18 Dec 2024 17:19 UTC

412 points

51 comments10 min readLW link

Biological risk from the mirror world

jasoncrawford12 Dec 2024 19:07 UTC

283 points

24 comments7 min readLW link

(newsletter.rootsofprogress.org)

Frontier Models are Capable of In-context Scheming

Marius Hobbhahn, AlexMeinke, Bronson Schoen, rusheb, Jérémy Scheurer and Mikita Balesni

5 Dec 2024 22:11 UTC

201 points

24 comments7 min readLW link

Understanding Shapley Values with Venn Diagrams

Carson L6 Dec 2024 21:56 UTC

190 points

29 comments1 min readLW link

(medium.com)

Communications in Hard Mode (My new job at MIRI)

tanagrabeast13 Dec 2024 20:13 UTC

185 points

23 comments5 min readLW link

Gradient Routing: Masking Gradients to Localize Computation in Neural Networks

cloud, Jacob G-W, Evzen, Joseph Miller and TurnTrout

6 Dec 2024 22:19 UTC

151 points

11 comments11 min readLW link

(arxiv.org)

o1: A Technical Primer

Jesse Hoogland9 Dec 2024 19:09 UTC

140 points

17 comments9 min readLW link

(www.youtube.com)

Subskills of “Listening to Wisdom”

Raemon9 Dec 2024 3:01 UTC

133 points

16 comments42 min readLW link

“Alignment Faking” frame is somewhat fake

Jan_Kulveit20 Dec 2024 9:51 UTC

122 points

4 comments6 min readLW link

The Dangers of Mirrored Life

Niko_McCarty and fin

12 Dec 2024 20:58 UTC

118 points

7 comments29 min readLW link

(www.asimov.press)

The Dream Machine

sarahconstantin5 Dec 2024 0:00 UTC

116 points

6 comments12 min readLW link

(sarahconstantin.substack.com)

The o1 System Card Is Not About o1

Zvi13 Dec 2024 20:30 UTC

116 points

5 comments16 min readLW link

(thezvi.wordpress.com)

o3

Zach Stein-Perlman20 Dec 2024 18:30 UTC

115 points

57 comments1 min readLW link

Sorry for the downtime, looks like we got DDosd

habryka2 Dec 2024 4:14 UTC

109 points

13 comments1 min readLW link

Ablations for “Frontier Models are Capable of In-context Scheming”

AlexMeinke, Bronson Schoen, Marius Hobbhahn, Mikita Balesni, Jérémy Scheurer and rusheb

17 Dec 2024 23:58 UTC

107 points

1 comment2 min readLW link

A shortcoming of concrete demonstrations as AGI risk advocacy

Steven Byrnes11 Dec 2024 16:48 UTC

103 points

27 comments2 min readLW link

What Goes Without Saying

sarahconstantin20 Dec 2024 18:00 UTC

102 points

2 comments5 min readLW link

(sarahconstantin.substack.com)

When Is Insurance Worth It?

kqr19 Dec 2024 19:07 UTC

99 points

12 comments4 min readLW link

(entropicthoughts.com)

MIRI’s 2024 End-of-Year Update

Rob Bensinger3 Dec 2024 4:33 UTC

98 points

2 comments4 min readLW link

How to replicate and extend our alignment faking demo

Fabien Roger19 Dec 2024 21:44 UTC

97 points

0 comments2 min readLW link

(alignment.anthropic.com)

The “Think It Faster” Exercise

Raemon11 Dec 2024 19:14 UTC

95 points

13 comments14 min readLW link

AIs Will Increasingly Attempt Shenanigans

Zvi16 Dec 2024 15:20 UTC

94 points

4 comments26 min readLW link

(thezvi.wordpress.com)

Takes on “Alignment Faking in Large Language Models”

Joe Carlsmith18 Dec 2024 18:22 UTC

92 points

8 comments62 min readLW link

2024 Unofficial LessWrong Census/Survey

Screwtape2 Dec 2024 5:30 UTC

91 points

42 comments1 min readLW link

Parable of the vanilla ice cream curse (and how it would prevent a car from starting!)

