In (highly contingent!) defense of interpretability-in-the-loop ML training

Steven Byrnes6 Feb 2026 16:32 UTC

68 points

3 comments3 min readLW link

Increasing AI Strategic Competence as a Safety Approach

Wei Dai3 Feb 2026 1:08 UTC

45 points

8 comments1 min readLW link

“Features” aren’t always the true computational primitives of a model, but that might be fine anyways

LawrenceC2 Feb 2026 18:41 UTC

18 points

0 comments5 min readLW link

Are there lessons from high-reliability engineering for AGI safety?

Steven Byrnes2 Feb 2026 15:26 UTC

90 points

6 comments8 min readLW link

Fitness-Seekers: Generalizing the Reward-Seeking Threat Model

Alex Mallen29 Jan 2026 19:42 UTC

83 points

4 comments17 min readLW link

AlgZoo: uninterpreted models with fewer than 1,500 parameters

Jacob_Hilton26 Jan 2026 17:30 UTC

176 points

7 comments10 min readLW link

(www.alignment.org)

New version of “Intro to Brain-Like-AGI Safety”

Steven Byrnes23 Jan 2026 16:21 UTC

58 points

1 comment19 min readLW link

When should we train against a scheming monitor?

Mary Phuong21 Jan 2026 20:48 UTC

22 points

4 comments5 min readLW link

No instrumental convergence without AI psychology

TurnTrout20 Jan 2026 22:16 UTC

68 points

7 comments6 min readLW link

(turntrout.com)

Pretraining on Aligned AI Data Dramatically Reduces Misalignment—Even After Post-Training

RogerDearnaley19 Jan 2026 21:24 UTC

102 points

12 comments11 min readLW link

(arxiv.org)

Could LLM alignment research reduce x-risk if the first takeover-capable AI is not an LLM?

Tim Hua19 Jan 2026 18:09 UTC

24 points

1 comment6 min readLW link

Desiderata of good problems to hand off to AIs

Jozdien19 Jan 2026 16:55 UTC

29 points

1 comment4 min readLW link

Gradual Paths to Collective Flourishing

Nora_Ammann19 Jan 2026 7:52 UTC

37 points

10 comments13 min readLW link

Test your interpretability techniques by de-censoring Chinese models

Khoi Tran, aryaj, Senthooran Rajamanoharan and Neel Nanda

15 Jan 2026 16:33 UTC

85 points

10 comments20 min readLW link

Apply to Vanessa’s mentorship at PIBBSS

Vanessa Kosoy14 Jan 2026 9:15 UTC

39 points

0 comments2 min readLW link

Global CoT Analysis: Initial attempts to uncover patterns across many chains of thought

Riya Tyagi, daria, Arthur Conmy and Neel Nanda

13 Jan 2026 20:40 UTC

48 points

0 comments18 min readLW link

Brief Explorations in LLM Value Rankings

Tim Hua, Josh Engels, Neel Nanda and Senthooran Rajamanoharan

12 Jan 2026 18:16 UTC

37 points

1 comment11 min readLW link

Practical challenges of control monitoring in frontier AI deployments

David Lindner and charlie_griffin

12 Jan 2026 16:45 UTC

19 points

0 comments1 min readLW link

(arxiv.org)

My 2003 Post on the Evolutionary Argument for AI Misalignment

Wei Dai6 Jan 2026 20:45 UTC

37 points

7 comments2 min readLW link

How hard is it to inoculate against misalignment generalization?

Jozdien6 Jan 2026 17:30 UTC

42 points

4 comments14 min readLW link