PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

18:24

“A summary of Condensation and its relation to...

LessWrong (30+ Karma) ·

2026/03/05

Short summary of Condensation Condensation is a theory of concepts by Sam Eisenstat. The paper can be read here. Abram wrote a review on Lesswrong, and a followup. The paper is very much worth reading, and can be skimmed to just understand the motivation if you’re time constrained. What is a concept? One rough pointer is "the little packets of information that are the meaning of a word or phrase". Another is the clusters found in the cluster structure of thingspace. Maybe we could think of them as

en-gb

15:24

“Maybe there’s a pattern here?” by dynomight

LessWrong (30+ Karma) ·

2026/03/04

1. It occurred to me that if I could invent a machine—a gun—which could by its rapidity of fire, enable one man to do as much battle duty as a hundred, that it would, to a large extent supersede the necessity of large armies, and consequently, exposure to battle and disease [would] be greatly diminished. Richard Gatling (1861) 2. In 1923, Hermann Oberth published The Rocket to Planetary Spaces, later expanded as Ways to Space Travel. This showed that it was possible to build machines that could

en-gb

8:31

“Is GDP a Kind of Factory?d Draft” by Benquo

LessWrong (30+ Karma) ·

2026/03/04

In 2021, economists Arvind Subramanian, Justin Sandefur, and Dev Patel announced that poor countries had finally started catching up to rich ones, vindicating the Solow growth model's prediction of "convergence." Now they say the rumors of Solow convergence's vindication were premature; the convergence trend reversed and poor countries are falling behind again. David Oks, writing about their retraction, argues that the whole convergence episode was a mirage produced by the Chinese commodities boom

en-gb

11:04

“Sacred values of future AIs” by Cleo Nardo

LessWrong (30+ Karma) ·

2026/03/04

Consider a future with many diverse AIs that need to coordinate with each other, or at least coexist without conflict. Such AIs would need shared values they can coordinate around. According to Hanson's theory, groups of diverse agents facing coordination pressure will tend to sacralize some shared value — seeing it in “far mode” so they can see it together. Unfortunately, this makes them systematically worse at making decisions about these things. This model suggests three claims: Helpfulness,

en-gb

9:25

“Mass surveillance, red lines, and a crazy weekend”...

LessWrong (30+ Karma) ·

2026/03/04

[These are my own opinions, and not representing OpenAI. Cross-posted on windowsontheory.] AI has so many applications, and AI companies have limited resources and attention span. Hence if it was up to me, I’d prefer we focus on applications that are purely beneficial— science, healthcare, education — or even commercial, before working on anything related to weapons or spying. If someone has to do it, I’d prefer it not to be my own company. Alas, we can’t always get what we want. This is a long-ish

en-gb

17:16

“Physics of RL: Toy scaling laws for the emergence of...

LessWrong (30+ Karma) ·

2026/03/04

TL;DR: When is or isn't reward the optimization target? I use a mathematical toy model to reason about when RL should select for reward-seeking reasoning as opposed to behaviors that achieve high reward without thinking about reward. Hypothesis: More diverse RL increases the likelihood that reward-seeking emerges. In my toy model, there are scaling laws for the emergence of reward-seeking: when decreasing prior probability of reward-seeking reasoning we can increase RL diversity enough that

en-gb

14:28

“OpenAI’s surveillance language has many potential...

LessWrong (30+ Karma) ·

2026/03/04

(The author is not affiliated with the Department of War or any major AI company.) There's a lot of disagreement about the new surveillance language in the OpenAI–Department of War agreement. Some people think it's a significant improvement over the previous language.[1] Others think it patches some issues but still leaves enough loopholes to not make a material difference. Reasonable people disagree about how a court will interpret the language, if push comes to shove. But here's something that

en-gb

3:50

“Mass Surveillance w/ LLMs is the Default Outcome....

LessWrong (30+ Karma) ·

2026/03/04

What's the best case scenario regarding OpenAI's contract w/ the Department of War (DoW)? We have access to the full contract It's airtight OAI's engineers are on top of things in case the DoW breaks the contract There's actual teeth for violations But even then, the DoW can simply switch vendors. Use Gemini. Use Grok. If these models aren't capable enough, then just wait a year. In the past[1], the DoW has purchased commercial location data, on Americans, w/o warrants. In recent negotiations,[2]

en-gb

32:28

“A Tale of Three Contracts” by Zvi

LessWrong (30+ Karma) ·

2026/03/04

The attempt on Friday by Secretary of War Pete Hegsted to label Anthropic as a supply chain risk and commit corporate murder had a variety of motivations.On its face, the conflict is a tale of three contracts and the associated working relationships. The contract Anthropic signed with the Department of War (DoW) in 2025. The new contract Anthropic was negotiating with DoW, that would have been modified to favor DoW, but where the parties could not reach agreement. The contract OpenAI was

en-gb

4:38

“I’m confused by the change in the METR trend” by...

LessWrong (30+ Karma) ·

2026/03/04

Measuring AI Ability to Complete Long Tasks - METR In their original 2025 paper, METR noticed that the slope (aka task horizon doubling time) of the trendline for models released in 2024 and later is different from the slope for <2024 models. First, I decided to check whether a piecewise linear function fits the data better than a simple linear function. If it doesn't, then this change in trend is a random fluke and there is nothing worth talking about. Here is the data so far (SOTA models

en-gb

4:23

“I Had Claude Read Every AI Safety Paper Since 2020”...

