PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

5:43

“Attend the 2025 Reproductive Frontiers Summit, June...

LessWrong (30+ Karma) ·

2025/05/10

There should be a community oriented towards the genomic emancipation of humanity. There isn't such a community, but there should be. It's a future worth investing our hope in—a future where parents are able to choose to give their future children the gift of genomic foundations for long, healthy, sane, capable lives.We're inaugurating this community with the Reproductive Frontiers Summit 2025 in Berkeley, CA, June 10—12. Come join us if you want to learn, connect, think, and coordinate about the

en-gb

39:50

“Cheaters Gonna Cheat Cheat Cheat Cheat Cheat” by Zvi

LessWrong (30+ Karma) ·

2025/05/09

Cheaters. Kids these days, everyone says, are all a bunch of blatant cheaters via AI. Then again, look at the game we are forcing them to play, and how we grade it. If you earn your degree largely via AI, that changes two distinct things. You might learn different things. You might signal different things. Both learning and signaling are under threat if there is too much blatant cheating. There is too much cheating going on, too blatantly. Why is that happening? Because the students are

en-gb

15:30

“Slow corporations as an intuition pump for AI R&D...

LessWrong (30+ Karma) ·

2025/05/09

How much should we expect AI progress to speed up after fully automating AI R&D? This post presents an intuition pump for reasoning about the level of acceleration by talking about different hypothetical companies with different labor forces, amounts of serial time, and compute. Essentially, if you'd expect an AI research lab with substantially less serial time and fewer researchers than current labs (but the same cumulative compute) to make substantially less algorithmic progress, you should

en-gb

54:33

“An alignment safety case sketch based on debate” by...

LessWrong (30+ Karma) ·

2025/05/09

Audio note: this article contains 32 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. This post presents a mildly edited form of a new paper by UK AISI's alignment team (the abstract, introduction and related work section are replaced with an executive summary). Read the full paper here. Executive summary AI safety via debate is a promising method for solving part of the alignment problem for ASI (artificial

en-gb

16:50

“Interest In Conflict Is Instrumentally Convergent”...

LessWrong (30+ Karma) ·

2025/05/09

Why is conflict resolution hard? I talk to a lot of event organizers and community managers. Handling conflict is consistently one of the things they find the most stressful, difficult, or time consuming. Why is that? Or, to ask a related question: Why is the ACX Meetups Czar, tasked with collecting and dispersing best practices for meetups, spending so much time writing about social conflict? This essay is not the whole answer, but it is one part of why this is a hard problem. Short answer:

en-gb

28:53

“Misalignment and Strategic Underperformance: An...

LessWrong (30+ Karma) ·

2025/05/08

In the future, we will want to use powerful AIs on critical tasks such as doing AI safety and security R&D, dangerous capability evaluations, red-teaming safety protocols, or monitoring other powerful models. Since we care about models performing well on these tasks, we are worried about sandbagging: that if our models are misaligned [1], they will intentionally underperform. Sandbagging is crucially different from many other situations with misalignment risk, because it involves models

en-gb

8:21

“Orienting Toward Wizard Power” by johnswentworth

LessWrong (30+ Karma) ·

2025/05/08

For months, I had the feeling: something is wrong. Some core part of myself had gone missing. I had words and ideas cached, which pointed back to the missing part. There was the story of Benjamin Jesty, a dairy farmer who vaccinated his family against smallpox in 1774 - 20 years before the vaccination technique was popularized, and the same year King Louis XV of France died of the disease. There was another old post which declared “I don’t care that much about giant yachts. I want a cure for aging

en-gb

20:52

“OpenAI Claims Nonprofit Will Retain Nominal Control”...

LessWrong (30+ Karma) ·

2025/05/07

Your voice has been heard. OpenAI has ‘heard from the Attorney Generals’ of Delaware and California, and as a result the OpenAI nonprofit will retain control of OpenAI under their new plan, and both companies will retain the original mission.Technically they are not admitting that their original plan was illegal and one of the biggest thefts in human history, but that is how you should in practice interpret the line ‘we made the decision for the nonprofit to retain control of OpenAI after hearing

en-gb

2:44

[Linkpost] “It’s ‘Well, actually...’ all the way...

LessWrong (30+ Karma) ·

2025/05/07

This is a link post. Some people (the “Boubas”) don’t like “chemicals” in their food. But other people (the “Kikis”) are like, “uh, everything is chemicals, what do you even mean?” The Boubas are using the word “chemical” differently than the Kikis, and the way they’re using it is simultaneously more specific and less precise than the way the Kikis use it. I think most Kikis implicitly know this, but their identities are typically tied up in being the kind of person who “knows what ‘chemical’

en-gb

60:02

“Please Donate to CAIP (Post 1 of 3 on AI...

LessWrong (30+ Karma) ·

2025/05/07

I am Jason Green-Lowe, the executive director of the Center for AI Policy (CAIP). Our mission is to directly convince Congress to pass strong AI safety legislation. As I explain in some detail in this post, I think our organization has been doing extremely important work, and that we’ve been doing well at it. Unfortunately, we have been unable to get funding from traditional donors to continue our operations. If we don’t get more funding in the next 30 days, we will have to shut down, which will

en-gb

22:23

“UK AISI’s Alignment Team: Research Agenda” by...

