Re: Brain protocol

[Kevin Crowston]

>        Then, let's think of it. I would, for starters, propose a
>single function,
>-(float*) feel: (float*);
>
>        This would be the basic function, to feed an information
>vector to the brain, and return an "action" vector.
>
>        Comments, ideas, criticism anyone?

It would help to know what you expect the "brain" of an agent to do.  What
I want is for my agents to listen for other agents to tell them things and
then to act on what they're told.  So, I'd almost go for something like

-(void) heard: (char *) ;

Except, that instead of a (char *) maybe there'd be some structure to the
communication, like lists of atoms, or even KQML constructs (not that I
know KQML).  Any actions taken would be done directly by other message
sends in the brain, so maybe there'd be some protocol for that end too; one
clear candidate would be "say".  I could also see connections for memory
and learning, to make it easier to test the effects of different
algorithms.

In other words, I guess I want something like Shoam's agents in Agent
Oriented Programming, and the brain protocol would encapsulate the
connections between the messages received and the messages sent in
response.

Kevin