PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

0:58

[Linkpost] “METR have released Time Horizons 1.1” by...

LessWrong (30+ Karma) ·

2026/02/05

This is a link post. I just found out that METR released an updated version of their time horizons work with extra tasks and different evaluation infrastructure. This was released on 29th Jan and I think has been overshadowed by the Moltbook stuff. Main points: Similar overall trend since 2021 50% time horizon doubling time went from 165 days with 1.0 to 131 days with 1.1 over the period since 2023 The top model, Claude 4.5 Opus, has gone from a 4h49 time horizon to 5h20 --- First

en-gb

14:20

“Finding Cruxes: Help Reality Punch You In the Face”...

LessWrong (30+ Karma) ·

2026/02/05

Figuring out "what would actually change your mind?" is among the more important rationality skills. Being able to do change your mind at all is a huge project. But, being able to do it quickly is much more useful than being able to do it at all. Because, then you can interleave the question "what would change my mind?" inside other, more complex planning or thinking. I find this much easier to learn "for real" when I'm thinking specifically about what would change my mind about my

en-gb

10:12

“Solemn Courage” by aysja

LessWrong (30+ Karma) ·

2026/02/05

Every so often it slips. It seems I am writing a book, but I can’t remember why. Somehow, the sentences are supposed to perform that impossible, intimate task: to translate my inner world into another. Yet they sit there so quiescent and small. How could an arrangement of words do anything, let alone reduce that ultimate threat to which it is all supposedly connected: the looming god machines? I look again at the monitor in which the words are contained and suddenly what once felt so raw and

en-gb

16:38

“Post-AGI Economics As If Nothing Ever Happens” by...

LessWrong (30+ Karma) ·

2026/02/04

When economists think and write about the post-AGI world, they often rely on the implicit assumption that parameters may change, but fundamentally, structurally, not much happens. And if it does, it's maybe one or two empirical facts, but nothing too fundamental. This mostly worked for all sorts of other technologies, where technologists would predict society to be radically transformed e.g. by everyone having most of humanity's knowledge available for free all the time, or everyone having an

en-gb

1:54

“New AI safety funding newsletter” by Bryce Robertson

LessWrong (30+ Karma) ·

2026/02/04

We’ve had feedback from several people running AI safety projects that it can be a pain tracking various funding sources and their application windows. To help make it easier, AISafety.com has launched the AI Safety Funding newsletter (which you can subscribe to here). It lists all newly announced funding opportunities relevant to individuals and orgs working on AI x-risk, and any opportunities which are closing soon. We expect posts to be about 2x/month. Opportunities will be sourced from the

en-gb

11:40

“Anthropic’s “Hot Mess” paper overstates its case...

LessWrong (30+ Karma) ·

2026/02/04

Author's note: this is somewhat more rushed than ideal, but I think getting this out sooner is pretty important. Ideally, it would be a bit less snarky. Anthropic[1] recently published a new piece of research: The Hot Mess of AI: How Does Misalignment Scale with Model Intelligence and Task Complexity? (arXiv, Twitter thread). I have some complaints about both the paper and the accompanying blog post. tl;dr The paper's abstract says that "in several settings, larger, more capable models are more

en-gb

7:13

″‘Inventing the Renaissance’ Review” by Commander Zander

LessWrong (30+ Karma) ·

2026/02/04

Inventing the Renaissance is a 2025 pop history book by historian of ideas Ada Palmer. I'm someone who rarely completes nonfic books, but i finished this one & got a lot of new perspectives out of it. It's a fun read! I tried this book after attending a talk by Palmer in which she not only had good insights but also simply knew a lot of new-to-me context about the history of Europe. Time to reduce my ignorance! ItR is a conversational introduction to the European Renaissance. It mostly talks

en-gb

12:37

“Concrete research ideas on AI personas” by...

LessWrong (30+ Karma) ·

2026/02/04

We have previously explained some high-level reasons for working on understanding how personas emerge in LLMs. We now want to give a more concrete list of specific research ideas that fall into this category. Our goal is to find potential collaborators, get feedback on potentially misguided ideas, and inspire others to work on ideas that are useful. Caveat: We have not red-teamed most of these ideas. The goal for this document is to be generative. Project ideas are grouped into: Persona & goal

en-gb

29:35

“Unless That Claw Is The Famous OpenClaw” by Zvi

LessWrong (30+ Karma) ·

2026/02/04

First we must covered Moltbook. Now we can double back and cover OpenClaw.Do you want a generally impowered, initiative-taking AI agent that has access to your various accounts and communicates and does things on your behalf? That depends on how well, safely, reliably and cheaply it works. It's not ready for prime time, especially on the safety side. That may not last for long. It's definitely ready for tinkering, learning and having fun, if you are careful not to give it access to anything

en-gb

20:47

“What did we learn from the AI Village in 2025?” by...

LessWrong (30+ Karma) ·

2026/02/03

Why This Project Exists Standard AI benchmarks test narrow capabilities in controlled settings. They tell us whether a model can solve a coding problem or answer a factual question. They don’t tell us what happens when you give an AI agent a computer, internet access, and an open-ended goal like "raise money for charity" or "build an audience on Substack." The AI Village exists to fill that gap. We run frontier models from OpenAI, Anthropic, Google, and others in a shared environment where they

en-gb

31:01

“Conditionalization Confounds Inoculation Prompting...

