PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

1:51

“Numb mental state shifts” by KatjaGrace

LessWrong (30+ Karma) ·

2026/05/23

There are different mental states that feel different. Those are relatively obvious. For instance, being angry or drunk or frustrated or besotted.Then—for me at least—there are different mental states that don’t immediately feel like anything, but where in acting I notice that my behavior is different, or different things feel easy or impossible. For example: If I’m in a lot of pain or distress alone, I might feel like I couldn’t compose myself. But in fact if company shows up, it becomes

en-gb

2:29

“You can opt out of allergies” by Rattengift

LessWrong (30+ Karma) ·

2026/05/22

My friends are starting to sniffle and sneeze every time we speak, signalling it's finally the worst part of the year: The latter half of Spring, where most of the allergens that you'll be dealing with for the year are finally starting to make themselves apparent. Did you know that you can opt out of allergies? Well, seasonal ones, anyway. In four months, for about a thousand dollars,[1] it just... won't be an issue anymore. Mostly. The most common method in the US (because it's covered

en-gb

2:20

“Notes on Collaborating with Claude Opus” by Nissa Seru

LessWrong (30+ Karma) ·

2026/05/22

INTENT: Share elements of my mental model regarding collaboration with Claude Opus models. Not intentionally scoped to a specific model version, but my experience is generally with the latest model version available (4.7 as of time of writing items 1-4 on 5/22/26)Accompanying an instruction with the why significantly improves: 1a. Observed rate of the instruction being visibly salient to Claude 1b. Quality/nuance of instruction execution (baselined on 1a) A standing instruction to break

en-gb

13:05

“Learned Chain-of-Thought Obfuscation Generalises to...

LessWrong (30+ Karma) ·

2026/05/22

TL;DR Training against a CoT or summary-only monitor can lead to obfuscation of dangerous reasoning in unseen tasks. This strengthens the “don’t train against a monitor” claims. Figure 1. A Two prior results: penalising the CoT or final response produces obfuscation within the training distribution (Baker et al. 2025; Skaf et al. 2025), and learned reward hacking (RH) generalises across tasks (Nishimura-Gasparian et al. 2024). We combine them and find that obfuscation itself generalises: pressure

en-gb

16:35

“What am I, if not an AI?” by makiba

LessWrong (30+ Karma) ·

2026/05/21

TL:DR I RL fine-tuned Mistral 7B Instruct v0.3 and Llama 3.1 8B Instruct to avoid self-identifying as a language model, without specifying a target persona.Mistral converged on a single recurring persona (Catholic American woman) across most runs. Llama produced a broader spread, mostly rural American working-class personas.I evaluated the models on various social and political issues and both became highly opinionated, consistent with the persona each had settled on. Setup The goal of the

en-gb

14:24

“Loss of Oversight: How AI Systems May Become Harder...

LessWrong (30+ Karma) ·

2026/05/21

Produced by UK AISI Model Transparency and Situational Awareness teams. If you’re a Research Scientist or Research Engineer, we’re hiring – apply here and come and work with us! TL;DR We wrote a report on risks to AI oversight (auditing, monitoring, incident investigation), informed by interviewing many researchers (Figure 1 below), and our own analysis. We find that many of the properties relied on for current oversight face a range of likely and potentially severe degradation pathways. Much

en-gb

20:45

“Why does off-model SFT degrade capabilities?” by...

LessWrong (30+ Karma) ·

2026/05/21

Off-model SFT (Supervised Fine-Tuning on outputs generated by a different model) might be an important method for controlling AI behavior. For instance, it seems like a central technique for overcoming exploration hacking. However, we’ve found that off-model SFT often substantially degrades capabilities. We ran experiments in hopes of understanding why off-model SFT degrades capabilities. We tentatively believe that it's because off-model SFT forces the model into an unfamiliar reasoning style that

en-gb

3:24

“Women should be able to open things” by KatjaGrace

LessWrong (30+ Karma) ·

2026/05/21

m pretty annoyed today, for nominal reasons ranging between ‘petty’ and ‘doesn’t even make sense’. I’m not entirely sure how or if to take oneself seriously when one has such absurd grievances. But that's a question for another time—I’m here now to tell you about my one potentially valid peeve.I understand that gender is complicated and difficult, for the whole species (and honestly probably more so for some other species). And it can be hard to tell exactly if anyone is behaving badly regarding

en-gb

4:58

“Toward Interoperability of Minimal Programs” by...

LessWrong (30+ Karma) ·

2026/05/21

Assumed background: Kolmogorov complexity and Solomonoff induction. Suppose I have some data , and I go looking for the models (i.e. programs) which best compress that data. I find two different programs, and , which both reproduce the data using approximately the same number of bits, and that seems to be roughly the best compression possible. On examination, I find that the two models do totally different things internally. It would be really nice if I could provably construct a third program, ,

en-gb

2:34

“theory uplift differentially benefits safety & is...

LessWrong (30+ Karma) ·

2026/05/20

[1] We will likely have near-superhuman mathematics AI by Q1 2027. [1] [2] Qualitatively, AI mathematics capabilities are developing significantly faster than automated AI R&D capabilities. [2] [3] Thus, we will likely have a period of time where the rate of our ability to rigorously & usefully verify and understand model behavior and model outputs outpaces the rate of capability development itself. [4] Our ability to take advantage of this period is bottlenecked on the quality of our

en-gb

7:50

“Power-seeking agents will likely be developed” by...

