PODCAST

LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

technology philosophy

All Episodes

4:38

“Five Recent AI Tutoring Studies” by Arjun Panickssery

LessWrong (30+ Karma) ·

2025/01/19

This is a link post.Last week some results were released from a 6-week study using AI tutors in Nigeria. Below I summarize the results of that and four other recent studies about AI tutoring (the dates reflect when the study was conducted rather than when papers were published): Summer 2024 — 15–16-year olds in Nigeria They had 800 students total. The treatment group studied with GPT-based Microsoft Copilot twice weekly for six weeks, studying English. They were just provided an initial prompt to

en-gb

9:23

“Scaling Wargaming for Global Catastrophic Risks with...

LessWrong (30+ Karma) ·

2025/01/19

We at Sentinel are developing an AI-enabled wargaming-tool, grim, to significantly scale up the number of catastrophic scenarios that concerned organizations can explore and to improve emergency response capabilities of, at least, Sentinel. Table of Contents How AI Improves on the State of the ArtImplementation Details, Limitations, and ImprovementsLearnings So FarGet Involved! How AI Improves on the State of the ArtIn a wargame, a group dives deep into a specific scenario in the hopes of

en-gb

8:52

“Beards and Masks?” by jefftk

LessWrong (30+ Karma) ·

2025/01/19

In general, you're not supposed to wear a beard with a respirator mask (N95, P100, etc), at least not in a way where you have facial hair under the seal: But how much worse is the fit? A P100 with a beard is going to filter less well than a P100 without a beard, but does it do as well as an N95? Or is it hopelessly compromised? And is stubble as bad as a full beard? Does length matter? In an emergency where I really need my mask for protection should I shave? Bottom line up

en-gb

24:27

“The quantum red pill or: They lied to you: we live...

LessWrong (30+ Karma) ·

2025/01/18

Audio note: this article contains 159 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. Today's post is in response to the post "Quantum without complications", which I think is a pretty good popular distillation of the basics of quantum mechanics. For any such distillation, there will be people who say "but you missed X important thing". The limit of appeasing such people is to turn your popular distillation into a

en-gb

3:06

“Don’t ignore bad vibes you get from people” by...

LessWrong (30+ Karma) ·

2025/01/18

I think a lot of people have heard so much about internalized prejudice and bias that they think they should ignore any bad vibes they get about a person that they can’t rationally explain. But if a person gives you a bad feeling, don’t ignore that. Both I and several others who I know have generally come to regret it if they’ve gotten a bad feeling about somebody and ignored it or rationalized it away. I’m not saying to endorse prejudice. But my experience is that many types of prejudice feel more

en-gb

10:31

“Timaeus is hiring researchers & engineers” by Jesse...

LessWrong (30+ Karma) ·

2025/01/17

TLDR: We're hiring for research & engineering roles across different levels of seniority. Hires will work on applications of singular learning theory to alignment, including developmental interpretability. About UsTimaeus' mission is to empower humanity by making breakthrough scientific progress on alignment. Our research focuses on applications of singular learning theory to foundational problems within alignment, such as interpretability (via "developmental interpretability"),

en-gb

19:24

“Meta Pivots on Content Moderation” by Zvi

LessWrong (30+ Karma) ·

2025/01/17

There's going to be some changes made.Table of Contents Out With the Fact Checkers. What Happened. Timing is Everything. Balancing Different Errors. Truth and Reconciliation. Fact Check Fact Check. Mistakes Will Be Made. Where We Go From Here. Out With the Fact Checkers Mark Zuckerberg has decided that with Donald Trump soon to be in office, he is allowed to care about free speech again. And he has decided it is time to admit that what was called ‘fact checking’ meant he had for years

en-gb

6:07

“Tax Price Gouging?” by jefftk

LessWrong (30+ Karma) ·

2025/01/17

In the aftermath of a disaster, there is usually a large shift in what people need, what is available, or both. For example, people normally don't use very much ice, but after a hurricane or other disaster that knocks out power, suddenly (a) lots of people want ice and (b) ice production is more difficult. Since people really don't want their food going bad, and they're willing to pay a lot to avoid that, In a world of pure economics, sellers would raise prices. This can have serious

en-gb

23:37

“Thoughts on the conservative assumptions in AI...

LessWrong (30+ Karma) ·

2025/01/17

Work that I’ve done on techniques for mitigating risk from misaligned AI often makes a number of conservative assumptions about the capabilities of the AIs we’re trying to control. (E.g. the original AI control paper, Adaptive Deployment of Untrusted LLMs Reduces Distributed Threats, How to prevent collusion when using untrusted models to monitor each other.) For example: The AIs are consistently trying to subvert safety measures.They’re very good at strategizing about how to subvert them, and can

en-gb

3:29

“A Novel Emergence of Meta-Awareness in LLM...

LessWrong (30+ Karma) ·

2025/01/17

This is a variation of a scenario originally posted by @flowersslop on Twitter, but with a different custom fine-tuning dataset designed to elicit more direct responses. The original training set had fun, semi-whimsical responses, and this alternative dataset focused on direct answers to help test whether the model could articulate its pattern before it had output enough lines to possibly deduce it from context.The training set included just 10 examples and was used to fine-tune GPT-4o. Here is an

en-gb

21:31

“Introducing the WeirdML Benchmark” by Håvard Tveit Ihle

LessWrong (30+ Karma) ·

2025/01/16

WeirdML websiteRelated posts:How good are LLMs at doing ML on an unknown dataset?o1-preview is pretty good at doing ML on an unknown dataset IntroductionHow good are Large Language Models (LLMs) at doing machine learning on novel datasets? The WeirdML benchmark presents LLMs with weird and unusual machine learning tasks, designed to require careful thinking and actual understanding to solve, and tests an LLM's ability to: Actually understand the properties of the data and the problemCome up with an

en-gb

72:44

“Numberwang: LLMs Doing Autonomous Research, and a...

