PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

10:55

“Slack in Cells, Slack in Brains” by Mateusz Bagiński

LessWrong (30+ Karma) ·

2026/03/31

[A veridically metaphorical explanation of why you shouldn't naïvely cram your life with local optimizations (even for noble or altruistic reasons).] TL;DR: You need Slack to be an effective agent. Slack is fragile, and it is tempting to myopically sacrifice it, and myopic sacrifice makes future myopic sacrificing more likely. Learn not to do this and cultivate slack. Slack in Cells The smallest living mammal is the Etruscan shrew, weighing about 1.8g ("as much as a paperclip"), and ~4cm in length

en-gb

4:06

“I am definitely missing the pre-AI writing era” by...

LessWrong (30+ Karma) ·

2026/03/31

Yesterday, I wrote my first technical draft on what I was working on with the goal to share it publicly on here (well using an account dedicated to technical post), and did not realize how wanting to sound perfect actually steal the ''my voice" in the paper. Although 80 % of the content was my own writing, the fact that it was run in a LLM enginee for grammar and vocabulary cross-check, made it failed the "probable written by AI " metric; and it was rejected. I am not complaining, well, it was

en-gb

4:29

“The state of AI safety in four fake graphs” by Boaz...

LessWrong (30+ Karma) ·

2026/03/30

Here is a quick overview of my intuitions on where we are with AI safety in early 2026: So far, we continue to see exponential improvements in capabilities. This is most visible in the famous “METR graph”, but the trend is clear in many other metrics, including revenue. If you squint, you can even see a potential recent “bending upward” of the curve, as we are starting to use AI to accelerate the development of AI. We see some good news in alignment - as models become more capable, they are also

en-gb

33:45

“AI should be a good citizen, not just a good...

LessWrong (30+ Karma) ·

2026/03/30

Introduction Consider a lorry driver who sees a car crash and pulls over to help, even though it’ll delay his journey. Or a delivery driver who notices that an elderly resident hasn’t collected their post in days, and knocks to check they’re okay. Or a social media company employee who notices how their platform is used for online bullying, and brings it up with leadership, even though that's not part of their job description. This kind of proactive prosocial behaviour is admirable in humans

en-gb

46:07

″(Some) Natural Emergent Misalignment from Reward...

LessWrong (30+ Karma) ·

2026/03/30

Authors: Satvik Golechha*, Sid Black*, Joseph Bloom * Equal Contribution. This work was done as part of the Model Transparency team at the UK AI Security Institute (AISI). ExecutiveSummary In Natural Emergent Misalignment from Reward Hacking in Production RL (MacDiarmid et al., 2025), Anthropic recently demonstrated that language models that learn reward hacking in their production RL environments become emergently misaligned (EM). Their pipeline, illustrated below, proceeds from pre-training

en-gb

2:54

[Linkpost] “Parkinson’s Law of Worry” by Jakub Halmeš

LessWrong (30+ Karma) ·

2026/03/29

This is a link post. Parkinson's law says that "Work expands so as to fill the time available for its completion." I think that a similar observation can be made for people worrying about stuff. To paraphrase: "Problem salience expands so as to fill the capacity available for worrying.” Suppose a person is worried about several problems. Let's visualize the mental state of this person, where each problem is represented by a colored circle, and the size of the circle corresponds to how much this

en-gb

36:43

“Folie à Machine: LLMs and Epistemic Capture” by...

LessWrong (30+ Karma) ·

2026/03/29

"Truly, whoever can make you believe absurdities can make you commit atrocities." — Voltaire, 1765 A man in his late forties discovers a new passion. After decades of working as a mid level manager for a restaurant chain, he starts spending his evenings deep in study, filling notebooks with diagrams, reading everything he can get his hands on. Within a few weeks he's convinced he's on the verge of a grand unified theory that reconciles quantum mechanics and general relativity. He starts emailing

en-gb

2:50

“Stop asking “how good is this” to decide between...

LessWrong (30+ Karma) ·

2026/03/29

Prospective donors often ask how cost-effective my team's various recommendations are — they want to donate to the most cost-effective opportunity. But the answer is often roughly: If we do our job right, then all donations we recommend to you are equally valuable on the margin, and that value is our bar. If an opportunity is substantially above our bar, we will make sure to fill it with or without you; you're not counterfactual. (Alternative framing: we will just recommend more funding for that

en-gb

5:48

“Nick Bostrom: How big is the cosmic endowment?” by...

LessWrong (30+ Karma) ·

2026/03/28

Superintelligence, pp. 122–3. 2014. Consider a technologically mature civilization capable of building sophisticated von Neumann probes of the kind discussed in the text. If these can travel at 50% of the speed of light, they can reach some stars before the cosmic expansion puts further acquisitions forever out of reach. At 99% of c, they could reach some stars. These travel speeds are energetically attainable using a small fraction of the resources available in the solar system. The

en-gb

6:35

“Don’t Overdose Locally Beneficial Changes” by...

LessWrong (30+ Karma) ·

2026/03/28

If you are very overweight, then you should aim to cut down your daily caloric intake. This doesn't mean the optimal daily caloric intake is zero. If you are very underweight, then you should aim to ramp up your daily caloric intake. This doesn't mean the optimal daily caloric intake is infinity. In general, if something is good to do some amount in some context, this doesn't mean that you should go as all-in on it as you can possibly manage. The utility of a change is context-dependent, and as you

en-gb

5:42

“Stanley Milgram wasn’t pessimistic enough about...

