LESSWRONG2022 MIRI Alignment Discussion
LW

2022 MIRI Alignment Discussion

A collection of MIRI write-ups and conversations about alignment released in 2022, following the Late 2021 MIRI Conversations.

277Six Dimensions of Operational Adequacy in AGI ProjectsΩ

Eliezer Yudkowsky

9mo

65

781AGI Ruin: A List of LethalitiesΩ

Eliezer Yudkowsky

10mo

657

261A central AI alignment problem: capabilities generalization, and the sharp left turnΩ

9mo

48

270On how various plans miss the hard bits of the alignment challengeΩ

8mo

78

170The inordinately slow spread of good AGI conversations in ML

9mo

61

181A note about differential technological developmentΩ

4mo

32

122Brainstorm of things that could force an AI team to burn their leadΩ

8mo

6

154AGI ruin scenarios are likely (and disjunctive)Ω

8mo

37

63Where I currently disagree with Ryan Greenblatt’s version of the ELK approachΩ

6mo

7

153Why all the fuss about recursive self-improvement?

9mo

61

53Humans aren't fitness maximizersΩ

6mo

45

112Warning Shots Probably Wouldn't Change The Picture MuchΩ

6mo

40

72What does it mean for an AGI to be 'safe'?Ω

6mo

29

101Don't leave your fingerprints on the futureΩ

6mo

30

105Niceness is unnaturalΩ

5mo

19

88Contra shard theory, in the context of the diamond maximizer problemΩ

5mo

17

56Notes on "Can you control the past"Ω

5mo

41

150Decision theory does not imply that we get to have nice thingsΩ

4mo

53

123Superintelligent AI is necessary for an amazing future, but far from sufficientΩ

4mo

47

96How could we know that an AGI system will have good consequences?Ω

5mo

24

67Distinguishing test from trainingΩ

4mo

10

294A challenge for AGI organizations, and a challenge for readersΩ

Rob Bensinger, Eliezer Yudkowsky

4mo

32

99Thoughts on AGI organizations and capabilities workΩ

Rob Bensinger, So8res

4mo

17