Check out The AI Doc, now streaming.
Updates Ramana Kumar and Scott Garrabrant argue that the AGI safety community should begin prioritizing “approaches that work well in the absence of human models”: [T]o the extent that human modelling is a good idea, it is important to do...
