LESSWRONGAI Safety Subprojects
LW

AI Safety Subprojects

Sep 20, 2021 by Stuart_Armstrong

There are the AI safety subprojects designed for elucidating "model splintering" and "learning the preferences of irrational agents".

32Immobile AI makes a move: anti-wireheading, ontology change, and model splinteringΩ

Stuart_Armstrong

2y

3

14AI, learn to be conservative, then learn to be less so: reducing side-effects, learning preserved features, and going beyond conservatismΩ

Stuart_Armstrong

2y

4

30AI learns betrayal and how to avoid itΩ

Stuart_Armstrong

2y

4

17Force neural nets to use models, then detect theseΩ

Stuart_Armstrong

1y

8

15Preferences from (real and hypothetical) psychology papersΩ

Stuart_Armstrong

1y

0

15Finding the multiple ground truths of CoinRun and image classificationΩ

Stuart_Armstrong

1y

3