Found 1 document(s)
Page 1 of 1
-
AI systems can become better at achieving the goals we give them without becoming better at doing what we actually want. Studying this alignment problem rigorously requires a framework connecting measurable performance t...