Protocol Learning & Unextractable Protocol Models This Friday at dAGI Summit, @_AlexanderLong, Founder of @PluralisHQ, presents his keynote: “Protocol learning and unextractable protocol models”. He will cover decentralized model parallel training, its feasibility, and the possibilities it opens. Learn how splitting a model across nodes keeps weights confidential, enables secure training, and creates new ways to monetize AI in open protocols. 📍 Oct 24, San Francisco, Clancy Hotel 🎟 Register here:

People forget Policy Gradient based RL is the most data-inefficient form of training. Going to be major algorithmic advances in RL'ing the base models, probably using something like artificial curiosity (). But the current methods are not there.