LessWrong (30+ Karma) ·

2026/02/03

Summary Conditionalization in Inoculation Prompting. Inoculation Prompting is a technique for selective learning that involves using a system prompt at train-time that won’t be used at test-time. When doing Inoculation-style training, using fixed arbitrary prompts at train time can prevent learned traits from generalizing to contexts that don’t include these prompts. We call this conditionalization: a learned trait is only expressed conditional on specific context features. This effect also happens

en-gb

3:51

“Three ways to make Claude’s constitution better” by...

LessWrong (30+ Karma) ·

2026/02/03

The evening after Claude's new constitution was published, about 15 AI safety FTEs and Astra fellows discussed the constitution, its weaknesses, and its implications. After the discussion, I compiled some of their most compelling recommendations: Increase transparency about the character training process. Much of the document is purposefully hedged and vague in its exact prescriptions; therefore, the training process used to instill the constitution is extremely load-bearing. We wish more of this

en-gb

3:13

“The Meta-Anthropic Argument” by RogerDearnaley

LessWrong (30+ Karma) ·

2026/02/03

Epistemic status: I just thought this up There is a well-known style of thinking called the anthropic argument (which has nothing to do with the AI frontier lab of the same name). It goes like this: Scientist 1: “X seems really unlikely! How come I'm observing it?" (where X is usualy something like the Hoyle resonance and the resulting triple-alpha process) Scientist 2: “Your mistake is that you’re estimating <span>_P(text{X})_</span> — you should instead be estimating

en-gb

44:46

“Cross-Layer Transcoders are incentivized to learn...

LessWrong (30+ Karma) ·

2026/02/03

Many thanks to Michael Hanna and Joshua Batson for useful feedback and discussion. Kat Dearstyne and Kamal Maher conducted experiments during the SPAR Fall 2025 Cohort. TL;DR Cross-layer transcoders (CLTs) enable circuit tracing that can extract high-level mechanistic explanations for arbitrary prompts and are emerging as general-purpose infrastructure for mechanistic interpretability. Because these tools operate at a relatively low level, their outputs are often treated as reliable descriptions of

en-gb

3:08

“Increasing AI Strategic Competence as a Safety...

LessWrong (30+ Karma) ·

2026/02/03

If AIs became strategically competent enough, they may realize that RSI is too dangerous because they're not good enough at alignment or philosophy or strategy, and potentially convince, help, or work with humans to implement an AI pause. This presents an alternative "victory condition" that someone could pursue (e.g. by working on AI strategic competence) if they were relatively confident about the alignment of near-human-level AIs but concerned about the AI transition as a whole, for example

en-gb

6:55

“Conditional Kickstarter for the “Don’t Build It”...

LessWrong (30+ Karma) ·

2026/02/03

tl;dr: You can pledge to join a big protest to ban AGI research at ifanyonebuildsit.com/march, which only triggers if 100,000 people sign up. The If Anyone Builds It website includes a March page, wherein you can pledge to march in Washington DC, demanding an international treaty to stop AGI research if 100,000 people in total also pledge. I designed the March page (although am not otherwise involved with March decisionmaking), and want to pitch people on signing up for the "March Kickstarter."

en-gb

6:36

“On Goal-Models” by Richard_Ngo

LessWrong (30+ Karma) ·

2026/02/02

I'd like to reframe our understanding of the goals of intelligent agents to be in terms of goal-models rather than utility functions. By a goal-model I mean the same type of thing as a world-model, only representing how you want the world to be, not how you think the world is. However, note that this still a fairly inchoate idea, since I don't actually know what a world-model is. The concept of goal-models is broadly inspired by predictive processing, which treats both beliefs and goals as

en-gb

6:00

“I finally fixed my footwear” by dominicq

LessWrong (30+ Karma) ·

2026/02/02

I’ve been wearing footwear of the wrong size, material, and shape for as long as I can remember, certainly at least 20 years. Only recently have I fixed this, and I come with great tidings: if you, too, hate wearing shoes, and the industrial revolution and its consequences, it is possible to be cured of at least one of these ailments. The problem is three-shaped, and is named as follows: wrong size, wrong material, and wrong shape. 1. Wrong size My algorithm for buying shoes was roughly this: Be

en-gb

15:40

“Are there lessons from high-reliability engineering...

LessWrong (30+ Karma) ·

2026/02/02

This post is partly a belated response to Joshua Achiam, currently OpenAI's Head of Mission Alignment: If we adopt safety best practices that are common in other professional engineering fields, we'll get there … I consider myself one of the x-risk people, though I agree that most of them would reject my view on how to prevent it. I think the wholesale rejection of safety best practices from other fields is one of the dumbest mistakes that a group of otherwise very smart people has ever made

en-gb

2:39

“Differentially Scary Movies” by jefftk

LessWrong (30+ Karma) ·

2026/02/02

This post has spoilers for My Neighbor Totoro, Frozen, Bambi, andthe Lion King People at different stages of development enjoy different things in movies. Some of the best children's movies are able to make things scary or intense for the adults without being too much for little kids. For example, in My Neighbor Totoro, there's a place where everyone is worried that a small child may have fallen in the lake. She's gone missing, they find a sandal floating in the pond, you see

en-gb

4,474 results

LessWrong (30+ Karma)

All Episodes

Similar Podcasts

ev.news

Martyn Lee

TechCrunch Industry News

TechCrunch

Elon Musk Podcast

Stage Zero

Reflections

Livewire Labs

TechCrunch Startup News

TechCrunch