All 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 20232024

AllJanFeb Mar Apr May Jun Jul Aug Sep Oct Nov

All1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

There is way too much serendipity

Malmesbury19 Jan 2024 19:37 UTC

365 points

56 comments7 min readLW link

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

evhub, Carson Denison, Meg, Monte M, David Duvenaud, Nicholas Schiefer and Ethan Perez

12 Jan 2024 19:51 UTC

305 points

95 comments3 min readLW link

(arxiv.org)

Gentleness and the artificial Other

Joe Carlsmith2 Jan 2024 18:21 UTC

291 points

33 comments11 min readLW link

The case for ensuring that powerful AIs are controlled

ryan_greenblatt and Buck

24 Jan 2024 16:11 UTC

258 points

66 comments28 min readLW link

MIRI 2024 Mission and Strategy Update

Malo5 Jan 2024 0:20 UTC

219 points

44 comments8 min readLW link

Toward A Mathematical Framework for Computation in Superposition

Dmitry Vaintrob, jake_mendel and Kaarel

18 Jan 2024 21:06 UTC

203 points

18 comments63 min readLW link

The impossible problem of due process

mingyuan16 Jan 2024 5:18 UTC

195 points

64 comments14 min readLW link

This might be the last AI Safety Camp

Remmelt and Linda Linsefors

24 Jan 2024 9:33 UTC

195 points

34 comments1 min readLW link

Introducing Alignment Stress-Testing at Anthropic

evhub12 Jan 2024 23:51 UTC

182 points

23 comments2 min readLW link

Without fundamental advances, misalignment and catastrophe are the default outcomes of training powerful AI

Jeremy Gillen and peterbarnett

26 Jan 2024 7:22 UTC

160 points

60 comments57 min readLW link

Making every researcher seek grants is a broken model

jasoncrawford26 Jan 2024 16:06 UTC

158 points

41 comments4 min readLW link

(rootsofprogress.org)

What’s up with LLMs representing XORs of arbitrary features?

Sam Marks3 Jan 2024 19:44 UTC

157 points

61 comments16 min readLW link

Apologizing is a Core Rationalist Skill

johnswentworth2 Jan 2024 17:47 UTC

152 points

42 comments5 min readLW link

Deep atheism and AI risk

Joe Carlsmith4 Jan 2024 18:58 UTC

146 points

22 comments27 min readLW link

What good is G-factor if you’re dumped in the woods? A field report from a camp counselor.

Hastings12 Jan 2024 13:17 UTC

137 points

22 comments1 min readLW link

Processor clock speeds are not how fast AIs think

Ege Erdil29 Jan 2024 14:39 UTC

132 points

55 comments2 min readLW link

The case for training frontier AIs on Sumerian-only corpus

Alexandre Variengien, Charbel-Raphaël and Jonathan Claybrough

15 Jan 2024 16:40 UTC

130 points

15 comments3 min readLW link

Notice When People Are Directionally Correct

Chris_Leong14 Jan 2024 14:12 UTC

129 points

8 comments2 min readLW link

An even deeper atheism

Joe Carlsmith11 Jan 2024 17:28 UTC

125 points

47 comments15 min readLW link

A Shutdown Problem Proposal

johnswentworth and David Lorell

21 Jan 2024 18:12 UTC

125 points

61 comments6 min readLW link

Steering Llama-2 with contrastive activation additions

Nina Panickssery, Wuschel Schulz, NickGabs, Meg, evhub and TurnTrout

2 Jan 2024 0:47 UTC

123 points

29 comments8 min readLW link

(arxiv.org)

Why I take short timelines seriously

NicholasKees28 Jan 2024 22:27 UTC

121 points

29 comments4 min readLW link

Gender Exploration

sapphire14 Jan 2024 18:57 UTC

113 points

25 comments5 min readLW link

(open.substack.com)

Four visions of Transformative AI success

Steven Byrnes17 Jan 2024 20:45 UTC

112 points

22 comments15 min readLW link

Practically A Book Review: Appendix to “Nonlinear’s Evidence: Debunking False and Misleading Claims” (ThingOfThings)

tailcalled3 Jan 2024 17:07 UTC

111 points

25 comments2 min readLW link

(thingofthings.substack.com)

