- We also have had some things that we have yet to fold back into
the main agent.
- One of the main problems with symbolic AI systems is learning
to ground symbols. Fawad has managed a system for labelling
the visual items.
- This has also been used to resolve ambiguous pronouns.
- Fawad has also done some stuff with learning shapes. The
2-D stuff is solid, and the 3-D stuff is getting that way.
- Kailash has done several associative memory tasks. Spatial
cognitive mapping (in CABot3) is a form of associative memory.
- We've also shown that any Finite State Automata can be duplicated
in fLIF neurons and that a fLIF system is Turing complete (out
for review at neurocomputing).
- Finally, Hina is doing stuff with sequence learning. It's
not hooked into a CABot, but it should be easily incorporated
as it runs on the basic CANT simulator.