PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

32:15

“Most likely you won’t be able to perform a...

LessWrong (30+ Karma) ·

2026/03/14

Suppose you’re not happy with the quality of your sleep. You’ve already stopped doing the obviously harmful things (no more coffee at night), and your sleep has improved - but you’d like to work on it further. A coworker gives you an herbal mix with St. John's wort and lavender. You try drinking it at night instead of coffee, and it does seem that sometimes your sleep really does get deeper than before. But sometimes it doesn’t. You’re willing to experiment, but how do you actually check whether

en-gb

14:11

“Cycle-Consistent Activation Oracles” by slavachalnev

LessWrong (30+ Karma) ·

2026/03/14

TL;DR: I train a model to translate LLM activations into natural language, using cycle consistency as a training signal (activation → description → reconstructed activation). The outputs are often plausible, but they are very lossy and are usually guesses about the context surrounding the activation, not good descriptions of the activation itself. This is an interim report with some early results.Overview I think Activation Oracles (Karvonen et al., 2025) are a super exciting research direction

en-gb

16:20

“Things that Go Boom” by sarahconstantin

LessWrong (30+ Karma) ·

2026/03/13

Workers at the Naval Surface Warfare Center in Indian Head, Maryland, the only facility where torpedo fuel is produced for the U.S. Navy In the event of a late-2020s Chinese attempt to invade Taiwan, leading to a US-China military conflict, what would be the most urgent bottlenecks in military equipment? Where is there the greatest need to scale up defense manufacturing? I am not an expert in military matters, so take all this with a grain of salt. But I looked into this question and it seems to

en-gb

96:05

“AI #159: See You In Court” by Zvi

LessWrong (30+ Karma) ·

2026/03/13

The conflict between Anthropic and the Department of War has now moved to the courts, where Anthropic has challenged the official supply chain risk designation as well as the order to remove it from systems across the government, claiming retaliation for protected speech. It will take a bit to work its way through the courts.Anthropic has the principles of law on its side, a maximally strong set of facts and absurdly strong amicus briefs. If Anthropic loses this case, there will be far reaching

en-gb

18:20

“Operationalizing FDT” by Vivek Hebbar

LessWrong (30+ Karma) ·

2026/03/13

This post is an attempt to better operationalize FDT (functional decision theory). It answers the following questions: given a logical causal graph, how do we define the logical do-operator? what is logical causality and how might it be formalized? how does FDT interact with anthropic updating? why do we need logical causality? why FDT and not EDT? Defining the logical do-operator Consider Parfit's hitchhiker:A logical causal graph for Parfit's hitchhiker, where blue nodes are logical facts An FDT

en-gb

17:13

“Are AIs more likely to pursue on-episode or...

LessWrong (30+ Karma) ·

2026/03/13

Consider an AI that terminally pursues reward. How dangerous is this? It depends on how broadly-scoped a notion of reward the model pursues. It could be: on-episode reward-seeking: only maximizing reward on the current training episode — i.e., reward that reinforces their current action in RL. This is what people usually mean by “reward-seeker” (e.g. in Carlsmith or The behavioral selection model…). beyond-episode reward-seeking: maximizing reward for a larger-scoped notion of “self” (e.g., all

en-gb

8:48

“Ideologies Embed Taboos Against Common Knowledge...

LessWrong (30+ Karma) ·

2026/03/12

LLMs are searchable holograms of the text corpus they were trained on. RLHF LLM chat agents have the search tuned to be person-like. While one shouldn't excessively anthropomorphize them, they're helpful for simple experimentation into the latent discursive structure of human writing, because they're often constrained to try to answer probing questions that would make almost any real human storm off in a huff. Previously, I explained a pattern of methodological blind spots in terms of an ideology

en-gb

18:42

“Why AI Evaluation Regimes are bad” by PranavG,...

LessWrong (30+ Karma) ·

2026/03/12

How the flagship project of the AI Safety Community ended up helping AI Corporations. I care about preventing extinction risks from superintelligence. This de facto makes me part of the “AI Safety” community, a social cluster of people who care about these risks. In the community, a few organisations are working on “Evaluations” (which I will shorten to Evals). The most notable examples are Apollo Research, METR, and the UK AISI. Evals make for an influential cluster of safety work, wherein

en-gb

3:34

“Conflicted on Ramsey” by jefftk

LessWrong (30+ Karma) ·

2026/03/12

People are often pretty short-sighted, spending money today thatthey'll want tomorrow. Debt makes it possible to prioritize your current self even more highly: you can spend money you haven't even earned yet. This is a trap many people fall into, and one different communities have built social defenses against. One of the more surprisingly successful approaches is the Financial Peace (Ramsey) system, popular in evangelical Christian communities. It has a series of rules, most prominently

en-gb

25:32

“Dwarkesh Patel on the Anthropic DoW dispute” by anaguma

LessWrong (30+ Karma) ·

2026/03/12

Below is the text of blog post that Dwarkesh Patel wrote on the Anthropic DoW dispute and related topics. He has also narrated it here. By now, I’m sure you’ve heard that the Department of War has declared Anthropic a supply chain risk, because Anthropic refused to remove redlines around the use of their models for mass surveillance and for autonomous weapons. Honestly I think this situation is a warning shot. Right now, LLMs are probably not being used in mission critical ways. But within 20

en-gb

116:47

“How well do models follow their constitutions?” by...

