LessWrong (30+ Karma)

LessWrong
  • 0.0 (0)
  • TECHNOLOGY
  • UPDATED DAILY

Audio narrations of LessWrong posts.

  1. 7H AGO

    “Bad Problems Don’t Stop Being Bad Because Somebody’s Wrong About Fault Analysis” by Linch

    Here's a dynamic I’ve seen at least a dozen times: Alice: Man that article has a very inaccurate/misleading/horrifying headline. Bob: Did you know, *actually* article writers don't write their own headlines? … But what I care about is the misleading headline, not your org chart __ Another example I’ve encountered recently is (anonymizing) when a friend complained about a prosaic safety problem at a major AI company that went unfixed for multiple months. Someone else with background information “usefully” chimed in with a long explanation of organizational limitations and why the team responsible for fixing the problem had limitations on resources like senior employees and compute, and actually not fixing the problem was the correct priority for them etc etc etc. But what I (and my friend) cared about was the prosaic safety problem not being fixed! And what this says about the company's ability to proactively respond to and fix future problems. We’re complaining about your company overall. Your internal team management was never a serious concern for us to begin with! __ A third example comes from Kelsey Piper. Kelsey wrote about the (horrifying) recent case where Hantavirus carriers in the recent [...] The original text contained 1 footnote which was omitted from this narration. --- First published: May 8th, 2026 Source: https://www.lesswrong.com/posts/PCsmhN9z65HtC4t5v/bad-problems-don-t-stop-being-bad-because-somebody-s-wrong --- Narrated by TYPE III AUDIO.

    6 min

  2. 12H AGO

    “Write Cause You Have Something to Say” by Logan Riggs

    The ones who are most successful at writeathons (Inkhaven, NaNoWriMo) are those with an overhang of things to say, usually in the form of: draft postsdaydreams When Scott Alexander said: "Whenever I see a new person who blogs every day, it's very rare that that never goes anywhere or they don't get good. That's like my best leading indicator for who's going to be a good blogger." (source). , it may seem you can just write every day, but that'd be Goodharting. There's something hidden in the writing process you can't see: they have something to say. They'll have an idea (somehow) and think it through by [writing it out/sitting quietly/etc]. This can then generate more ideas, some of which aren't even related to the original idea! At this point, though, my imaginary interlocutor would like to say: I'm trying to publish a blog post every day, so of course I'll eventually be bottlenecked on ideas! How do you generate them though? Catching Ideas Have an idea? Write down the idea. This is equivalent to giving your idea-generating process a cookie, reinforcing the habit of generating ideas. Sometimes, when I'm writing one post, a different idea will [...] --- Outline: (01:18) Catching Ideas (02:32) Just \[Write\] and Nobody Will Get Hurt The original text contained 1 footnote which was omitted from this narration. --- First published: May 8th, 2026 Source: https://www.lesswrong.com/posts/h5n3rscJ7he3yLseo/write-cause-you-have-something-to-say-1 --- Narrated by TYPE III AUDIO.

    4 min

  3. 13H AGO

    “AI is Breaking Two Vulnerability Cultures” by jefftk

    A week ago the Copy Fail vulnerability came out, and Hyunwoo Kim immediately realized that the fixes were insufficient, sharing a patch the same day. In doing this he followed standard procedure for Linux, especially within networking: share the security impact with a closed list of Linux security engineers, while fixing the bug quietly and efficiently in the open. His goal was that with only the raw fix public, the knowledge that a serious vulnerability existed could be "embargoed": the people in a position to address it know, but they've agreed not to say anything for a few days. Someone else noticed the change, however, realized the security implications, and shared it publicly. Since it was now out, the embargo was deemed over, and we can now see the full details. It's interesting to see the tension here between two different approaches to vulnerabilities, and think about how this is likely to change with AI acceleration. On one side you have "coordinated disclosure" culture. This is probably the most common approach in computer security. When you discover a security bug you tell the maintainers privately and give them some amount of time (often 90d) [...] --- First published: May 8th, 2026 Source: https://www.lesswrong.com/posts/wKzWGMoubHoHRC4ng/ai-is-breaking-two-vulnerability-cultures --- Narrated by TYPE III AUDIO.

    4 min

  4. 14H AGO

    “Is ProgramBench Impossible?” by frmsaul

    ProgramBench is a new coding benchmark that all frontier models fail spectacularly. We’ve been on a quest for “hard benchmarks” for a while so it's refreshing to see a benchmark where top models do badly. Unfortunately, ProgramBench has one big problem: it's impossible! What is ProgramBench? ProgramBench tests if a model can recreate a program from a “clean room” environment. The model is given only a bit of documentation and black-box access to the program (all the programs are CLIs), then tasked with re-implementing it. How does ProgramBench know if the implementation is correct? It also generates a bunch of unit tests for the program[1]. The re-implementing coding agent doesn't have access to any of those tests. The coding agent only considers a task “resolved” if it passes all of the tests and “almost resolved” if it passes 95% of them. Why is this problematic? Obscure behavior can enter the unit tests without being in the clean room path. An extreme version of this is a backdoor: program that behaves in one way most of the time but behaves totally differently when exposed to a specific string. This wouldn't make a task literally impossible, just incredibly hard in [...] --- Outline: (00:37) What is ProgramBench? (02:41) This seems like a theoretical issue, does it actually happen? (03:11) What can we do differently? The original text contained 4 footnotes which were omitted from this narration. --- First published: May 8th, 2026 Source: https://www.lesswrong.com/posts/3pdyxFi6JS389nptu/is-programbench-impossible --- Narrated by TYPE III AUDIO. --- Images from the article: Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

