Aligning AGI Values to Human Values
Lex Fridman asks Ilya Sutskever if he thinks about the problems of continued alignment as AI systems are developed, to which Ilya Sutskever responds that there are definite ideas of how to train a value function that will be trained separately to recognize and internalize human judgments.In this clip
From this podcast

Lex Fridman Podcast
Ilya Sutskever: Deep Learning | Lex Fridman Podcast #94
Related Questions