Other Things and after CABot5
- Some other things we're hoping to do in CABot5 include:
- make it more realistic neurobiologically with specific ad hoc
simulations,
- translate it to chip; Steve Furber says he'll have a chip (well 500)
capable of simulating a billion of our fLIF neurons in real time,
- speech recogntion,
- make a neuro-cognitive architecture (which means propose an initial
one).
- Beyond CABot5 I'd like to make an industrially viable system.
- Current AI is good at domain specific problems. I think
a post-CABot5 agent would be really good at learning a relatively
broad domain. This could be industrially viable. Any ideas for
domains?
- Improve on the neural cognitive architecture and models.
- Put it in a physical robot. I think our vision and language
understanding will be good, but motor skills and physical sensation
will be weak.
- Try the Turing test.
- Think about new problems to solve.
- I really don't think we'll pass the Turing test in 5 or even 10 years,
so we're not going to be running out of problems.