LessWrong (30+ Karma) ·

2025/01/16

SummaryCan LLMs science? The answer to this question can tell us important things about timelines to AGI. In this small pilot experiment, we test frontier LLMs on their ability to perform a minimal version of scientific research, where they must discover a hidden rule about lists of integers by iteratively generating and testing hypotheses. Results are ambiguous: they're mostly pretty bad at it but top systems show apparent signs of life. We're working on a larger, more rigorous experiment in

en-gb

0:39

“Gaming TruthfulQA: Simple Heuristics Exposed Dataset...

LessWrong (30+ Karma) ·

2025/01/16

This is a link post.(Explanation. Also I have no reason to think they hate me.)https://turntrout.com/original-truthfulqa-weaknesses ---First published: January 16th, 2025 Source: https://www.lesswrong.com/posts/57k6xNcWtAtsSTcor/gaming-truthfulqa-simple-heuristics-exposed-dataset --- Narrated by TYPE III AUDIO. ---Images from the article:Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts,

en-gb

46:27

“What Is The Alignment Problem?” by johnswentworth

LessWrong (30+ Karma) ·

2025/01/16

So we want to align future AGIs. Ultimately we’d like to align them to human values, but in the shorter term we might start with other targets, like e.g. corrigibility.That problem description all makes sense on a hand-wavy intuitive level, but once we get concrete and dig into technical details… wait, what exactly is the goal again? When we say we want to “align AGI”, what does that mean? And what about these “human values” - it's easy to list things which are importantly not human values (like

en-gb

7:28

“New, improved multiple-choice TruthfulQA” by...

LessWrong (30+ Karma) ·

2025/01/16

TLDR: There is a potential issue with the multiple-choice versions of our TruthfulQA benchmark (a test of truthfulness in LLMs), which could lead to inflated model scores. This issue was analyzed by a helpful post by Alex Turner (@TurnTrout). We created a new multiple-choice version of TruthfulQA that fixes the issue. We compare models on the old and new versions and find very similar performance. This suggests that models are not exploiting the issue in the old versions to a significant extent, and

en-gb

2:32

“Voluntary Salary Reduction” by jefftk

LessWrong (30+ Karma) ·

2025/01/15

Until recently I thought Julia and I were digging a bit into savings to donate more. With the tighter funding climate for effective altruism we thought it was worth spending down a bit, especially considering that our expenses should decrease significantly in 1.5y when our youngest starts kindergarten. I was surprised, then, when I ran the numbers and realized that despite donating 50% of a reduced income, we were $9k (0.5%) [1] richer than when I left Google two years earlier. This

en-gb

16:33

“On the OpenAI Economic Blueprint” by Zvi

LessWrong (30+ Karma) ·

2025/01/15

Table of Contents Man With a Plan. Oh the Pain. Actual Proposals. For AI Builders. Think of the Children. Content Identification. Infrastructure Week. Paying Attention. Man With a Plan The primary Man With a Plan this week for government-guided AI prosperity was UK Prime Minister Keir Starmer, with a plan coming primarily from Matt Clifford. I’ll be covering that soon. Today I will be covering the other Man With a Plan, Sam Altman, as OpenAI offers its Economic Blueprint. Cyrps1s (CISO

en-gb

0:38

“Lecture Series on Tiling Agents” by abramdemski

LessWrong (30+ Karma) ·

2025/01/15

For my AISC, I'll[1] be presenting more details about the research every Thursday for approximately the next three months. If you are interested in listening in, here is a calendar link.^Maybe there will be guest speakers at some point, EG, the AISC mentees. The original text contained 1 footnote which was omitted from this narration. --- First published: January 14th, 2025 Source:

en-gb

26:09

“Inference-Time-Compute: More Faithful? A Research...

LessWrong (30+ Karma) ·

2025/01/15

Figure 1: Left: Example of models either succeeding or failing to articulate a cue that influences their answer. We edit an MMLU question by prepending a Stanford professor's opinion. For examples like this where the cue changes the model answer, we measure how often models articulate the cue in their CoT. (Here we show only options A and B, rather than all four.) Right: Inference-Time-Compute models articulate the cue more often. The ITC version of Qwen refers to QwQ-32b-Preview, and non-ITC refers

en-gb

7:50

“We probably won’t just play status games with each...

LessWrong (30+ Karma) ·

2025/01/15

There is a view I’ve encountered fairly often,[1] which can be summarized as follows: After the widespread deployment of advanced AGI, assuming humanity survives, material scarcity will largely disappear. Everyone will have sufficient access to necessities like food, housing, and other basic resources. Therefore, the only scarce resource remaining will be “status". As a result, the primary activity humans will engage in will be playing status games with other humans.I want to challenge this idea

en-gb

4,491 results

LessWrong (30+ Karma)

All Episodes

Similar Podcasts

The Automated Daily

TrendTeller

All On AI

All On AI

Elon Musk Podcast

Stage Zero

Reflections

Livewire Labs

科技早8点

之远不知远