CAs doing Multiple Tasks
- Feed Forward nets solve one problem. CA nets should solve lots.
(Unsupervised learning: Weaver and Kaplan vs. Hanson and Barr)
- Grand Challenges of Computing Science: Mind Brain Architecture
- Virtual Agent: get an agent to move about in a virtual environment
and learn a host of things.
- We might start doing this with games (next year or later).
- Robots: this grand challenge thing seems to like Robots.
- They're difficult for many reasons (e.g. IO), but they
would be great to get working.
- In both virtual and real agents, the point would be to get
the parts of the network working together.
- This kind of distributed processing and storage is
just not understood.
- (Oh, you should be able to build parallel hardware pretty
easily.)
