PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

22:19

“Tracing the Thoughts of a Large Language Model” by...

LessWrong (30+ Karma) ·

2025/03/27

[This is our blog post on the papers, which can be found at https://transformer-circuits.pub/2025/attribution-graphs/biology.html and https://transformer-circuits.pub/2025/attribution-graphs/methods.html.] Language models like Claude aren't programmed directly by humans—instead, they‘re trained on large amounts of data. During that training process, they learn their own strategies to solve problems. These strategies are encoded in the billions of computations a model performs for every word it

en-gb

47:27

“Third-wave AI safety needs sociopolitical thinking”...

LessWrong (30+ Karma) ·

2025/03/27

At EA Global Boston last year I gave a talk on how we're in the third wave of EA/AI safety, and how we should approach that. This post contains a (lightly-edited) transcript and slides. Note that in the talk I mention a set of bounties I'll put up for answers to various questions—I still plan to do that, but am taking some more time to get the questions right. Hey everyone. Thanks for coming. You're probably looking at this slide and thinking, how to change a changing world. That's a weird title

en-gb

2:45

[Linkpost] “Center on Long-Term Risk: Summer Research...

LessWrong (30+ Karma) ·

2025/03/27

This is a link post. Summary: CLR is hiring for our Summer Research Fellowship. Join us for eight weeks to work on s-risk motivated empirical AI safety research. Apply here by Tuesday 15th April 23:59 PT. We, the Center on Long-Term Risk, are looking for Summer Research Fellows to explore strategies for reducing suffering in the long-term future (s-risks) and work on technical AI safety ideas related to that. For eight weeks, fellows will be part of our team while working on their own research

en-gb

6:35

“Avoid the Counterargument Collapse” by marknm

LessWrong (30+ Karma) ·

2025/03/27

Selective listening is a real problem. It's really hard to listen to someone when you think you already know what they’re going to say; the words get filtered through our biases and expectations, and sometimes those filters remove the important bits. I want to talk about a specific way I've seen this happen. I don't know if there's a name for it already, so I'm going to call it the counterargument collapse until someone suggests a better term. The pattern I’m talking about is three levels of

en-gb

8:06

“Automated Researchers Can Subtly Sandbag” by...

LessWrong (30+ Karma) ·

2025/03/27

Twitter thread here. tl;dr When prompted, current models can sandbag ML experiments and research decisions without being detected by zero-shot prompted monitors. Claude 3.5 Sonnet (new) can only sandbag effectively when seeing a one-shot example, while Claude 3.7 Sonnet can do this without an example (zero-shot). We are not yet worried about sabotage in today's models, as the sandbagging we observe today would be noticed by humans. Introduction AI models might soon assist with large parts of AI and

en-gb

6:22

“Conceptual Rounding Errors” by Jan_Kulveit

LessWrong (30+ Karma) ·

2025/03/26

Epistemic status: Reasonably confident in the basic mechanism. Have you noticed that you keep encountering the same ideas over and over? You read another post, and someone helpfully points out it's just old Paul's idea again. Or Eliezer's idea. Not much progress here, move along. Or perhaps you've been on the other side: excitedly telling a friend about some fascinating new insight, only to hear back, "Ah, that's just another version of X." And something feels not quite right about that response,

en-gb

11:42

“Eukaryote Skips Town - Why I’m leaving DC” by eukaryote

LessWrong (30+ Karma) ·

2025/03/26

I’ve spent the past 7 years living in the DC area. I moved out there from the Pacific Northwest to go to grad school – I got my masters in Biodefense from George Mason University, and then I stuck around, trying to move into the political/governance sphere. That sort of happened. But I will now be sort of doing that from rural California rather than DC, and I’ll be looking for something else – maybe something more unusual – do to next. A friend asked if this means I’m leaving biosecurity behind

en-gb

32:00

“Goodhart Typology via Structure, Function, and...

LessWrong (30+ Karma) ·

2025/03/26

Audio note: this article contains 127 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. (Work done at Convergence Analysis. The ideas are due to Justin. Mateusz wrote the post. Thanks to Olga Babeeva for feedback on this post.) In this post, we introduce the typology of structure, function, and randomness that builds on the framework introduced in the post Goodhart's Law Causal Diagrams. We aim to present a

en-gb

1:19

[Linkpost] “Latest map of all 40 copyright suits v....

LessWrong (30+ Karma) ·

2025/03/26

This is a link post. Download the latest PDF with links to court dockets here. ---First published: March 26th, 2025 Source: https://www.lesswrong.com/posts/AAxKscHCM4YnfQrzx/latest-map-of-all-40-copyright-suits-v-ai-in-u-s Linkpost URL:https://chatgptiseatingtheworld.com/2025/03/24/latest-map-of-all-40-copyright-suits-v-ai-in-u-s/ --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify

en-gb

53:09

“An overview of areas of control work” by...

LessWrong (30+ Karma) ·

2025/03/26

In this post, I'll list all the areas of control research (and implementation) that seem promising to me. This references framings and abstractions discussed in Prioritizing threats for AI control. First, here is a division of different areas, though note that these areas have some overlaps: Developing and using settings for control evaluations (including technique iteration) Better understanding control-relevant capabilities and model properties Developing specific

en-gb

18:46

“Subversion Strategy Eval: Can language models...

