The cringey style discouraged me for a moment but as you powered your way through the video I saw what you were doing and stayed for the straight-shooting expla

The cringey style discouraged me for a moment but as you powered your way through the video I saw what you were doing and stayed for the straight-shooting explanations. It is a great description of the experiment from scratch, fast and entertaining enough that even knowing many of those things it did not get tedious. I’m very interested in the neurosymbolic approach, like you did using neural networks for perceiving the world and then logic rules for deciding what to do. I’ve been preaching about it for years but did not do anything by myself and this made me realize how much time has past. I’m now reading about DeepProbLog which I did not know about. I have some experience with Prolog and thought that I would have to build something like this, but now that I know it exists I’ll try it out. The most surprising part was to learn that neurosymbolic AI requires training data that follows the rules to begin with.