LESSWRONG"Why Not Just..."
LW

"Why Not Just..."

Aug 08, 2022 by johnswentworth

A compendium of rants about alignment proposals, of varying charitability.

138Deep Learning Systems Are Not Less Interpretable Than Logic/Probability/Etc
Ω
johnswentworth
1y
Ω
52
149Godzilla Strategies
Ω
johnswentworth
1y
Ω
67
75Rant on Problem Factorization for Alignment
Ω
johnswentworth
1y
Ω
48
126Interpretability/Tool-ness/Alignment/Corrigibility are not Composable
Ω
johnswentworth
1y
Ω
13
161How To Go From Interpretability To Alignment: Just Retarget The Search
Ω
johnswentworth
1y
Ω
32
91Oversight Misses 100% of Thoughts The AI Does Not Think
Ω
johnswentworth
1y
Ω
50
73Human Mimicry Mainly Works When We’re Already Close
Ω
johnswentworth
1y
Ω
16
177Worlds Where Iterative Design Fails
Ω
johnswentworth
1y
Ω
27
151Why Not Just... Build Weak AI Tools For AI Alignment Research?
Ω
johnswentworth
7mo
Ω
17
124Why Not Just Outsource Alignment Research To An AI?
Ω
johnswentworth
7mo
Ω
45
137OpenAI Launches Superalignment Taskforce
Zvi
3mo
38