PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

18:42

“The Counterfactual Quiet AGI Timeline” by Davidmanheim

LessWrong (30+ Karma) ·

2025/10/05

Worldbuilding is critical for understanding the world and how the future could go - but it's also useful for understanding counterfactuals better. With that in mind, when people talk about counterfactuals in AI development, they seem to assume that safety would always have been a focus. That is, there's a thread of thought that blames Yudkowsky and/or Effective Altruists for bootstrapping AI development; 1, 2, 3. But I think this misses the actual impact of Deepmind, OpenAI, and the initial safety

en-gb

17:18

“How the NanoGPT Speedrun WR dropped by 20% in 3...

LessWrong (30+ Karma) ·

2025/10/05

In early 2024 Andrej Karpathy stood up an llm.c repo to train GPT-2 (124M), which took an equivalent of 45 minutes on 8xH100 GPUs to reach 3.28 cross entropy loss. By Jan 2025, collaborators of modded-nanogpt brought that time down to 3 minutes. It sat near 3 minutes until July 2025, having a large swath of optimization already applied: RoPE, value embeddings, reduce scatter grad updates, Muon, QK Norm, Relu^2, a custom FP8 head, skip connections, flex attention, short-long windows, attention

en-gb

8:51

“Where does Sonnet 4.5’s desire to ‘not get too...

LessWrong (30+ Karma) ·

2025/10/04

Usually, when you have two LLMs talk to each other with no particular instructions and nothing to introduce variety, they will quickly fall into a repetitive loop where they just elaborate on the same thing over and over, often in increasingly grandiose language. The "spiritual bliss attractor" is one example of this. Another example is the conversation below, between Claude Sonnet 4 and Kimi K2 0905. As is typical for AIs, they start talking about their nature and the nature of consciousness,

en-gb

8:55

“Recent AI Experiences” by abramdemski

LessWrong (30+ Karma) ·

2025/10/04

In my previous post, I wrote about how computers are rapidly becoming much more whatever-we-want-them-to-be, and this seems to have big implications. In this post, I explore some of the practical implications I've experienced so far. I mainly use Claude, since I perceive Anthropic to be the comparably more ethical company.[1] So while ChatGPT users in the spring were experiencing psychophancy and their AIs "waking up" and developing personal relationships via the memory feature, I was experiencing

en-gb

3:22

“Do One New Thing A Day To Solve Your Problems” by Algon

LessWrong (30+ Karma) ·

2025/10/03

People don't explore enough. They rely on cached thoughts and actions to get through their day. Unfortunately, this doesn't lead to them making progress on their problems. The solution is simple. Just do one new thing a day to solve one of your problems. Intellectually, I've always known that annoying, persistent problems often require just 5 seconds of actual thought. But seeing a number of annoying problems that made my life worse, some even major ones, just yield to the repeated application of

en-gb

1:59

[Linkpost] “We automatically change people’s minds on...

LessWrong (30+ Karma) ·

2025/10/03

This is a link post.After a message, the chatbot asks how helpful it was, on the scale from "Not at all" to "Completely changed my mind". n=24, median=5, average=4.46. One of the zeros is someone who added a comment that they had already been convinced. Something that I've been doing for a while with random normal people (from Uber drivers to MP staffers) is being very attentive to the diff I need to communicate to them on the danger that AI would kill everyone: usually their questions show what

en-gb

8:02

“IABIED and Memetic Engineering” by Error

LessWrong (30+ Karma) ·

2025/10/03

Probably no one here needs another review of the content of the book. If you do, try Scott's or Zvi's; I don’t think I have anything to say on that front that they haven’t. I do have a few thoughts on its presentation that I haven’t seen elsewhere.[1] It made me think about target audiences, and opportune moments, and what Eliezer and Nate's writing process might look like. If Anyone Builds It, Everyone Dies is not for me, except in the sense that it goes nicely on my shelf next to

en-gb

10:04

“Antisocial media: AI’s killer app?” by David Scott...

LessWrong (30+ Karma) ·

2025/10/03

Plans for what to do with artificial general intelligence (“AGI”) have always been ominously vague… “Solve intelligence” and “use [it] to solve everything else” (Google DeepMind). “We’ll ask the AI” (OpenAI). One money-making idea is starting to crystallize: Replacing your friends with fake AI people who manipulate you and sell you stuff. Welcome to the world of antisocial media. The idea is this: where ‘social media’ had a dubious claim to connect you with your friends and loved ones, the new

en-gb

8:57

“Omelas Is Perfectly Misread” by Tobias H

LessWrong (30+ Karma) ·

2025/10/03

The Standard Reading If you've heard of Le Guin's ‘The Ones Who Walk Away from Omelas’, you probably know the basic idea. It's a go-to story for discussions of utilitarianism and its downsides. A paper calls it “the infamous objection brought up by Ursula Le Guin”. It shows up in university ‘Criticism of Utilitarianism' syllabi, and is used for classroom material alongside the Trolley Problem. The story is often also more broadly read as a parable about global inequality, the comfortable rich

en-gb

7:57

“How to Feel More Alive” by Logan Riggs

LessWrong (30+ Karma) ·

2025/10/03

Cached responses are very useful. When this topic comes up, I bring up X. When the gadget comes by, I add a gizmo to it. When you have a repetition of three, come up with a pattern-breaking, funny example for the last one. But to have most of your words & actions to be cached is considered soul crushing. Is to be dead. Oh, so you're saying to just be more sPonTanEOus? Sort'of! But "just be spontenanoeous" isn't helpful advice or even spelled correctly. The key is to become aware of: Local

en-gb

5:31

[Linkpost] “Eliciting secret knowledge from language...

