[FOM] Talk on Formal Verification

Walt Read walt.read at gmail.com
Wed Feb 17 20:33:48 EST 2016

> At the moment, machine learning can solve problems such as the following: consider a class C of hard problems (large polynomials to factor, for example). We have two algorithms at our disposal, F and G. We know that F outperforms G on some problems while G outperforms F on others. And it is simply too expensive to try both. It frequently happens that machine learning can be used to decide between F and G more effectively than human-written code. It’s a valuable technology, but it’s not intelligence.

Cf Wolpert and Macready's No Free Lunch results:


> Larry Paulson


More information about the FOM mailing list