Video motivatring applying neuroevolution to AI
- I recently gave an invited talk on "The Case for Evolution in Engineering Brains" at the "MSU Symposium on New Frontiers in Cognitive, Evolutionary, and Computational Models of the Mind." It was placed online and may be of interest to this group because I tried to argue for why neuroevolution will be an important tool if we are eventually going to be able to create brain-like behavior in computers. Note that the talk is mostly aimed at those who are more familiar with or partial to other techniques (so it isn't really aimed at people in this group who already like the idea of neuroevolution), but it still may be interesting as an example of trying to motivate what we do to people from outside the area. In particular, I tried to argue that learning methods that focus on a single key principle (like deep learning or reinforcement learning) can't be the whole story because of the high chance that any fundamental principle will nevertheless still need to vary in innumerable subtle ways across the substrate of the brain:
There is also a question-and-answer session at the end.