Joe_Carlsmith

AI for AI safety

Paths and waysta­tions in AI safety

When should we worry about AI power-seek­ing?

• Joe_CarlsmithFeb 19, 2025, 8:13 AM

  4 points

  0 ∶ 0

  in reply to: Lizka’s comment on: Fake think­ing and real thinking

  Very glad to hear it, Lizka :) -- and thanks for letting me know.

What is it to solve the al­ign­ment prob­lem?

How do we solve the al­ign­ment prob­lem?

Fake think­ing and real thinking

Takes on “Align­ment Fak­ing in Large Lan­guage Models”

In­cen­tive de­sign and ca­pa­bil­ity elicitation

Op­tion control

Mo­ti­va­tion control

How might we solve the al­ign­ment prob­lem? (Part 1: In­tro, sum­mary, on­tol­ogy)

Video and tran­script of pre­sen­ta­tion on Oth­er­ness and con­trol in the age of AGI

What is it to solve the al­ign­ment prob­lem? (Notes)

Value frag­ility and AI takeover

A frame­work for think­ing about AI power-seeking

Lov­ing a world you don’t trust

On “first crit­i­cal tries” in AI alignment

• Joe_CarlsmithApr 9, 2024, 7:40 PM

  8 points

  0 ∶ 0

  on: Shar­ing Real­ity with Walt Whit­man [Video]

  Thanks for making this, Michel :)

On attunement