lovetheusers answers Don’t you think RLHF solves outer alignment?

lovetheusers 30 Dec 2022 4:08 UTC
2 points
−2
- RLHF increases desire to seek power
https://twitter.com/AnthropicAI/status/1604883588628942848
- RLHF causes models to say what we want to hear (behave sycophantically)
https://twitter.com/goodside/status/1605471061570510848