All 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 20252026

All Jan Feb Mar Apr May JunJulAug Sep

All 1 2 3 4 5 6 7 8 91011 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Natural Language Autoencoders are summarizers, but do they have to be?

Andrey Anurin9 Jul 2026 19:54 UTC

10 points

1 comment22 min readLW link

Where Do LLM Values Come From?

lilysun004, Arthur Conmy and Josh Engels

9 Jul 2026 19:54 UTC

17 points

1 comment18 min readLW link

Selective Optimism: a critique of AI 2040

Richard_Ngo9 Jul 2026 19:43 UTC

231 points

13 comments8 min readLW link

(www.mindthefuture.info)

Rogue ASI Can’t Stay Aligned to Itself

absenteewarlord9 Jul 2026 19:38 UTC

8 points

2 comments1 min readLW link

Criticism against “unembedded FDT” doesn’t apply to FDT

Fernand09 Jul 2026 18:35 UTC

2 points

5 comments3 min readLW link

Your Prompt-Injection Defense Metric Might Be Lying to You

sahilraut9 Jul 2026 17:53 UTC

5 points

0 comments8 min readLW link

When is misalignment just a bug?

Yoav Hollander9 Jul 2026 17:07 UTC

16 points

5 comments11 min readLW link

(blog.foretellix.com)

What is the computational substance of the axiom of choice?

tailcalled9 Jul 2026 16:53 UTC

26 points

2 comments6 min readLW link

Skeptical of the TESCREAL Acronym? Read This.

philosophytorres9 Jul 2026 16:27 UTC

−41 points

12 comments20 min readLW link

AI 2040: Plan A

Daniel Kokotajlo, elifland, Thomas Larsen, romeo, brendanhalstead and ryan_greenblatt

9 Jul 2026 16:25 UTC

568 points

135 comments1 min readLW link

(www.ai-2040.com)

How big is the Sun? How could you figure it out?

Elliott Thornley9 Jul 2026 16:24 UTC

101 points

9 comments6 min readLW link

Some Thoughts on The Environment Problem in Agent Training

TheVinci9 Jul 2026 15:41 UTC

10 points

2 comments3 min readLW link

(www.tarantulabs.com)

The Cube Theory of Partially Grasped Concepts

Mateusz Bagiński9 Jul 2026 15:40 UTC

24 points

2 comments5 min readLW link

Debate with Self-Play Best-of-N Optimization

Dewi Gould, Sam Martin, Alejandro Aristizabal, Simon Marshall and Jacob Pfau

9 Jul 2026 15:29 UTC

53 points

2 comments14 min readLW link

AI #176 Part 1: Doing It Live

Zvi9 Jul 2026 14:31 UTC

36 points

2 comments31 min readLW link

(thezvi.wordpress.com)

Optimiser Choice Can Amplify or Suppress Emergent Misalignment

Jason R Brown, Patrick Leask and Lev McKinney

9 Jul 2026 10:00 UTC

63 points

2 comments4 min readLW link

Announcing our $160M grant from Coefficient Giving

Geoffrey Irving, Jesse Hoogland, Alex HT, Jacob Pfau, Daniel Murfet, Marco Cozzi and Stan van Wingerden

9 Jul 2026 7:58 UTC

50 points

0 comments2 min readLW link

Interpretability is becoming increasingly uninterpretable

jcksanderson9 Jul 2026 7:04 UTC

34 points

1 comment4 min readLW link

(jcksanderson.com)

Persistent Latent Misalignment, a new dimension of misalignment?

Florian_Dietz9 Jul 2026 4:56 UTC

15 points

2 comments1 min readLW link

Because 8 ≈ e², Anthropic’s researcher uplift is plausibly >2x

Thomas Kwa9 Jul 2026 4:30 UTC

54 points

9 comments12 min readLW link

(metr.org)

Transformers Resist Their Own Architecture

Zach Baker9 Jul 2026 0:57 UTC

11 points

0 comments15 min readLW link

Modular Pretraining Enables Access Control

E.Roland and cloud

9 Jul 2026 0:03 UTC

71 points

2 comments9 min readLW link

(alignment.anthropic.com)

There Should Be More AI Safety Hubs

Seth Lifland8 Jul 2026 23:09 UTC

13 points

1 comment3 min readLW link

Models are blind outside the J-space. NLAs aren’t.

