“Demis Hassabis believes that at most one or two major undiscovered ideas are needed to achieve AGI on top of existing techniques.”