our current thoughts on hard questions about how AI systems should behave: 1) less biased defaults, 2) lots of user customization within very broad bounds, 3) public input on bounds and defaults

→ View original post on X — @sama,