PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

25:24

“Research Notes: Running Claude 3.7, Gemini 2.5 Pro,...

LessWrong (30+ Karma) ·

2025/04/21

Disclaimer: this post was not written by me, but by a friend who wishes to remain anonymous. I did some editing, however. So a recent post my friend wrote has made the point quite clearly (I hope) that LLM performance on the simple task of playing and winning a game of Pokémon Red is highly dependent on the scaffold and tooling provided. In a way, this is not surprising—the scaffold is there to address limitations in what the model can do, and paper over things like lack of long-term context,

en-gb

21:28

“How Close We Are to a Complete List of Imprinted...

LessWrong (30+ Karma) ·

2025/04/20

This post summarizes some of the research I have been doing forBootstrap Bio AKA kman and Genesmith TL;DR We currently don’t have a list of all imprinted genes that are important in human development, but further long-read sequencing in adults and in the human placenta is going to get us close to that goal. Measuring DNA methylation in a low input DNA context is really annoying, because usually we can’t amplify methylated DNA. I explore the only way that appears tractable to me to amplify

en-gb

15:18

“Impact, agency, and taste” by benkuhn

LessWrong (30+ Karma) ·

2025/04/20

I’ve been thinking recently about what sets apart the people who’ve done the best work at Anthropic. You might think that the main thing that makes people really effective at research or engineering is technical ability, and among the general population that's true. Among people hired at Anthropic, though, we’ve restricted the range by screening for extremely high-percentile technical ability, so the remaining differences, while they still matter, aren’t quite as critical. Instead, people's

en-gb

12:21

“Is Gemini now better than Claude at Pokémon?” by...

LessWrong (30+ Karma) ·

2025/04/20

Background: With the release of Claude 3.7 Sonnet, Anthropic promoted a new benchmark: beating Pokémon. Now, Google claims Gemini 2.5 Pro has substantially surpassed Claude's progress on that benchmark. TL:DR: We don't know if Gemini is better at Pokémon than Claude because their playthroughs can't be directly compared. The Metrics Here are Anthropic's and Google's charts: [1]Unfortunately these are using different x and y axes, but it's roughly accurate to say that Gemini has made it nearly twice

en-gb

1:15

“Why Should I Assume CCP AGI is Worse Than USG AGI?”...

LessWrong (30+ Karma) ·

2025/04/19

Though, given my doomerism, I think the natsec framing of the AGI race is likely wrongheaded, let me accept the Dario/Leopold/Altman frame that AGI will be aligned to the national interest of a great power. These people seem to take as an axiom that a USG AGI will be better in some way than CCP AGI. Has anyone written justification for this assumption? I am neither an American citizen nor a Chinese citizen. What would it mean for an AGI to be aligned with "Democracy" or "Confucianism" or "Marxism

en-gb

7:47

“Scaffolding Skills” by Screwtape

LessWrong (30+ Karma) ·

2025/04/19

Epistemic status: Exploratory A scaffold is a lightweight, temporary construction whose point is to make working on other buildings easier. Maybe you could live and work on a scaffold, but why would you? The point is the building. When the building is done, you can take down the scaffold, though you might put it back up if you need to do repairs. If your scaffold was shaky and unsteady, or fell over when you were halfway up, this would make it harder to work on the building.This essay is what we

en-gb

4:42

“What Makes an AI Startup ‘Net Positive’ for Safety?”...

LessWrong (30+ Karma) ·

2025/04/19

In light of the recent news from Mechanize/Epoch and the community discussion it sparked, I'd like to open a conversation about a question some of us grapple with: What constitutes a net-positive AI startup from an AI safety perspective, and what steps can founders take to demonstrate goodwill and navigate the inherent pressures? This feels like an important conversation to have because AI safety seems to have been increasingly encouraging people in the community to build startups (due to a lack of

en-gb

2:19

“The Russell Conjugation Illuminator” by TimmyM

LessWrong (30+ Karma) ·

2025/04/18

Bertrand Russell noted how people often describe the same factual behavior using emotionally opposite language depending on perspective — e.g. I am firm, you are obstinate, he is pigheaded. This framing tactic is now called a Russell Conjugation, and once you start noticing them, they’re everywhere — especially in politics and media. For the past year and a half, I’ve been training a finetuned ChatGPT model, and building a tool to automatically highlight Russell Conjugations in text and suggest

en-gb

23:56

“Handling schemers if shutdown is not an option” by Buck

LessWrong (30+ Karma) ·

2025/04/18

What if getting strong evidence of scheming isn't the end of your scheming problems, but merely the middle? In most of our research and writing on AI control, we’ve emphasized the following situation: The AI developer is deploying a model that they think might be scheming, but they aren’t sure. The objective of the safety team is to ensure that ifthe model is scheming, it will be caught before successfully causing a catastrophe.For example, in our original paper on AI control, the methodology we

en-gb

10:47

“Training AGI in Secret would be Unsafe and...

LessWrong (30+ Karma) ·

2025/04/18

Subtitle: Bad for loss of control risks, bad for concentration of power risks I’ve had this sitting in my drafts for the last year. I wish I’d been able to release it sooner, but on the bright side, it’ll make a lot more sense to people who have already read AI 2027. There's a good chance that AGI will be trained before this decade is out. By AGI I mean “An AI system at least as good as the best human X’ers, for all cognitive tasks/skills/jobs X.” Many people seem to be dismissing this hypothesis

en-gb

8:16

“Three Months In, Evaluating Three Rationalist Cases...