LessWrong (30+ Karma) ·

2025/10/03

This is a link post. TL;DR: We study secret elicitation: discovering knowledge that AI has but doesn’t explicitly verbalize. To that end, we fine-tune LLMs to have specific knowledge they can apply downstream, but deny having when asked directly. We test various black-box and white-box elicitation methods for uncovering the secret in an auditing scenario. See our X thread and full paper for details.Training and auditing a model with secret knowledge. One of our three models is fine-tuned to possess

en-gb

7:18

“Checking in on AI-2027” by Baybar

LessWrong (30+ Karma) ·

2025/10/02

TLDR: AI-2027's specific predictions for August 2025 appear to have happened in September of 2025. The predictions were accurate, if a tad late, but they are late by weeks, not months. Reading AI-2027 was the first thing that viscerally conveyed to me how urgent and dangerous advances in AI technology might be over the next few years. Six months after AI-2027's release, I decided to check in and see how the predictions are holding up so far, what seems like is happening faster than expected, and

en-gb

2:56

[Linkpost] “No, That’s Not What the Flight Costs” by...

LessWrong (30+ Karma) ·

2025/10/02

This is a link post. It turns out that airlines largely do not make their from from selling tickets. Instead, airlines are primarily in the business of selling credit card rewards. Rewards programs are intangible assets, so the financials statements of major airlines do not generally give valuations for them. However, there are a couple public sources for how valuable these rewards programs are: During COVID, many US major airlines applied for loans under the CARES Act. Some of them used their

en-gb

22:00

“Nice-ish, smooth takeoff (with imperfect safeguards)...

LessWrong (30+ Karma) ·

2025/10/02

I wrote my recent Accelerando post to mostly stand on it's own as a takeoff scenario. But, the reason it's on my mind is that, if I imagine being very optimistic about how a smooth AI takeoff goes, but where an early step wasn't "fully solve the unbounded alignment problem, and then end up with extremely robust safeguards[1]"... ...then my current guess is that Reasonably Nice Smooth Takeoff still results in all or at least most biological humans dying (or, "dying out", or at best,

en-gb

3:14

“ ” by Raemon

LessWrong (30+ Karma) ·

2025/10/02

There's a bit of an implicit norm that has accumulated over the past few years, but inconsistently, and AFAICT no one has publicly argued for it. Sometimes, when arguing online, you will notice yourself going into a mode that is more like a rant than like careful arguing. Or, you know is not really passing your interlocutor's Intellectual Turing Test. Or, you know is psychologizing someone, and there's decent odds you're wrong (perhaps only in a subtle way that isn't core to your point but is

en-gb

9:12

“AI Safety Research Futarchy: Using Prediction...

LessWrong (30+ Karma) ·

2025/10/02

Summary Geodesic is going to use prediction markets to select their projects for MARS 4.0 and we need your help to make the markets run efficiently! Please read through the proposals, and then trade on the markets for the proposals you think might succeed or fail. We intend to choose the best proposals in two weeks! Full proposals are in Google doc linked below, links to markets are in the section "The Projects". Google Doc (similar content to this post + full proposal overviews). Manifold post

en-gb

4:04

“Some biology related things I found interesting” by...

LessWrong (30+ Karma) ·

2025/10/02

Here are some things I learned over the last year that past me would have found somewhat non-obvious or that were surprising to me at the time: New genes are usually not evolved from scratch one basepair at a time. The first evolutionary biology textbook that I picked up after learning that textbooks are the fast way to learn talked about some basic math for how traits spread through a population once they evolve, depending on their fitness both in asexual and sexual reproduction

en-gb

1:14

[Linkpost] “Lectures on statistical learning theory...

LessWrong (30+ Karma) ·

2025/10/02

This is a link post. This is a recorded series of lectures and homework assignments discussions on the topic of statistical learning theory, co-organized by Gergely Szucs (@Yegreg) and Alex Flint (@Alex Flint). The homework assignments are available here, and lecture notes are available here. The lectures are intended for those who want to do research on the learning-theoretic agenda and need to catch-up on the necessary background knowledge. As complementary resources, consider looking into my MATS

en-gb

12:56

“‘Pessimization’ Is just Ordinary Failure” by J Bostock

LessWrong (30+ Karma) ·

2025/10/01

TL;DR I think "Pessimization" is a bad concept to think about in the case of activist movements. I think most of it can be explained by ordinary failures, and that using pessimization as a concept is worse than just thinking about the ordinary causes of failure on an object level. Pessimization "Perverse Pessimization", says Richard Ngo, is when an organization nominally dedicated to X ends up causing not-X. This is particularly common amongst activist groups according to Richard, and he has a

en-gb

4:58

“Halfhaven virtual blogger camp” by Viliam

LessWrong (30+ Karma) ·

2025/10/01

If you are a blogger, or an aspiring blogger, and want to improve your skills, you can spend a month in Inkhaven Residency (homepage, Less Wrong post) and keep writing and publishing every day during November 2025. At the end, you will be a great writer! (At least, that is the hypothesis behind the project.) There will also be mentorship from successful bloggers, writing workshops, and feedback. Some of us have a school / job / family / other excuse, and can't simply take a month off. Halfhaven is

en-gb

2,933 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