Yuxi on the Wired (@layer07_yuxi) 's Twitter Profile
Yuxi on the Wired

@layer07_yuxi

[email protected]:~$ whoami
tech enby (agi/asi pronouns)
(math ⊗ physic ⊗ comput).ist

ID: 1123471073775882240

linkhttps://yuxi-liu-wired.github.io/ calendar_today01-05-2019 06:15:51

1,1K Tweet

1,1K Followers

55 Following

Yuxi on the Wired (@layer07_yuxi) 's Twitter Profile Photo

the problem is that anything that finally works in psychology stops getting called psychology and becomes neuroscience/psychometry/behavioral sciences the problem is that anything that finally works in philosophy stops getting called philosophy and becomes a science

Yuxi on the Wired (@layer07_yuxi) 's Twitter Profile Photo

"Alignment faking" is folk psychology. Interpreting Reinforcement Learning folk-psychologically won't work because biological RL is very different (unclear if superior or inferior) to current RL algorithms. Even TD learning seems different from biological TD (with dopamine).

Yuxi on the Wired (@layer07_yuxi) 's Twitter Profile Photo

[a few days ago] mutual: "Why is grok retarded? It attributes this painting to a Frenchman." us: "Due to extensive RLEF (RL to Elon's Feedback), the most modal personality of Grok is someone that makes stuff up all the time despite what the sources say."

[a few days ago]
mutual: "Why is grok retarded? It attributes this painting to a Frenchman."
us: "Due to extensive RLEF (RL to Elon's Feedback), the most modal personality of Grok is someone that makes stuff up all the time despite what the sources say."