LESSWRONG
LW

2744
"Why Not Just..."

"Why Not Just..."

Aug 08, 2022 by johnswentworth

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

161Deep Learning Systems Are Not Less Interpretable Than Logic/Probability/Etc
Ω
johnswentworth
3y
Ω
56
165Godzilla Strategies
Ω
johnswentworth
3y
Ω
72
106Rant on Problem Factorization for Alignment
Ω
johnswentworth
3y
Ω
53
148Interpretability/Tool-ness/Alignment/Corrigibility are not Composable
Ω
johnswentworth
3y
Ω
13
212How To Go From Interpretability To Alignment: Just Retarget The Search
Ω
johnswentworth
3y
Ω
34
113Oversight Misses 100% of Thoughts The AI Does Not Think
Ω
johnswentworth
3y
Ω
49
82Human Mimicry Mainly Works When We’re Already Close
Ω
johnswentworth
3y
Ω
16
220Worlds Where Iterative Design Fails
Ω
johnswentworth
3y
Ω
30
187Why Not Just... Build Weak AI Tools For AI Alignment Research?
Ω
johnswentworth
3y
Ω
18
155Why Not Just Outsource Alignment Research To An AI?
Ω
johnswentworth
3y
Ω
50
150OpenAI Launches Superalignment Taskforce
Zvi
2y
40