Pranav Viswanath8 Jul 2026 23:09 UTC

28 points

4 comments9 min readLW link

Solving the BlueDot Puzzle TAIS: The Velocity Ring

Karine Levonyan8 Jul 2026 23:03 UTC

8 points

0 comments4 min readLW link

(karinelevonyan.github.io)

Action as Choice Expressed Through Movement Toward a Goal: a Framework for Overcoming Inaction

techandsundry8 Jul 2026 23:02 UTC

7 points

0 comments4 min readLW link

Optimum number of items to inspect before buying one

bilibili8 Jul 2026 21:23 UTC

18 points

0 comments4 min readLW link

Free will as a model parameter

darshanav8 Jul 2026 21:11 UTC

10 points

1 comment6 min readLW link

Find funding, fast

Austin Chen8 Jul 2026 21:10 UTC

43 points

0 comments3 min readLW link

(manifund.substack.com)

Childhood and Education #20: Phones and Screens

Zvi8 Jul 2026 19:41 UTC

37 points

1 comment17 min readLW link

(thezvi.wordpress.com)

AI Safety Hong Kong Reading Group: Machine Learning and Human Values

Schizoid Rentoid8 Jul 2026 19:19 UTC

1 point

0 comments1 min readLW link

Can the U.S. and China Deny AI?

Felix Choussat and Oscar

8 Jul 2026 18:45 UTC

20 points

0 comments14 min readLW link

(secondstrike.substack.com)

A Real-Life Example of an Aligned System Killing Hundreds of People

London L.8 Jul 2026 18:43 UTC

3 points

8 comments4 min readLW link

Notes on technical alignment via human-like social drives

Steven Byrnes8 Jul 2026 18:30 UTC

74 points

17 comments42 min readLW link

Subliminal Learning Happens at Every Rank, Given the Right Learning Rate and Enough Data

Lawrence Feng8 Jul 2026 17:32 UTC

60 points

5 comments7 min readLW link

AI Safety at the Frontier: Paper Highlights of May & June 2026

gasteigerjo8 Jul 2026 17:19 UTC

16 points

0 comments10 min readLW link

Why study proto-training gaming as an adversarial alignment failure mode?

Puria, Edward James Young and Cam

8 Jul 2026 17:07 UTC

55 points

0 comments7 min readLW link

Why study alignment interventions on pre-RL checkpoints?

Edward James Young, Puria and Cam

8 Jul 2026 17:07 UTC

66 points

2 comments6 min readLW link

Artificial Ambassadors: The Three Cases for Sending LLMs to Deep Space

Ekaterina Matson8 Jul 2026 16:08 UTC

7 points

0 comments12 min readLW link

(medium.com)

Reframing LessWrong-style decision theory as “commitment theory”

Elias Schmied8 Jul 2026 15:25 UTC

61 points

33 comments11 min readLW link

Why I’m a moral anti-realist but may be unable to convince you

Nina Panickssery8 Jul 2026 15:23 UTC

23 points

12 comments7 min readLW link

(blog.ninapanickssery.com)

AI Forecasting in 2026: What 11 Analyses Say

Ben Wilson8 Jul 2026 14:39 UTC

25 points

0 comments18 min readLW link

(www.metaculus.com)

Why models game evals might matter as much as whether they do it

Igor Ivanov8 Jul 2026 14:30 UTC

30 points

8 comments4 min readLW link

The Empathetic Poor

a unemployed pastor- de S Brito Gabriel8 Jul 2026 14:21 UTC

5 points

2 comments16 min readLW link

How Valuable are Traditions?

martinkunev8 Jul 2026 13:44 UTC

10 points

5 comments5 min readLW link

How much slower does takeoff go with 10× less compute?

brendanhalstead8 Jul 2026 10:04 UTC

46 points

4 comments4 min readLW link

Human Empowerment in an AI Society

Alvin Ånestrand8 Jul 2026 9:01 UTC

24 points

0 comments36 min readLW link

The mosquito bucket of doom works

dominicq8 Jul 2026 8:18 UTC

134 points

10 comments5 min readLW link

(blog.d11r.eu)

[Linkpost]Six Story Prompts I Want to Read

Linch8 Jul 2026 6:12 UTC

15 points

0 comments11 min readLW link

(linch.substack.com)

the allston lab rat existential framing

somayya8 Jul 2026 2:45 UTC

−3 points

0 comments2 min readLW link