AI ALIGNMENT FORUM
"Why Not Just..."
AF

"Why Not Just..."

Aug 08, 2022 by johnswentworth

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

52Deep Learning Systems Are Not Less Interpretable Than Logic/Probability/Etc

3y

14

47Godzilla Strategies

3y

19

42Rant on Problem Factorization for Alignment

2y

32

56Interpretability/Tool-ness/Alignment/Corrigibility are not Composable

2y

4

76How To Go From Interpretability To Alignment: Just Retarget The Search

2y

24

38Oversight Misses 100% of Thoughts The AI Does Not Think

2y

23

37Human Mimicry Mainly Works When We’re Already Close

2y

4

66Worlds Where Iterative Design Fails

2y

17

67Why Not Just... Build Weak AI Tools For AI Alignment Research?

2y

2

Review

42Why Not Just Outsource Alignment Research To An AI?

2y

7

Review

48OpenAI Launches Superalignment Taskforce

1y

0