LessWrong (30+ Karma) ·

2025/03/26

We recently released Subversion Strategy Eval: Can language models statelessly strategize to subvert control protocols?, a major update to our previous paper/blogpost, evaluating a broader range of models (e.g. helpful-only Claude 3.5 Sonnet) in more diverse and realistic settings (e.g. untrusted monitoring). Abstract An AI control protocol is a plan for usefully deploying AI systems that prevents an AI from intentionally causing some unacceptable outcome. Previous work evaluated protocols by

en-gb

0:47

“23andMe potentially for sale for $23M” by lemonhope

LessWrong (30+ Karma) ·

2025/03/25

It seems the company has gone bankrupt and wants to be bought and you can probably get their data if you buy it. I'm not sure how thorough their transcription is, but they have records for at least 10 million people (who voluntarily took a DNA test and even paid for it!). Maybe this could be useful for the embryo selection/editing people. --- First published: March 25th, 2025 Source:

en-gb

14:41

“Notes on countermeasures for exploration hacking...

LessWrong (30+ Karma) ·

2025/03/25

If we naively apply RL to a scheming AI, the AI may be able to systematically get low reward/performance while simultaneously not having this behavior trained out because it intentionally never explores into better behavior. As in, it intentionally puts very low probability on (some) actions which would perform very well to prevent these actions from being sampled and then reinforced. We'll refer to this as exploration hacking. In this post, I'll discuss a variety of countermeasures for

en-gb

2:09

“Analyzing long agent transcripts (Docent)” by...

LessWrong (30+ Karma) ·

2025/03/25

This is a brief overview of a recent release by Transluce. You can see the full write-up on the Transluce website. AI systems are increasingly being used as agents: scaffolded systems in which large language models are invoked across multiple turns and given access to tools, persistent state, and so on. Understanding and overseeing agents is challenging, because they produce a lot of text: a single agent transcript can have hundreds of thousands of tokens. At Transluce, we built a system, Docent,

en-gb

48:42

“An overview of control measures” by ryan_greenblatt

LessWrong (30+ Karma) ·

2025/03/25

We often talk about ensuring control, which in the context of this doc refers to preventing AIs from being able to cause existential problems, even if the AIs attempt to subvert our countermeasures. To better contextualize control, I think it's useful to discuss the main countermeasures (building on my prior discussion of the main threats). I'll focus my discussion on countermeasures for reducing existential risk from misalignment, though some of these countermeasures will help with other problems

en-gb

4:18

“Policy for LLM Writing on LessWrong” by jimrandomh

LessWrong (30+ Karma) ·

2025/03/24

LessWrong has been receiving an increasing number of posts and contents that look like they might be LLM-written or partially-LLM-written, so we're adopting a policy. This could be changed based on feedback. Humans Using AI as Writing or Research Assistants Prompting a language model to write an essay and copy-pasting the result will not typically meet LessWrong's standards. Please do not submit unedited or lightly-edited LLM content. You can use AI as a writing or research assistant when writing

en-gb

11:13

“AI ‘Deep Research’ Tools Reviewed” by sarahconstantin

LessWrong (30+ Karma) ·

2025/03/24

Midjourney: “an artificially intelligent researcher, library, posthuman archivist, mapping the noosphere” As regular readers are aware, I do a lot of informal lit review. So I was especially interested in checking out the various AI based “deep research” tools and seeing how they compare. I did a side-by-side comparison, using the same prompt, of Perplexity Deep Research, Gemini Deep Research, ChatGPT-4o Deep Research, Elicit, and PaperQA. General Impressions The Deep Research bots are useful, but

en-gb

14:30

“Recent AI model progress feels mostly like bullshit”...

LessWrong (30+ Karma) ·

2025/03/24

About nine months ago, I and three friends decided that AI had gotten good enough to monitor large codebases autonomously for security problems. We started a company around this, trying to leverage the latest AI models to create a tool that could replace at least a good chunk of the value of human pentesters. We have been working on this project since since June 2024. Within the first three months of our company's existence, Claude 3.5 sonnet was released. Just by switching the portions of our

en-gb

7:49

“Will Jesus Christ return in an election year?” by...

LessWrong (30+ Karma) ·

2025/03/24

Thanks to Jesse Richardson for discussion. Polymarket asks: will Jesus Christ return in 2025? In the three days since the market opened, traders have wagered over $100,000 on this question. The market traded as high as 5%, and is now stably trading at 3%. Right now, if you wanted to, you could place a bet that Jesus Christ will not return this year, and earn over $13,000 if you're right. There are two mysteries here: an easy one, and a harder one. The easy mystery is: if people are willing to bet

en-gb

9:29

“We need (a lot) more rogue agent honeypots” by Ozyrus

LessWrong (30+ Karma) ·

2025/03/24

Epistemic status: Uncertain in writing style, but reasonably confident in content. Want to come back to writing and alignment research, testing waters with this. Current state and risk level I think we're in a phase in AI>AGI>ASI development where rogue AI agents will start popping out quite soon. Pretty much everyone has access to frontier LLMs/VLMs, there are options to run LLMs locally, and it's clear that there are people that are eager to "let them out"; truth terminal is one example of

en-gb

4,489 results

LessWrong (30+ Karma)

All Episodes

Similar Podcasts

Talking Tech w/ Wayne

Wayne Hunter

The Everything Feed - All Packet Pushers Pods

Packet Pushers

The Good Tech Companies

HackerNoon

Engadget News + Next

Engadget

Cybersecurity Tech Brief By HackerNoon

HackerNoon