The case for more ambitious language model evals

Jozdien30 Jan 2024 0:01 UTC

110 points

30 comments5 min readLW link

′ petertodd’’s last stand: The final days of open GPT-3 research

mwatkins22 Jan 2024 18:47 UTC

109 points

16 comments45 min readLW link

Being nicer than Clippy

Joe Carlsmith16 Jan 2024 19:44 UTC

109 points

32 comments27 min readLW link

2023 in AI predictions

jessicata1 Jan 2024 5:23 UTC

107 points

35 comments5 min readLW link

Catching AIs red-handed

ryan_greenblatt and Buck

5 Jan 2024 17:43 UTC

104 points

22 comments17 min readLW link

Deceptive AI ≠ Deceptively-aligned AI

Steven Byrnes7 Jan 2024 16:55 UTC

96 points

19 comments6 min readLW link

Almost everyone I’ve met would be well-served thinking more about what to focus on

Henrik Karlsson5 Jan 2024 21:01 UTC

95 points

8 comments11 min readLW link

(www.henrikkarlsson.xyz)

RAND report finds no effect of current LLMs on viability of bioterrorism attacks

StellaAthena25 Jan 2024 19:17 UTC

94 points

14 comments1 min readLW link

(www.rand.org)

On the abolition of man

Joe Carlsmith18 Jan 2024 18:17 UTC

88 points

18 comments41 min readLW link

The Aspiring Rationalist Congregation

maia10 Jan 2024 22:52 UTC

86 points

23 comments10 min readLW link

Sparse Autoencoders Work on Attention Layer Outputs

Connor Kissane, robertzk, Arthur Conmy and Neel Nanda

16 Jan 2024 0:26 UTC

83 points

9 comments18 min readLW link

Some Vacation Photos

johnswentworth4 Jan 2024 17:15 UTC

82 points

0 comments1 min readLW link

An Introduction To The Mandelbrot Set That Doesn’t Mention Complex Numbers

Yitz17 Jan 2024 9:48 UTC

82 points

11 comments9 min readLW link

Palworld development blog post

bhauth28 Jan 2024 5:56 UTC

81 points

12 comments1 min readLW link

(note.com)

Survey of 2,778 AI authors: six parts in pictures

KatjaGrace6 Jan 2024 4:43 UTC

80 points

1 comment2 min readLW link

Universal Love Integration Test: Hitler

Raemon10 Jan 2024 23:55 UTC

76 points

65 comments9 min readLW link

When “yang” goes wrong

Joe Carlsmith8 Jan 2024 16:35 UTC

72 points

6 comments13 min readLW link

We need a Science of Evals

Marius Hobbhahn and Jérémy Scheurer

22 Jan 2024 20:30 UTC

71 points

13 comments9 min readLW link

The True Story of How GPT-2 Became Maximally Lewd

Writer and Jai

18 Jan 2024 21:03 UTC

70 points

7 comments6 min readLW link

(youtu.be)

[Repost] The Copenhagen Interpretation of Ethics

mesaoptimizer25 Jan 2024 15:20 UTC

70 points

4 comments5 min readLW link

(web.archive.org)

Epistemic Hell

rogersbacon27 Jan 2024 17:13 UTC

70 points

20 comments14 min readLW link

InterLab – a toolkit for experiments with multi-agent interactions

Tomáš Gavenčiak, Ada Böhm and Jan_Kulveit

22 Jan 2024 18:23 UTC

69 points

0 comments8 min readLW link

(acsresearch.org)

[Question] Will quantum randomness affect the 2028 election?

Thomas Kwa and habryka

24 Jan 2024 22:54 UTC

66 points

52 comments1 min readLW link

OpenAI’s Preparedness Framework: Praise & Recommendations

Akash2 Jan 2024 16:20 UTC

66 points

1 comment7 min readLW link

The Perceptron Controversy

Yuxi_Liu10 Jan 2024 23:07 UTC

65 points

18 comments1 min readLW link

(yuxi-liu-wired.github.io)