Importance of Neural and Psychological Fidelity
- There are a lot of different algorithms and approaches that are
being used for AI.
- There are also an incredible number of problems that AI systems need
to deal with. The frame problem, learning, dynamic processing,
stability plasticity dilemma, knowledge representation, sensing,
language, symbol grounding ......
- While we may get to AI by some other mechanism, the human model
is the obvious one to follow. Human's (and animals) are the only
really intelligent things we know of. (A rat can learn a lot more
than any existing machine.)
- Of course we can be inspired by humans, but I think a better methodology
is to follow them.
- Psychologically we can develop and use cogntive architectures to
account for a range of human behaviour.
- Neurally, we can develop models that approximate single neuron
- We can also develop models that approximate suites of neural
- If we use these components as the basis of our AI systems,
we can look to the human solution when the models fail.