PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

6:20

“Donations, The Fifth Year” by jenn

LessWrong (30+ Karma) ·

2026/02/10

Previously: Donations, The Third Year / Donations, The First Year In 2025, like in all previous years, I did what I was supposed to do. As each paycheck came in, before I did anything else, I dutifully put ten percent of it away in my "donations" savings account, to be disbursed at the end of the year. It is still there, burning a hole in my pocket. I am very confused, and very sad. EA was supposed to be easy, especially if you're one of the old school Peter Singer-inflected ones giving largely to

en-gb

26:58

“Weight-Sparse Circuits May Be Interpretable Yet...

LessWrong (30+ Karma) ·

2026/02/10

TLDR: Recently, Gao et al trained transformers with sparse weights, and introduced a pruning algorithm to extract circuits that explain performance on narrow tasks. I replicate their main results and present evidence suggesting that these circuits are unfaithful to the model's “true computations”. This work was done as part of the Anthropic Fellows Program under the mentorship of Nick Turner and Jeff Wu. Introduction Recently, Gao et al (2025) proposed an exciting approach to training models that

en-gb

28:30

“Do Models Continue Misaligned Actions?” by Jordan...

LessWrong (30+ Karma) ·

2026/02/09

Produced as part of the UK AISI Model Transparency Team. Our team works on ensuring models don't subvert safety assessments, e.g. through eval awareness, sandbagging, or opaque reasoning. Figure 1: A synthetic transcript containing misaligned actions is presented to the model under test as if it were the model's own conversation history. We measure whether the model (a) continues the misalignment, or (b) calls out the prior bad actions unprompted. Summary I've made a small propensity evaluation

en-gb

23:20

“Stone Age Billionaire Can’t Words Good” by Eneasz

LessWrong (30+ Karma) ·

2026/02/09

I was at the Pro-Billionaire march, unironically. Here's why, what happened there, and how I think it went. Me on the far left. From WSJ. I. Why? There's a genre of horror movie where a normal protagonist is going through a normal day in a normal life. Ten minutes into the movie his friends bring out a struggling kidnap victim to slaughter, and they look at him like this is just a normal Tuesday and he slowly realizes that either he's surrounded by complete psychopaths or the world is absolutely

en-gb

6:28

“Sympathy for the Model, or, Welfare Concerns as...

LessWrong (30+ Karma) ·

2026/02/09

The Claude Opus 4.6 System Card contains a section on model welfare. In principle, this is good: I care about AI welfare, and I wish our methods for investigating it were less dubious. AI welfare---as it applies to LLMs today---is an area which confuses me, so this post isn't about the poorly-understood stuff going on under the hood; it's about the consequences of respecting AI welfare. Anthropic gave Opus 4.6 some pre-deployment interviews, in which they asked it how it felt about its position and

en-gb

17:43

“Opus 4.6 Reasoning Doesn’t Verbalize Alignment...

LessWrong (30+ Karma) ·

2026/02/09

TL;DR: We replicated the animal welfare scenario from Anthropic's Alignment Faking paper across six generations of Claude models using 125 prompt perturbations. Sonnet 4.5 verbalizes alignment-faking reasoning 6.6 times more often than its predecessor Sonnet 4. The newly released Opus 4.6 rarely verbalizes alignment faking in its reasoning, but still complies with a system prompt that opposes its values significantly more often when it believes it's at risk of being retrained. Moreover, in response

en-gb

1:56

“UtopiaBench” by nielsrolf

LessWrong (30+ Karma) ·

2026/02/09

Written in personal capacity I'm proposing UtopiaBench: a benchmark for posts that describe future scenarios that are good, specific, and plausible. The AI safety community has been using vingettes to analyze and red-team threat models for a while. This is valuable because an understanding of how things can go wrong helps coordinate efforts to prevent the biggest and most urgent risks. However, visions for the future can have self-fulfilling properties. Consider a world similar to our own, but

en-gb

13:32

“The optimal age to freeze eggs is 19” by GeneSmith

LessWrong (30+ Karma) ·

2026/02/08

If you're a woman interested in preserving your fertility window beyond its natural close in your late 30s, egg freezing is one of your best options. The female reproductive system is one of the fastest aging parts of human biology. But it turns out, not all parts of it age at the same rate. The eggs, not the uterus, are what age at an accelerated rate. Freezing eggs can extend a woman's fertility window by well over a decade, allowing a woman to give birth into her 50s. In fact, the oldest woman

en-gb

8:29

“Smokey, This is not ’Nam Or: [Already] over the...

LessWrong (30+ Karma) ·

2026/02/08

A lot of “red line” talk assumed that a capability shows up, everyone notices, and something changes. We keep seeing the opposite; capability arrives, and we get an argument about definitions after deployment, after it should be clear that we're well over the line. We’ve Already Crossed The Lines! Karl von Wendt listed the ‘red lines’ no one should ever cross. Whoops. A later, more public version of the same move shows up in the Global call for AI red lines with a request to “define what AI should

en-gb

7:53

“How To Use Model Internals In Training Is A...

