Cognitive Models
- What we really want to do is build cognitive models from neurons.
- So, we want to build systems that do cognitive (thinking) things
like people do.
- Roman and I have a neural model of the two choice task from
Friedman.
- We've got a model of the prepositional phrase attachment ambiguity
resolution. (I saw the girl with the telescope.)
- I've also got the best cognitive model of natural language parsing from
neurons. It parses in the correct time, generates semantic output,
and handles PP attachment ambiguity.
- What's nice about neural cognitive models compared to non-neural models
is that they have another constraint: neurons.
- We could even go so far as to start to link this directly to
human neural behaviour on the task.
- Of course our models have at most on the order of 100,000 neurons, not
the 100,000,000,000 neurons in a human brain.
