Tuesday, December 06, 2005

Metrics For Faulty Inferences

It is my firm conviction that with current logic based inference techniques it is not possible to get “internet scale” or “Turing test able” reasoning. I also believe that for this kind of reasoning both soundness and completeness will have to be given up.
To develop algorithms doing incomplete and unsound reasoning we need metrics that tell us just HOW unsound and incomplete this algorithms results are.

Tags: , ,