Mati_Roy8 Dec 2024 6:57 UTC

84 points

20 comments3 min readLW link

Circling as practice for “just be yourself”

Kaj_Sotala16 Dec 2024 7:40 UTC

83 points

4 comments4 min readLW link

(kajsotala.fi)

Deep Causal Transcoding: A Framework for Mechanistically Eliciting Latent Behaviors in Language Models

Andrew Mack and TurnTrout

3 Dec 2024 21:19 UTC

83 points

7 comments41 min readLW link

Remap your caps lock key

bilalchughtai15 Dec 2024 14:03 UTC

80 points

15 comments1 min readLW link

Testing which LLM architectures can do hidden serial reasoning

Filip Sondej16 Dec 2024 13:48 UTC

79 points

9 comments4 min readLW link

Should you be worried about H5N1?

gw5 Dec 2024 21:11 UTC

79 points

2 comments5 min readLW link

(www.georgeyw.com)

Should there be just one western AGI project?

rosehadshar and Tom Davidson

3 Dec 2024 10:11 UTC

78 points

72 comments15 min readLW link

Best-of-N Jailbreaking

John Hughes, saraprice, Aengus Lynch, Rylan Schaeffer, Fazl, Henry Sleight, Ethan Perez and mrinank_sharma

14 Dec 2024 4:58 UTC

78 points

6 comments2 min readLW link

(arxiv.org)

Effective Evil’s AI Misalignment Plan

lsusr15 Dec 2024 7:39 UTC

77 points

9 comments3 min readLW link

Matryoshka Sparse Autoencoders

Noa Nabeshima14 Dec 2024 2:52 UTC

75 points

10 comments11 min readLW link

SAEBench: A Comprehensive Benchmark for Sparse Autoencoders

Can, Adam Karvonen, Johnny Lin, Curt Tigges, Joseph Bloom, chanind, Yeu-Tong Lau, Eoin Farrell, Arthur Conmy, CallumMcDougall, Kola Ayonrinde, Matthew Wearden, Sam Marks and Neel Nanda

11 Dec 2024 6:30 UTC

71 points

1 comment2 min readLW link

(www.neuronpedia.org)

The 2023 LessWrong Review: The Basic Ask

Raemon4 Dec 2024 19:52 UTC

71 points

25 comments9 min readLW link

Drexler’s Nanotech Software

PeterMcCluskey2 Dec 2024 4:55 UTC

65 points

9 comments4 min readLW link

(bayesianinvestor.com)

A Qualitative Case for LTFF: Filling Critical Ecosystem Gaps

Linch3 Dec 2024 21:57 UTC

64 points

2 comments1 min readLW link

RL, but don’t do anything I wouldn’t do

Gunnar_Zarncke7 Dec 2024 22:54 UTC

63 points

5 comments1 min readLW link

(arxiv.org)

A case for donating to AI risk reduction (including if you work in AI)

tlevin2 Dec 2024 19:05 UTC

61 points

2 comments1 min readLW link

Zen and The Art of Semiconductor Manufacturing

Recurrented9 Dec 2024 17:19 UTC

61 points

2 comments9 min readLW link

(futuring.substack.com)

Cognitive Work and AI Safety: A Thermodynamic Perspective

Daniel Murfet8 Dec 2024 21:42 UTC

61 points

7 comments4 min readLW link

Intricacies of Feature Geometry in Large Language Models

7vik, Lucius Bushnaq and Nandi

7 Dec 2024 18:10 UTC

59 points

0 comments12 min readLW link

An Illustrated Summary of “Robust Agents Learn Causal World Model”

Dalcy14 Dec 2024 15:02 UTC

57 points

2 comments10 min readLW link

Retrospective: PIBBSS Fellowship 2024

DusanDNesic, clem_acs and Lucas Teixeira

20 Dec 2024 15:55 UTC

54 points

1 comment4 min readLW link

o1 Turns Pro

Zvi10 Dec 2024 17:00 UTC

53 points

3 comments14 min readLW link

(thezvi.wordpress.com)

Luck Based Medicine: No Good Very Bad Winter Cured My Hypothyroidism

Elizabeth8 Dec 2024 20:10 UTC

53 points

3 comments2 min readLW link

(acesounderglass.com)

Anthropic leadership conversation

Zach Stein-Perlman20 Dec 2024 22:00 UTC

52 points

10 comments6 min readLW link

(www.youtube.com)

I Finally Worked Through Bayes’ Theorem (Personal Achievement)

keltan5 Dec 2024 2:04 UTC

51 points

6 comments9 min readLW link

A toy evaluation of inference code tampering

Fabien Roger9 Dec 2024 17:43 UTC

50 points

0 comments9 min readLW link

(alignment.anthropic.com)