MIke,

> How would your approach replicate the classic visual problem for infants of:
> where has the toy car gone?  (when the parent puts it behind a box} - and oh
> look, it's still there, when the parent takes away the box ?

Not sure why you think that learning "object permanence" via experience
would be hard for any probabilistic learning system....  That is
pretty elementary
probabilistic reasoning, after the infant sees a few examples of
previously obscured
objects still being there...

> The first idea that my recent musings suggest about looking at any scene -
> is that the first thing a real world agent must look for in any scene, is
> what is moving - because that is the greatest source of both potential
> danger and interest.

Salience detection is a well known component of computer vision systems,
and ironically (in light of your negative view of probability theory) is often
handled using information-theoretic methods (which involve logarithms
of probabilities..)

-- Ben


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to