Rob gave an interesting talk titled “F%ck Computational Linguistics.” I had some questions about ‘chunking’ and why it’s used in NLP.
Today we had a guest speaker: Doug Lenat. He is one of leading researchers in common sense reasoning in artificial intelligence. His company Cycorp released OpenCyc, which is another common sense knowledge base. He presented some interesting examples of Cyc in practice. It’s too bad is stay was short. I had some general questions that I didn’t ask, as I didn’t want to take up the little time he had.
In class, Dr. Lieberman presented some of the differences in methodology between Cyc and Common Sense. I wonder if, in application, does it matter that a common sense assertion is factually false? Is this another aspect of “fail-soft?”
He also showed an example of the flaws of common sense problem solving:
Cheap are apartments are rare.
Rare things are expensive.
Therefore… cheap apartments are expensive.