All 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 20242025

All Jan Feb MarAprMay Jun Jul Aug

All 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 242526 27 28 29 30

LLM Pareto Frontier But Live

winstonBosanApr 24, 2025, 9:22 PM

8 points

0 comments1 min readLW link

Modifying LLM Beliefs with Synthetic Document Finetuning

RowanWang, Johannes Treutlein, Avery, Ethan Perez, Fabien Roger and Sam Marks

Apr 24, 2025, 9:15 PM

70 points

12 comments2 min readLW link

(alignment.anthropic.com)

This prompt (sometimes) makes ChatGPT think about terrorist organisations

jakub_krysApr 24, 2025, 9:15 PM

30 points

13 comments1 min readLW link

Severe control over AI agents as a tool for mass-surveillance

Andrey SeryakovApr 24, 2025, 8:27 PM

2 points

0 comments3 min readLW link

Token and Taboo

GuiveApr 24, 2025, 8:17 PM

31 points

6 comments4 min readLW link

(guive.substack.com)

Trouble at Miningtown: Prologue

QuinnApr 24, 2025, 7:09 PM

19 points

0 comments4 min readLW link

Training-time schemers vs behavioral schemers

Alex MallenApr 24, 2025, 7:07 PM

44 points

9 comments6 min readLW link

Reward hacking is becoming more sophisticated and deliberate in frontier LLMs

KeiApr 24, 2025, 4:03 PM

94 points

6 comments1 min readLW link

Finding an Error-Detection Feature in DeepSeek-R1

keith_wynroeApr 24, 2025, 4:03 PM

15 points

0 comments7 min readLW link

Anticipating AI: Keeping Up With What We Build

Alvin ÅnestrandApr 24, 2025, 3:23 PM

2 points

0 comments11 min readLW link

(forecastingaifutures.substack.com)

Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?

Matrice JacobineApr 24, 2025, 2:11 PM

12 points

4 comments1 min readLW link

(limit-of-rlvr.github.io)

Academia as a happy place?

jow and pchvykov

Apr 24, 2025, 2:03 PM

9 points

0 comments19 min readLW link

“The Era of Experience” has an unsolved technical alignment problem

Steven ByrnesApr 24, 2025, 1:57 PM

115 points

48 comments23 min readLW link

AI #113: The o3 Era Begins

ZviApr 24, 2025, 1:40 PM

38 points

4 comments62 min readLW link

(thezvi.wordpress.com)

The Intelligence Curse: an essay series

L Rudolf L and lukedrago

Apr 24, 2025, 12:59 PM

68 points

10 comments2 min readLW link

Personal evaluation of LLMs, through chess

Karthik TadepalliApr 24, 2025, 7:01 AM

20 points

4 comments2 min readLW link

Intelligence explosion

samuelshadrachApr 24, 2025, 6:35 AM

2 points

0 comments4 min readLW link

(samuelshadrach.com)

Cognitive Dissonance is Mentally Taxing

SorenJApr 24, 2025, 12:38 AM

4 points

0 comments4 min readLW link

My Favorite Productivity Blog Posts

Parker ConleyApr 24, 2025, 12:32 AM

53 points

0 comments1 min readLW link

(parconley.com)

What Physically Distinguishes a Brain with False Beliefs Using a Swimming Pool Example

YanLyutnevApr 24, 2025, 12:01 AM

6 points

0 comments7 min readLW link

OpenAI Alums, Nobel Laureates Urge Regulators to Save Company’s Nonprofit Structure

garrisonApr 23, 2025, 11:01 PM

66 points

0 comments8 min readLW link

(garrisonlovely.substack.com)

What AI safety plans are there?

MichaelDickensApr 23, 2025, 10:58 PM

16 points

3 comments1 min readLW link

o3 Is a Lying Liar

ZviApr 23, 2025, 8:00 PM

84 points

26 comments9 min readLW link

(thezvi.wordpress.com)

Putting up Bumpers

Sam BowmanApr 23, 2025, 4:05 PM

52 points

14 comments2 min readLW link

The AI Belief-Consistency Letter

Knight LeeApr 23, 2025, 12:01 PM

−6 points

15 comments4 min readLW link

Jaan Tallinn’s 2024 Philanthropy Overview

jaanApr 23, 2025, 11:06 AM

223 points

8 comments1 min readLW link

(jaan.info)

[Question] Are we “being poisoned”?

