Integrate Learning
- Use Human-like components (simulated neurons)
- Keep in touch with reality by
- Solve problems like humans do (cognitive models and architecture)
- Develop interesting systems (AI programs)
- Integrate Learning
- Take advantage of emergent algorithms
- Repeat
- Turing himself said that he didn't expect a full-fledged adult
AI to be built, but instead to have a child AI built that
learned.
- Learning is clearly going to be necessary.
- Not just as simple facts, but at the very basis of cognition.
- Learning what things are, categories, symbols, and language are
all going to be necessary for a full-fledged AI.
- Fortunately neural systems are good at learning.
- Moreover there is a good deal of research in biological neural
learning and Hebbian learning is well explored computationally.
- My group has done a fair amount of work with learning, but our current
CABot agent does less than we'd like.
- We have ideas about how to proceed, but it's clear that
- learning will be central!
