On Fri, Nov 8, 2013 at 5:11 PM, Piaget Modeler
<[email protected]> wrote:
>
> Achieve infant intelligence a la Piaget.

Is the goal to test Piaget's theories of childhood development, or is
the goal to hope that the infant AI will develop into an adult AI?

I'm not sure how you could conduct such tests. For example, in the
A-not-B test ( http://en.wikipedia.org/wiki/A-not-B_error ) there are
competing explanations, but it seems you could build a model to
produce whatever results you want without answering the question.

> see Serving up Minds paper on the site (http://piagetmodeler.tumblr.com).

Really, you put your essay behind a paywall?

> The design is complete for now. There are always open questions in basic 
> research, we will tackle them as they arise but for now I think we have a 
> well defined system to do (i) ontology formation (ii) goal selection (ii) 
> goal achievement.

Do you need that for an infant intelligence? It seems that one thing
you do need, even at the sensory-motor stage, is a lot of computing
power to handle vision, hearing, and movement. I realize that neural
pattern recognition circuits are still shallow in young brains, but
this doesn't make the problem easier. A child's brain has more
synapses than an adult. Have you thought about how to test your models
if you can't run experiments lasting years on a 1 petaflop computer?

Also, we tend to thing of the infant mind as a blank slate, but it
really is very complex. Human DNA has about the same information
content as 300 million lines of code.

-- 
-- Matt Mahoney, [email protected]


-------------------------------------------
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