Charbel-Raphaël comments on Don’t you think RLHF solves outer alignment?

[Charbel-Raphaël]

RLHF gives rewards which can withstand more optimization before producing unintended

Do you have a link for that please?