    5 min

  5. 18H AGO

    “Bringing More Expertise to Bear on Alignment” by Edmund Lau, Geoffrey Irving, Cameron Holmes, David Africa

    Preamble The preamble is less useful for the typical AlignmentForum/LessWrong reader, who may want to skip to Adversaria vs Basinland section. On 28th of October 2025, Geoffrey Irving, Chief Scientist of the UK AI Security Institute, gave a keynote talk (slides) at the Alignment Conference. The conference was organised by the UK AISI and FAR.AI as part of the Alignment Project, which aims to bring experts from relevant fields to make progress on the alignment problem. TLDR: Adversaria vs Basinland. We might be in one of two worlds. One where alignment is adversarial (a security problem), one where it is navigational (a search for good basins of training behaviour). We don't know which world we are in, and how we train and deploy AIs may determine this.We need new disciplines. The field is small, thinly resourced and approached from only a handful of angles. A few well-placed ideas from other disciplines could disproportionately shift what's achievable.Even if this all fails, evidence of hardness is valuable. Moving past broad framing to details Alignment means ensuring that AI systems do what humans want. This is the broad framing. There is, of course, a lot of complexity [...] --- Outline: (00:12) Preamble (01:25) Moving past broad framing to details (02:15) We should plan for superintelligence (03:33) Adversaria vs Basinland (07:04) Which world are we in? (08:39) Why a few ideas might be enough (10:38) Write down problems (11:25) Multiple new ideas, fitting together (13:18) Spherical cows vs the mess (15:03) Conclusion (15:56) Acknowledgement The original text contained 1 footnote which was omitted from this narration. --- First published: May 8th, 2026 Source: https://www.lesswrong.com/posts/cWFsCFyCttsiJwn2j/bringing-more-expertise-to-bear-on-alignment --- Narrated by TYPE III AUDIO. --- Images from the article: Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

    17 min

  6. 1D AGO

    [Linkpost] “How to prevent AI’s 2008 moment (We’re hiring)” by felixgaston

    This is a link post. TL;DR; CeSIA, the French Center for AI Safety is recruiting. French not necessary. Apply by 22 May 2026; Paris or remote in Europe/UK. On August 27, 2005, at an annual symposium in Jackson Hole, Raghuram Rajan, then chief economist of the International Monetary Fund, argued in front of central bank governors and top officials that the innovations of the previous decade in banking had not made the world safer. The financial instruments built over the previous decade, he argued, had become so intricate that even their creators no longer fully understood the risks they carried. Risk had migrated to institutions the supervisory system was not designed to watch. And the people running those institutions were compensated in ways that rewarded short-term performance over long-term stability. The reception was hostile. Lawrence Summers, a former U.S. Treasury Secretary at the time, rose from the audience to attack the paper, calling its premise "slightly Luddite" and "largely misguided," and warning that the kind of changes Rajan argued for would only reduce the productivity of the financial sector. Three years after Jackson Hole, major banks collapsed, first Bear Stearns, then Lehman Brothers, then Merrill Lynch, then AIG. [...] --- First published: May 7th, 2026 Source: https://www.lesswrong.com/posts/gnZyTQFqLhiHdHELC/how-to-prevent-ai-s-2008-moment-we-re-hiring Linkpost URL:https://forum.effectivealtruism.org/posts/7nq5vK2xo85e9GZjC/we-re-hiring-three-people-to-prevent-ai-s-2008-moment --- Narrated by TYPE III AUDIO.

    5 min

  7. 1D AGO

    “AI #167: The Prior Restraint Era Begins” by Zvi

    The era of training frontier models and then releasing them whenever you wanted? That was fun while it lasted. It looks likely to be over now. The White House wants to get an advance look and have the option to veto your release decisions, and it has used this veto on an expansion of access to Mythos. We have additional clarity on what that might mean and it does not look good. Hassett explicitly used the FDA as a parallel, which is the actual worst option unless your goal is to strange or pause AI development in America, without a parallel action from China. That doesn’t seem like a great plan to me and Susie Wiles is out doing damage control. The part where we are talking to China to coordinate model access restrictions does seem better. Anthropic continues its explosive growth, and it continues to strike compute deals. In addition to a long term expanded deal with Google, Anthropic is now leasing SpaceX's Colossus 1, which has let them expand usage limits immediately, and Elon Musk is now speaking positively about Anthropic, including its motivations. This comes as we get testimony in the Musk [...] --- Outline: (01:45) Language Models Offer Mundane Utility (02:45) Language Models Dont Offer Mundane Utility (05:09) Huh, Upgrades (05:37) Grok 4.3 Exists But xAI Kind Of Doesnt (07:02) Show Me The Compute (13:23) On Your Marks (14:05) Copyright Confrontation (14:19) