Oddbean new post about login | logout model architectures that can adjust their weights on the fly similar to how the human brain does will be the path to super intelligence
It’ll need a feedback loop, and understanding if the info is correct or not, so that is the main obstacle to overcome. Unlike humans with our nervous system to provide feedback, AI is still a bit behind 🐶🐾🫡
there's a company working on a device that retrains itself every night, but I don't believe that this is the same idea. however, it does have similar goals
Retraining is not the hard part, getting the correct feedback is 🐶🐾🫡