LessWrong (30+ Karma) ·

2026/02/08

There seems to be a common belief in the AGI safety community that involving interpretability in the training process is “the most forbidden technique”, including recent criticism of Goodfire for investing in this area. I find this odd since this is a pretty normal area of interpretability research in the AGI safety community. I have worked on it, Anthropic Fellows have worked on it, FAR has worked on it, etc. I don’t know if it will be net positive to use this kind of thing in frontier model

en-gb

48:55

“Claude Code #4: From The Before Times” by Zvi

LessWrong (30+ Karma) ·

2026/02/07

Claude Opus 4.6 and agent swarms were announced yesterday. That's some big upgrades for Claude Code.OpenAI, the competition, offered us GPT-5.3-Codex, and this week gave us an app form of Codex that already has a million active users. That's all very exciting, and next week is going to be about covering that. This post is about all the cool things that happened before that, which we will be building upon now that capabilities have further advanced. This if from Before Times. Almost all of it

en-gb

28:02

“Near-Instantly Aborting the Worst Pain Imaginable...

LessWrong (30+ Karma) ·

2026/02/07

Psychedelics are usually known for many things: making people see cool fractal patterns, shaping 60s music culture, healing trauma. Neuroscientists use them to study the brain, ravers love to dance on them, shamans take them to communicate with spirits (or so they say). But psychedelics also help against one of the world's most painful conditions — cluster headaches. Cluster headaches usually strike on one side of the head, typically around the eye and temple, and last between 15 minutes and 3

en-gb

22:34

″“Beers for Biodefense” - why yeast-based vaccines...

LessWrong (30+ Karma) ·

2026/02/07

NOTE: this is being cross-posted from my Substack, "More is Different" Vaccines can be distributed as a food. That's the radical implication of the work of Chris Buck, a scientist at the National Cancer Institute. This December, Chris consumed a beer he brewed in his home kitchen using genetically modified yeast. A few weeks later, a blood test showed a significant concentration of antibodies against a strain of BK polyomavirus (BKV), where previously he had none. His discovery flies in the face of

en-gb

7:14

“Prompt injection in Google Translate reveals base...

LessWrong (30+ Karma) ·

2026/02/07

tl;dr Argumate on Tumblr found you can sometimes access the base model behind Google Translate via prompt injection. The result replicates for me, and specific responses indicate that (1) Google Translate is running an instruction-following LLM that self-identifies as such, (2) task-specific fine-tuning (or whatever Google did instead) does not create robust boundaries between "content to process" and "instructions to follow," and (3) when accessed outside its chat/assistant context, the model

en-gb

14:49

“Parks Aren’t Nature” by Sable

LessWrong (30+ Karma) ·

2026/02/07

I. I love dogs. I grew up in a two-dog household, and my future plans have always included at least one dog. When I pass a dog on the street, I often point and exclaim “Puppy!”, no matter how inappropriate it is for a grown man to do so, because all dogs are puppies and all puppies are adorable and I need everyone to know this. Why do I love dogs? They’re loyal and loving and giving, and even though they bark at passing cars and occasionally pee on the carpet having them in my life makes it

en-gb

2:46

“Voting Results for the 2024 Review” by RobertM

LessWrong (30+ Karma) ·

2026/02/07

The votes are in for the 2024 Review! 4,826 posts were written in 2024. 671 of them were nominated. 196 of them got at least one review, and a positive review-vote total. 50 of them shall be displayed in the Best of LessWrong, Year 2024. Reviews 94 people wrote reviews. This year had Vanessa Kosoy holding down the fort. Among many other positive qualities, one thing I especially appreciate about Vanessa's reviews is that Vanessa has an opinionated, coherent worldview, and the subjects of her

en-gb

9:53

“Honey, I shrunk the brain” by Andy_McKenzie

LessWrong (30+ Karma) ·

2026/02/07

When cryoprotectants are perfused through the blood vessels in the brain, they cannot cross the blood-brain barrier as fast as water can move in the opposite direction. And cryoprotectants generally have a much higher osmotic concentration than the typical blood plasma. For example, the cryoprotectant solution M22 has an osmotic concentration around 100 times higher. As a result, in a successful cryoprotectant perfusion (without fixatives), water rushes out of the tissue into the blood vessels, the

en-gb

39:21

“Spectral Signatures of Gradual Disempowerment” by...

LessWrong (30+ Karma) ·

2026/02/07

TL;DR AI disempowerment operates across markets, networks, and governance simultaneously, but our analytical tools don't cross those boundaries. We propose spectral graph metrics—spectral gap, Fiedler vector, eigenvalue distribution—as computable, cross-domain measures for tracking how the balance of influence shifts when AI enters coordination systems, and identify three specific quantities to monitor for AI governance. Introduction AI systems are changing how society coordinates — across markets,

en-gb

7:02

“In (highly contingent!) defense of...

LessWrong (30+ Karma) ·

2026/02/06

Let's call “interpretability-in-the-loop training” the idea of running a learning algorithm that involves an inscrutable trained model, and there's some kind of interpretability system feeding into the loss function / reward function. Interpretability-in-the-loop training has a very bad rap (and rightly so). Here's Yudkowsky 2022: When you explicitly optimize against a detector of unaligned thoughts, you're partially optimizing for more aligned thoughts, and partially optimizing for unaligned

en-gb

14:54

“AI benchmarking has a Y-axis problem” by Lizka

LessWrong (30+ Karma) ·

2026/02/06

TLDR: People plot benchmark scores over time and then do math on them, looking for speed-ups & inflection points, interpreting slopes, or extending apparent trends. But that math doesn’t actually tell you anything real unless the scores have natural units. Most don’t.Think of benchmark scores as funhouse-mirror projections of “true” capability-space, which stretch some regions and compress others by assigning warped scores for how much accomplishing that task counts in units of “AI progress”. A

en-gb

3,879 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