation to 
'measure' intelligence?  Well, sure.  Someone might be able to devise a 
psychometric measure of speculative potential or something like that.  But this 
does not translate into an objective measure of intelligence until it is 
compared with thousands of subjects and integrated into a system that indicates 
that this particular measure of speculative
 potential can be correlated with other measures of intelligence and 
achievement.
Sometimes a compression algorithm is just a compression algorithm.
Jim Bromer


      

-------------------------------------------
agi
Archives: http://www.listbox.com/member/archive/303/=now
RSS Feed: http://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
http://www.listbox.com/member/?member_id=8660244&id_secret=103754539-40ed26
Powered by Listbox: http://www.listbox.com

Reply via email to