Last Modified May 19, 1995

Joanna's Cog Page

I [was!] working with a bunch of other people on Cog, the humanoid robot. We are trying to create something approaching human intelligence, at least of a six-month old child. The reason we are doing this on a robot is that we believe that human intelligence, like all animal intelligence, is fundamentally composed of behaviors designed to move our bodies in an intelligent way. Eventually, we develop more advanced intellectual capabilities by learning, which we do by experimenting with our environment. The Cog group believes you can't really have something like human intelligence without having a similar body and similar formative experiences as humans do.

To read some papers about Cog, (and see a lot of pictures) trace the pointer above there. Here is a paper I wrote about Cog's learning called The Role and Nature of Learning for an Artifact. It talks about a lot of the biological, psychological, and mostly ethological research that our approach is based on, and then does some rampant speculation on what we might do with Cog that is pretty fun.

Here is some Q & A about Cog from an over-the-net interview I did -- other surfers might have similar questions.

Early this year [1995] I did some basic vision work for Cog -- letting it perceive depth using parallax and using this to establish the fundamental concept of "background," here's a paper about that which also talks about the problems of integrating behavior modules on a large scale project.


page author: Joanna Bryson