Re: [agi] Bayesian surprise attracts human attention

2009-01-15 Thread Bob Mottram
2009/1/15 Ronald C. Blue ronb...@u2ai.us:
 Bayesian surprise attracts human attention http://tinyurl.com/77p9xo



Sounds interesting.  In my opinion any research carried out at
universities using public money should be available to the public,
without additional charges.


---
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=8660244id_secret=126863270-d7b0b0
Powered by Listbox: http://www.listbox.com


Re: [agi] Bayesian surprise attracts human attention

2009-01-15 Thread Richard Loosemore

Bob Mottram wrote:

2009/1/15 Ronald C. Blue ronb...@u2ai.us:

Bayesian surprise attracts human attention http://tinyurl.com/77p9xo




In my opinion any research carried out at
universities using public money should be available to the public,
without additional charges.


Agreed.

 Sounds interesting.

Well... not so much.

This is what I have gleaned from the abstract.

The researchers demolished an old idea about attention (from the 1950s) 
and pretended to test a new idea.


In fact the new idea is so well-established that everyone takes it for 
granted:  attention shift is driven by surprise or novelty or 
unexpectedness.


Perspective:

Recall that there was once a theory that heat was a fluid that passed 
between bodies (caloric).  Then, that old theory was superceded by the 
new idea that heat and temperature were just average molecular motion.


Now imagine that someone came along and published a paper claiming to 
have discovered this second idea, years after it had become common 
knowledge.  But what you find, when you read the details, is that what 
they actual did was just take a thermometer and draw a scale on the 
outside of it - a completely arbitrary scale that they made up - and 
then declare that BECAUSE they slapped the scale on the outside, 
THEREFORE they have validated or proved or demonstrated the idea of 
temperature being molecular motion, rather than the movements of caloric 
fluid.


This is exactly what these people have just done with the notion of 
surprise.  It adds nothing useful to what we know.


If they had shown that there is a mechanism that actually computes the 
bayesian probabilities, then governs the attention shift using the 
results of that calculation, that would have been progress.  But just 
finding something that covaries with novelty is like shooting fish in a 
barrel.


Of course, it's not like these are the only people making this kind of 
non-progress  ;-)






Richard Loosemore


---
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=8660244id_secret=126863270-d7b0b0
Powered by Listbox: http://www.listbox.com