LessWrong (30+ Karma) ·

2025/04/18

I recall seeing three “rationalist” cases for Trump: Richard Ngo on Twitter and elsewhere focused on the realignment of elite coalitions, observing that “most elite institutions have become leftist monocultures” but speculating that “over the next 5–10 years Silicon Valley will become the core of the Republicans.” The left-wing monoculture catastrophically damaged institutional integrity when public-health officials lied during the pandemic and when bureaucrats used threats and intimidation to

en-gb

4:19

“D&D.Sci Tax Day: Adventurers and Assessments” by aphyer

LessWrong (30+ Karma) ·

2025/04/17

This is an entry in the 'Dungeons & Data Science' series, a set of puzzles where players are given a dataset to analyze and an objective to pursue using information from that dataset. Estimated Complexity: 3/5 (this is a guess, I will update based on feedback/seeing how the scenario goes) STORY It's that time of year again. The time when the Tithe Assessment Exactors demand that all adventurers pay taxes on the various monster parts they have hacked off and sold in the past year. And, more

en-gb

10:28

“Can SAE steering reveal sandbagging?” by jordine,...

LessWrong (30+ Karma) ·

2025/04/17

Summary We conducted a small investigation into using SAE features to recover sandbagged capabilities. We used the Goodfire API to pick out 15 SAE features, 5 related to sandbagging, 5 related to general expertise, and 5 unrelated features. We used these features to steer a Llama 3.3 70b Instruct that has been prompted to sandbag We observed that steering can recover up to ~25% of true capabilities We observed that using sandbagging/deception/evaluation-awareness relevant features were slightly

en-gb

7:02

“ALLFED emergency appeal: Help us raise $800,000 to...

LessWrong (30+ Karma) ·

2025/04/17

SUMMARY: ALLFED is making an emergency appeal here due to a serious funding shortfall. Without new support, ALLFED will be forced to cut half our budget in the coming months, drastically reducing our capacity to help build global food system resilience for catastrophic scenarios like nuclear winter, a severe pandemic, or infrastructure breakdown. ALLFED is seeking $800,000 over the course of 2025 to sustain its team, continue policy-relevant research, and move forward with pilot projects that

en-gb

15:23

“AI-enabled coups: a small group could use AI to...

LessWrong (30+ Karma) ·

2025/04/16

We’ve written a new report on the threat of AI-enabled coups. I think this is a very serious risk – comparable in importance to AI takeover but much more neglected. In fact, AI-enabled coups and AI takeover have pretty similar threat models. To see this, here's a very basic threat model for AI takeover: Humanity develops superhuman AI Superhuman AI ismisaligned and power-seeking Superhuman AI seizes powerfor itself And now here's a closely analogous threat model for AI-enabled coups: Humanity

en-gb

24:21

“Ctrl-Z: Controlling AI Agents via Resampling” by...

LessWrong (30+ Karma) ·

2025/04/16

Audio note: this article contains 37 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. Redwood Research has released a new paper, Ctrl-Z: Controlling AI Agents via Resampling. This is the largest and most intricate study of control techniques to date: that is, techniques that aim to prevent catastrophic failures even if egregiously misaligned AIs attempt to subvert the techniques. We extend control protocols to a

en-gb

1:44

“OpenAI rewrote its Preparedness Framework” by Zach...

LessWrong (30+ Karma) ·

2025/04/16

New: https://openai.com/index/updating-our-preparedness-framework/ Old: https://cdn.openai.com/openai-preparedness-framework-beta.pdf Summary Thresholds & responses: https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf#page=5. High and Critical thresholds trigger responses, like in the old PF; responses to Critical thresholds are not yet specified. Three main categories of capabilities: Bio/chem: High capabilities trigger security controls and (for

en-gb

22:43

“OpenAI #13: Altman at TED and OpenAI Cutting Corners...

LessWrong (30+ Karma) ·

2025/04/16

Three big OpenAI news items this week were the FT article describing the cutting of corners on safety testing, the OpenAI former employee amicus brief, and Altman's very good TED Interview.The FT detailed OpenAI's recent dramatic cutting back on the time and resources allocated to safety testing of its models. In the interview, Chris Anderson made an unusually strong effort to ask good questions and push through attempts to dodge answering. Altman did a mix of giving a lot of substantive content

en-gb

39:21

“ASI existential risk: Reconsidering Alignment as a...

LessWrong (30+ Karma) ·

2025/04/16

A post by Michael Nielsen that I found quite interesting. I decided to reproduce the full essay content here, since I think Michael is fine with that, but feel free to let me know to only excerpt it. This is the text for a talk exploring why experts disagree so strongly about whether artificial superintelligence (ASI) poses an existential risk to humanity. I review some key arguments on both sides, emphasizing that the fundamental danger isn't about whether "rogue ASI" gets out of control: it's

en-gb

35:52

“Surprising LLM reasoning failures make me think we...

LessWrong (30+ Karma) ·

2025/04/15

Introduction Writing this post puts me in a weird epistemic position. I simultaneously believe that: The reasoning failures that I'll discuss are strong evidence that current LLM- or, more generally, transformer-based approaches won't get us AGI As soon as major AI labs read about the specific reasoning failures described here, they might fix them But future versions of GPT, Claude etc. succeeding at the tasks I've described here will provide zero evidenceof their ability to reach AGI. If someone

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