LessWrong (30+ Karma) ·

2026/03/12

This work was conducted during the MATS 9.0 program under Neel Nanda and Senthooran Rajamanoharan. There's been a lot of buzz around Claude's 30K word constitution ("soul doc"), and unusual ways Anthropic is integrating it into training. If we can robustly train complex and nuanced values into a model, this would be a big deal for safety! But does it actually work? This is a preliminary investigation we did to test this. We decomposed the soul doc into 205 testable tenets and ran adversarial

en-gb

120:17

“How Hard a Problem is Alignment? (My Opinionated...

LessWrong (30+ Karma) ·

2026/03/11

Epistemic status: We really need to know. (And I have an opinionated answer.) TL;DR: Comparing person-years of effort, I argue that AI Safety seems harder than for steam engines, but probably less hard than the Apollo program or . I discuss why I suspect superalignment might not be super-hard. My has come down over the last half-decade, primarily because of properties of LLMs, and progress we’ve made in aligning them: I explain why certain previous concerns don’t apply to LLMs, and summarize what

en-gb

9:15

“AIs will be used in “unhinged” configurations” by...

LessWrong (30+ Karma) ·

2026/03/11

Writing up a probably-obvious point that I want to refer to later, with significant writing LLM writing help. TL;DR: 1) A common critique of AI safety evaluations is that they occur in unrealistic settings, such as excessive goal conflict, or are obviously an evaluation rather than “real deployment”.[1] I argue that 2) “real deployment” actually includes many unrealistic and unhinged configurations, due to both widespread prompting techniques, and scaffolding choices and bugs. 1) Background AI

en-gb

5:58

“Can models gradient hack SFT elicitation?” by...

LessWrong (30+ Karma) ·

2026/03/11

TL;DR: Using evidence from tamper resistance, we argue that it would be hard for current models to gradient hack SFT elicitation. Suppose you want to conservatively measure the dangerous capabilities of your LLM and will decide whether to deploy it based on these capabilities, but you’re worried that the LLM might sandbag your evaluation. In the best case, you have a dataset of question-answer pairs, which you can use for supervised fine-tuning (SFT) elicitation on a training set, and then measure

en-gb

39:56

“The Lethal Reality Hypothesis” by Ihor Kendiukhov

LessWrong (30+ Karma) ·

2026/03/11

The epistemic status thing Please read this section because it is not a disclaimer for the sake of it. I think the model I describe in this post is more likely wrong than right. I also think the probability of it being substantially correct is non-negligible and even more than that, and that this alone warrants writing it up and thinking about it seriously. Throughout this post, I will write "X is the case" rather than "it seems plausible that X might be the case, though I am not certain." Every

en-gb

10:15

“The Day After Move 37” by Eneasz

LessWrong (30+ Karma) ·

2026/03/11

I was a few months into 21 years old when a hijacked plane crashed into the first World Trade Center tower. I was commuting in to work listening to the radio (as was the style at the times). I couldn’t figure out how the heck a plane could hit the tower. Was the pilot drunk? How did he even get into the middle of New York City? I was imagining a Cessna because the idea of a passenger plane running into the building was actually unimaginable. I was barely starting to realize “Wait… are they talking

en-gb

14:12

“Less Dead” by Aurelia

LessWrong (30+ Karma) ·

2026/03/11

Come with me if you want to live. – The Terminator 'Close enough' only counts in horseshoes and hand grenades. – Traditional After 10 years of research my company, Nectome, has created a new method for whole-body, whole-brain, human end-of-life preservation for the purpose of future revival. Our protocol is capable of preserving every synapse and every cell in the body with enough detail that current neuroscience says long-term memories are preserved. It's compatible with traditional funerals at

en-gb

3:05

“What do we know about AI company employee giving?”...

LessWrong (30+ Karma) ·

2026/03/11

Many Anthropic employees, especially, are sympathetic to AI safety and (will) have lots of money. This is something that is being talked about a lot (semi-)privately, but I haven't seen any public discussion of it. I find that striking. It seems like the topic is worthy of extensive public discussion, and it seems to me that perhaps this community is inheriting anti-helpful cultural norms against publicly discussing how individuals make use of their money. It also seems likely that many/most of

en-gb

20:28

“AuditBench: Evaluating Alignment Auditing Techniques...

LessWrong (30+ Karma) ·

2026/03/11

TL;DR We release AuditBench, an alignment auditing benchmark. AuditBench consists of 56 language models with implanted hidden behaviors—such as sycophantic deference, opposition to AI regulation, or hidden loyalties—which they do not confess to when asked. We also develop an agent that audits models using a configurable set of tools. Using this agent, we study which tools are most effective for auditing. Introduction Alignment auditing—investigating AI systems to uncover hidden or unintended

en-gb

105:40

“Interview with Steven Byrnes on His Mainline Takeoff...

LessWrong (30+ Karma) ·

2026/03/10

After using the latest version of Claude Code and being surprised how capable it's become while still behaving friendly and corrigibly, I wanted to reflect on how this new observation should update my world model and my P(Doom). So I reached out to Dr. @Steven Byrnes, the highly polymathic AGI safety researcher who I last interviewed in August about Brain-like AGI, foom & doom, and solving technical alignment. We discussed: Steve's high P(Doom) “Brainlike AGI”: A qualitatively different

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