TigerlilyApr 23, 2025, 5:11 AM

16 points

2 comments2 min readLW link

To Understand History, Keep Former Population Distributions In Mind

Arjun PanicksseryApr 23, 2025, 4:51 AM

240 points

13 comments2 min readLW link

(arjunpanickssery.substack.com)

Fish and Faces

EggsApr 23, 2025, 3:35 AM

8 points

6 comments2 min readLW link

Is alignment reducible to becoming more coherent?

Cole WyethApr 22, 2025, 11:47 PM

19 points

0 comments3 min readLW link

The EU Is Asking for Feedback on Frontier AI Regulation (Open to Global Experts)—This Post Breaks Down What’s at Stake for AI Safety

Katalina HernandezApr 22, 2025, 8:39 PM

60 points

13 comments9 min readLW link

Corrupted by Reasoning: Reasoning Language Models Become Free-Riders in Public Goods Games

David Guzman Piedrahita, Yongjin Yang and Zhijing Jin

Apr 22, 2025, 7:25 PM

24 points

3 comments5 min readLW link

Alignment from equivariance II—language equivariance as a way of figuring out what an AI “means”

hamishtodd1Apr 22, 2025, 7:04 PM

5 points

0 comments3 min readLW link

There is no Red Line

TachikomaApr 22, 2025, 6:28 PM

−13 points

1 comment3 min readLW link

Manifund 2025 Regrants

Austin ChenApr 22, 2025, 5:36 PM

21 points

0 comments5 min readLW link

(manifund.substack.com)

AISN#52: An Expert Virology Benchmark

Corin Katzke and Dan H

Apr 22, 2025, 5:08 PM

6 points

0 comments4 min readLW link

(newsletter.safe.ai)

Intuition in AI

Priyanka BharadwajApr 22, 2025, 3:15 PM

−1 points

2 comments2 min readLW link

Problems with Bayesianism: A Socratic Dialogue

B JacobsApr 22, 2025, 2:09 PM

3 points

1 comment14 min readLW link

(bobjacobs.substack.com)

Societal and technological progress as sewing an ever-growing, ever-changing, patchy, and polychrome quilt

Joel Z. Leibo, Wilcunningham, Seb Krier and Manfred Diaz

Apr 22, 2025, 1:21 PM

47 points

24 comments25 min readLW link

You Better Mechanize

ZviApr 22, 2025, 1:10 PM

74 points

6 comments20 min readLW link

(thezvi.wordpress.com)

Experimental testing: can I treat myself as a random sample?

avturchinApr 22, 2025, 12:34 PM

9 points

41 comments4 min readLW link

Family-line selection optimizer

lemonhopeApr 22, 2025, 7:16 AM

2 points

0 comments1 min readLW link

Accountability Sinks

Martin SustrikApr 22, 2025, 5:00 AM

423 points

57 comments15 min readLW link

(250bpm.substack.com)

Most AI value will come from broad automation, not from R&D

Matthew BarnettApr 22, 2025, 3:22 AM

10 points

6 comments2 min readLW link

(epoch.ai)

Estimat (8 Identities)

P. JoãoApr 22, 2025, 2:42 AM

4 points

0 comments3 min readLW link

A Letter to His Highness Louis XV, the King of France

testingthewatersApr 22, 2025, 12:51 AM

2 points

0 comments1 min readLW link

(aclevername.substack.com)

10 Principles for Real Alignment

AdriaanApr 21, 2025, 10:18 PM

−7 points

0 comments7 min readLW link

AE Studio is hiring!

Trent HodgesonApr 21, 2025, 8:35 PM

20 points

2 comments2 min readLW link

$500 Bounty Problem: Are (Approximately) Deterministic Natural Latents All You Need?

johnswentworth and David Lorell

Apr 21, 2025, 8:19 PM

92 points

24 comments3 min readLW link

More Than Just A, T, C, and G: Screening for Hidden Dangers in DNA Sequences

sgdApr 21, 2025, 8:12 PM

1 point

0 comments11 min readLW link