LessWrong (30+ Karma) ·

2025/05/07

The UK's AI Security Institute published its research agenda yesterday. This post gives more details about how the Alignment Team is thinking about our agenda. Summary: The AISI Alignment Team focuses on research relevant to reducing risks to safety and security from AI systems which are autonomously pursuing a course of action which could lead to egregious harm, and which are not under human control. No known technical mitigations are reliable past AGI. Our plan is to break down promising

en-gb

36:28

“Will protein design tools solve the snake antivenom...

LessWrong (30+ Karma) ·

2025/05/07

Another note: Just yesterday, the same day this article was released, the New York Times put this out: Universal Antivenom May Grow Out of Man Who Let Snakes Bite Him 200 Times. I was scooped! Somewhat. I added an addendum section discussing this paper at the bottom. Introduction There has been a fair bit of discussion over this recent ‘creating binders against snake venom protein’ paper from the Baker Lab that came out earlier this year, including this article from Derek Lowe. For a quick recap

en-gb

6:07

“Global Risks Weekly Roundup #18/2025: US tariff...

LessWrong (30+ Karma) ·

2025/05/07

Executive summary Shortages of some goods expected in the US starting later this month, because of tariffs. The power outage on the Iberian Peninsula was the worst in Europe since 2003, and Europe may have been close to a more widespread blackout. Trump wants to involve the military in policing within the US. Negotiations between the US and Iran over its nuclear weapons program have stalled. Microsoft will host xAI's Grok model in its cloud service, in a move to reduce Microsoft's dependency on

en-gb

18:55

“Negative Results on Group SAEs” by Josh Engels

LessWrong (30+ Karma) ·

2025/05/07

Introduction Soon after we released Not All Language Model Features Are One-Dimensionally Linear, I started working with @Logan Riggs and @Jannik Brinkmann on a natural followup to the paper: could we build a variant of SAEs that could find multi-dimensional features directly, instead of needing to cluster SAE latents post-hoc like we did in the paper. We worked on this for a few months last summer and tried a bunch of things. Unfortunately, none of our results were that compelling, and eventually

en-gb

19:52

“Zuckerberg’s Dystopian AI Vision” by Zvi

LessWrong (30+ Karma) ·

2025/05/06

You think it's bad now? Oh, you have no idea. In his talks with Ben Thompson and Dwarkesh Patel, Zuckerberg lays out his vision for our AI future.I thank him for his candor. I’m still kind of boggled that he said all of it out loud. We will start with the situation now. How are things going on Facebook in the AI era? Oh, right. Sakib: Again, it happened again. Opened Facebook and I saw this. I looked at the comments and they’re just unsuspecting boomers congratulating the fake AI gen

en-gb

6:45

“$500 + $500 Bounty Problem: An (Approximately)...

LessWrong (30+ Karma) ·

2025/05/06

Audio note: this article contains 61 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. A lot of our work involves "redunds". A random variable _Gamma_ is a(n exact) redund over two random variables _X_1, X_2_ exactly when both _X_1 rightarrow X_2 rightarrow Gamma_ _X_2 rightarrow X_1 rightarrow Gamma_ Conceptually,

en-gb

0:55

“Nonprofit to retain control of OpenAI” by Archimedes

LessWrong (30+ Karma) ·

2025/05/06

The OpenAI Board has an updated plan for evolving OpenAI's structure. OpenAI was founded as a nonprofit, and is today overseen and controlled by that nonprofit. Going forward, it will continue to be overseen and controlled by that nonprofit. Our for-profit LLC, which has been under the nonprofit since 2019, will transition to a Public Benefit Corporation (PBC)–a purpose-driven company structure that has to consider the interests of both shareholders and the mission. --- First

en-gb

11:39

“Five Hinge‑Questions That Decide Whether AGI Is Five...

LessWrong (30+ Karma) ·

2025/05/06

For people who care about falsifiable stakes rather than vibes TL;DR All timeline arguments ultimately turn on five quantitative pivots. Pick optimistic answers to three of them and your median forecast collapses into the 2026–2029 range; pick pessimistic answers to any two and you drift past 2040. The pivots (I think) are: Which empirical curve matters (hardware spend, algorithmic efficiency, or revenue) Whether software‑only recursive self‑improvement (RSI) can accelerate capabilities faster

en-gb

29:58

“GPT-4o Sycophancy Post Mortem” by Zvi

LessWrong (30+ Karma) ·

2025/05/06

Last week I covered that GPT-4o was briefly an (even more than usually) absurd sycophant, and how OpenAI responded to that. Their explanation at that time was paper thin. It didn’t tell us much that we did not already know, and seemed to suggest they had learned little from the incident. Rolling Stone has a write-up of some of the people whose delusions got reinforced by ChatGPT, which has been going on for a while – this sycophancy incident made things way worse but the pattern isn’t new. Here's

en-gb

5:25

[Linkpost] “Tsinghua paper: Does RL Really...

LessWrong (30+ Karma) ·

2025/05/05

This is a link post. arXiv | project page | Authors: Yang Yue, Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang This paper from Tsinghua find that RL on verifiable rewards (RLVR) just increases the frequency at which capabilities are sampled, rather than giving a base model new capabilities. To do this, they compare pass@k scores between a base model and an RLed model. Recall that pass@k is the percentage of questions a model can solve at least once given k attempts at

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