hypothetical frameworks for safe and scalable ai research Your browser does not support the video tag. View original