LessWrong (30+ Karma) ·

2026/03/04

Click here if you just want to see the Database I made of all[1] AI safety papers written since 2020 and not read the methodology. Over the last month I have been trying to see just how much I can learn and do from a cold start[2] in the world of AI safety. A large part of this has been frantically learning mech interp, but I've picked up two projects that I think are worth sharing! There are a lot of AI Safety papers. When I started working on more hands on projects there wasn't a clear way to

en-gb

19:37

“Game Recognizes Game” by eva_

LessWrong (30+ Karma) ·

2026/03/03

Epistemic Status: Further Research Needed, would be a shorter essay if I thought about it for longer. Conspiracies exist. Some of them are quite large, involve people who met as adults and agreed to do crimes together, and do many heinous things that multiple co-conspirators know about for years without the police noticing. Some organised crime groups start within families. If you were all raised in the same family-first culture and they're your cousin, you can be pretty sure they're not a cop

en-gb

13:01

“An Alignment Journal: Coming Soon” by Dan MacKinlay,...

LessWrong (30+ Karma) ·

2026/03/03

tl;dr We’re incubating an academic journal for AI alignment: rapid peer-review of foundational Alignment research that the current publication ecosystem underserves. Key bets: paid attributed review, reviewer-written synthesis abstracts, and targeted automation. Contact us if you’re interested in participating as an author, reviewer, or editor, or if you know someone who might be. Experimental Infrastructure for Foundational Alignment Research This is the first in a series of “build-in-the-open”

en-gb

34:30

“Current activation oracles are hard to use” by...

LessWrong (30+ Karma) ·

2026/03/03

This work was conducted during the MATS 9.0 program under Neel Nanda and Senthooran Rajamanoharan. tldr; Activation oracles (Karvonen et al.) are a recent technique where a model is finetuned to answer natural language questions about another model's activations. They showed some promising signs of generalising to tasks fairly different to what they were trained for so we decided to try them on a bunch of safety relevant tasks. Overall, activation oracles showed some signal on tasks close to

en-gb

19:05

“In-context learning of representations can be...

LessWrong (30+ Karma) ·

2026/03/03

This is a crosspost of my ICLR 2026 blogpost track post. All code and experiments are available at github.com/andyrdt/iclr_induction. Summary Park et al., 2025 show that when large language models (LLMs) process random walks on a graph, their internal representations come to mirror the underlying graph's structure. The authors interpret this broadly, suggesting that LLMs can "manipulate their representations in order to reflect concept semantics specified entirely in-context". In this post, we take

en-gb

3:30

“Epstein and my world model” by Eye You

LessWrong (30+ Karma) ·

2026/03/02

Have you guys heard about this Epstein stuff? Shit's pretty crazy. Note: I'm not going to provide a summary of the situation or talk about evidence; this piece is for people that already know these things. I'm going to avoid specifics about what Epstein and co did, and instead will use vague terms like "Epstein stuff". This is a short post about how I've updated my world model. Particular things that I find very surprising: that so many people basically knew what was going on and didn't say

en-gb

172:45

“Secretary of War Tweets That Anthropic is Now a...

LessWrong (30+ Karma) ·

2026/03/02

This is the long version of what happened so far. I will strive for shorter ones later, when I have the time to write them. Most of you should read the first two sections, then choose the remaining sections that are relevant to your interests. But first, seriously, read Dean Ball's post Clawed. Do that first. I will not quote too extensively from it, because I am telling all of you to read it. Now. You’re not allowed to keep reading this or anything else until after you do. I’m not kidding.

en-gb

5:29

“War Claude” by PeterMcCluskey

LessWrong (30+ Karma) ·

2026/03/02

What a weekend. Two new wars in Asia don't qualify as top news. My first reaction to Hegseth's conflict with Anthropic was along the lines of: I expected an attempt at quasi-nationalization of AI, but not this soon. And I expected it to look like it was managed by national security professionals. Hegseth doesn't look like he's trying to avoid the role of cartoon villain. On closer inspection, it doesn't look very much like nationalization. A significant part of what's going on is bribery

en-gb

25:04

“How to Design Environments for Understanding Model...

LessWrong (30+ Karma) ·

2026/03/02

Authors: Gerson Kroiz*, Aditya Singh*, Senthooran Rajamanoharan, Neel Nanda Gerson and Aditya are co-first authors. This work was conducted during MATS 9.0 and was advised by Senthooran Rajamanoharan and Neel Nanda. TL;DR Understanding why a model took an action is a key question in AI Safety. It is a difficult question, as often many motivations could plausibly lead to the same action. We are particularly concerned about the case of model incrimination, i.e. observing a model taking a bad action

en-gb

16:48

“Tools to generate realistic prompts help...

LessWrong (30+ Karma) ·

2026/03/02

TLDR We train and many-shot prompt base models to generate user prompts that are harder to distinguish from deployment (WildChat) prompts. Then we give Petri, an automated auditing agent, a tool to use a prompt generator model for sycophancy audits. It doesn’t help with making the full audit transcripts harder to distinguish from deployment transcripts, despite more realistic single-turn prompts. We find evidence that full audit realism is bottlenecked not by prompt quality, but by higher-level

en-gb

3,860 results

LessWrong (30+ Karma)

All Episodes

Similar Podcasts

Lead With AI

Tamara Nall

Linux in the Ham Shack

Black Sparrow Media

Data Engineering Podcast

Tobias Macey

Eye On A.I.

Craig S. Smith

Reflections

Livewire Labs