Dwarkesh Patel interviews Ilya Sutskever on Safe Superintelligence's strategy, the problems with pre-training, how to improve generalization and how to ensure AGI goes well.