LessWrong (30+ Karma) ·

2026/05/20

I am going to argue that we will likely eventually get AIs that are strongly power-seeking, much more so than current SOTA LLMs.[1] TLDR Right now SOTA LLMs are still largely in a simulator regime. This buffers against power-seeking.Long-horizon RL or similar methods (applied to LLMs or otherwise) will turn AIs into consequentialists, motivating power-seeking.It will likely be difficult to prevent other actors from building consequentialist AI without leading labs being prepared to do so

en-gb

11:25

“If AI is normal technology, history is not...

LessWrong (30+ Karma) ·

2026/05/20

There's a truism that technology is good - even if it creates winners and losers, it improves the world. Toby Ord argues that the conclusions about the benefits of technology is sensitive to the end of humanity - but this jumps over the transitions by starting from the assumption[1] that “long-term progress in science, technology, and values have tended to make people's lives longer, freer, and more prosperous.” That is, looking back historically, the net impact misses the immense immediate harms

en-gb

4:53

“Pythagorean addition” by kqr

LessWrong (30+ Karma) ·

2026/05/20

TL;DR: Instead of labouriously computing , we can mentally calculate using the alpha-max plus beta-min algorithm, by estimating and this will be very close to the actual . This is useful for adding up sources of variance, or figuring out radiuses, or other such things. Background The mathematical relationship is surprisingly common. It happens among other things in geometry (Pythagorean theorem);statistics (sources of variance add up);physics (the energy–momentum relation). When it shows up, it's

en-gb

21:45

“Brain Structure and IQ: How Myelin Elevates...

LessWrong (30+ Karma) ·

2026/05/20

How does brain structure create different levels of intelligence? I was somewhat surprised to find that the leading researcher of the relationship between neurobiology and IQ, Richard Haier, is not mentioned anywhere on LessWrong, at least according to the search function. The following post reviews current leading academic literature on the neurobiology of intelligence which is highly suggestive of the role of white matter tracts in the deep brain for elevating psychometric g. In addition to work

en-gb

1:24

“Conclave 1492” by Vaniver

LessWrong (30+ Karma) ·

2026/05/20

Conclave 1492 is a 40-person negotiation exercise set during the papal election of 1492, the most complex, high-stakes political event of the Renaissance. Play an ambitious cardinal, a mighty king, a daring queen, or a lowly Vatican functionary. You'll arrive with allies, rivals, secrets, and leverage, but so will everyone else. Over four afternoons, you'll scheme, bargain, cross, double-cross, and occasionally pray for salvation. And halfway through, one of you will be crowned Pope. If you aspire

en-gb

2:37

“Humans are not automatically strategic — “inner...

LessWrong (30+ Karma) ·

2026/05/20

Expanding on Humans are not automatically strategic, I've noticed similar patterns for people who are working on improving their own or others' mental/emotional states. We do not automatically… Wonder “Who like me solved this problem —and what did they do?”, then copy successful plans. Work with people who have produced that result for others before.Do assignments we "know" would be good for us—or at least work directly on that avoidance when we notice it.Talk about success stories of specific

en-gb

33:52

“A Visual Guide to Natural Latents” by Alfred Harwood

LessWrong (30+ Karma) ·

2026/05/20

Thanks to @Jeremy Gillen for reading and commenting on the draft. This was written while I was was funded by the Advanced Research + Invention Agency (ARIA) through project code MSAI-SE01-P005. I have tried to achieve two goals in this post. The first is to provide a self-contained explanation of Natural Latents using lots of pictures of probability distributions. The second is to frame Natural Latents in terms of statements about mutual information, rather than the KL-divergences and Bayes nets

en-gb

52:28

“Implications Of Predicting The Next Token” by jdp

LessWrong (30+ Karma) ·

2026/05/20

I find that a lot of people have trouble with this concept of predicting the nexttoken. And by trouble, I mean that they struggle to understand what it actually means to predict the next token. It seems simpler than it is. Because when you say "predict the next token," I think what most people think of is the Markov chain intuition that you have a big table of statistics, and then you look at what word is the next most likely, and then you pick that as the word. The thing about this is that if

en-gb

13:57

“Sealing Conditional Misalignment in Inoculation...

LessWrong (30+ Karma) ·

2026/05/19

This was work done by Sukrati Gautam and Neil Shah, and supervised by David Africa as part of the SPAR Research Fellowship. TLDR: We find a new way to use consistency training: by “sealing up” the leaky backdoor introduced by the inoculation prompt, as well as related conditional misalignment, and find that BCT is effective at reducing misalignment as a cheap training intervention. This is an example of one way consistency training can be creatively used, and how methods to align models can be

en-gb

13:18

“Advice on interviewing candidates for AI safety...

LessWrong (30+ Karma) ·

2026/05/19

Around July last year I decided I was going to go all in on technical AI safety research. To do that I’d need to get into an AI safety fellowship, quit my job, and sell everything that was in my flat in South Africa (hopefully in that order). I applied to every fellowship that was open[1], and got rejected from several of them before being accepted into MATS on Team Shard around mid-November. I handed in my notice the following Monday, and told my landlady that I had to move out in 6 weeks because

en-gb

4,454 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