Neural Networks can do it all after they emerge symbolic representations, a way to represent logic and mathematics. A neural network that just "feels" that Fer

Neural Networks can do it all after they emerge symbolic representations, a way to represent logic and mathematics. A neural network that just "feels" that Fermat's Last Theorem is true, is much less intelligent than one that can produce the proof of it and present that to us, so we can trust that what it's saying is true. If you can't do symbolic manipulation, you are not really intelligent, artificial or otherwise, I would say.