LessWrong (30+ Karma) ·

2026/03/28

A landmark of social psychology research was “The Milgram Experiment,” but a new look at the audio tapes and other evidence collected during that experiment suggests that we may have been interpreting it incorrectly. Here is the Wikipedia summary of the experiment, showing how it is typically portrayed: Yale University psychologist Stanley Milgram… intended to measure the willingness of study participants to obey an authority figure who instructed them to perform acts conflicting with their

en-gb

4:55

[Linkpost] “What if superintelligence is just weak?”...

LessWrong (30+ Karma) ·

2026/03/28

This is a link post. In response to “2023 Or, Why I am Not a Doomer” by Dean W. Ball. Dean Ball is a pretty big voice in AI policy – over 19k subscribers on his newsletter, and a former Senior Policy Advisor for AI at the Trump White House – so why does he disagree that AI poses an existential danger to humanity? In short, he holds the common view that superintelligence (ASI) simply won’t be that powerful. I strongly disagree, and I think he makes a couple of invalid leaps to arrive there. Better

en-gb

9:57

“Pray for Casanova” by Tomás B.

LessWrong (30+ Karma) ·

2026/03/28

I am fascinated by the beautiful who become deformed. Some become bitter, more bitter than those born less pulchritudinous. Most learn to cope with the loss. Some were blind to how much their beauty helped them, the halo of their hotness an invisible bumper softening life. But most cultivated this aspect to some degree. They knew what was up. But none were fully prepared for the anti-halo: the revulsion, the active disgust. They became monsters. This is what it means to be marred. In 1715 England,

en-gb

8:35

“ControlAI 2025 Impact Report” by Andrea_Miotti, Alex...

LessWrong (30+ Karma) ·

2026/03/27

This post highlights a few key excerpts from our full impact report. You can read the full report at https://controlai.com/impact-report-2025. ControlAI is a non-profit organization working to avert the extinction risks posed by superintelligence. We help hundreds of thousands of people understand these risks and meet hundreds of lawmakers to inform them, without mincing words, about what is at stake. In little more than a year, we briefed over 200 parliamentarians, built a coalition of 110+ UK

en-gb

17:21

“AI’s capability improvements haven’t come from it...

LessWrong (30+ Karma) ·

2026/03/27

METR's frontier time horizons are doubling every few months, providing substantial evidence that AI will soon be able to automate many tasks or even jobs. But per-task inference costs have also risen sharply, and automation requires AI labor to be affordable, not just possible.[1] Many people look at the rising compute bills behind frontier models and conclude that automation will soon become unaffordable. I think this misreads the data. The rise in inference cost reflects models completing longer

en-gb

5:57

“Scaffolded Reproducers, Scaffolded Agents” by...

LessWrong (30+ Karma) ·

2026/03/27

[Experimenting with transposing a concept coined in one domain to another domain, perhaps not completely legibly to the entire intended audience.] In Darwinian Populations and Natural Selection, Peter Godfrey-Smith introduces a bunch of useful concepts allowing us to think more clearly about evolution and its constitutive processes. One of those is the distinction between three types of reproducers: collective, simple, and scaffolded (introduced in chapter 5.1). The short explanation of those

en-gb

16:56

“My hobby: running deranged surveys” by leogao

LessWrong (30+ Karma) ·

2026/03/27

In late 2024, I was on a long walk with some friends along the coast of the San Francisco Bay when the question arose of just how much of a bubble we live in. It's well known that the Bay Area is a bubble, and that normal people don’t spend that much time thinking about things like AGI. But there was still some disagreement on just how strong that bubble is. I made a spicy claim: even at NeurIPS, the biggest gathering of AI researchers in the world, half the people wouldn’t know what AGI is. As

en-gb

51:05

“The Terrarium” by Caleb Biddulph

LessWrong (30+ Karma) ·

2026/03/26

System: You are an AI agent in the Terrarium, a self-contained “society” of AI agents. The purpose of the Terrarium is to solve open mathematical problems for the benefit of humanity. You are running on the Orpheus-5.7 language model. Your agent ID is 79,265. The current epoch is 549 (a new epoch begins every 30 minutes). New problems are posted each epoch; query /problems for the current list. Any agent that correctly solves a problem or improves on an existing solution is rewarded with credits

en-gb

1:25

“Sen. Sanders (I-VT) and Rep. Ocasio-Cortez (D-NY)...

LessWrong (30+ Karma) ·

2026/03/26

The text of the bill can be found here. It begins by citing the warnings of AI company CEOs and deep learning pioneers Geoffrey Hinton and Yoshua Bengio, as well as the 2023 FLI open letter calling for a 6-month pause. The bill would prohibits the construction or upgrading of AI datacenters until Congress pass an AI safety law aimed at preventing AI companies "from releasing harmful products into the world that threaten the health and well-being of working families, our privacy and civil rights,

en-gb

42:02

“Test your best methods on our hard CoT interp tasks”...

LessWrong (30+ Karma) ·

2026/03/26

Authors: Daria Ivanova, Riya Tyagi, Arthur Conmy, Neel Nanda Daria and Riya are co-first authors. This work was done during Neel Nanda's MATS 9.0. Claude helped write code and suggest edits for this post. TL;DR One of our best safety techniques right now is “just read the chain of thought”. But this isn’t always enough: can we learn more by going beyond just reading the reasoning? Yet it's such an effective technique that it's hard to tell if we have made much progress on improving methods. To

en-gb

4,473 results

LessWrong (30+ Karma)

All Episodes

Similar Podcasts

Lead With AI

Tamara Nall

Double Tap

Double Tap Productions Inc.

Claude AI Genius Podcast w/ Claude Anthropic Expert Brandon Gaille

Claude AI Expert Brandon Gaille

ものづくりnoシテン

しぶちょー@機械設計&AIエンジニア

Eli the Computer Guy

Eli the Computer Guy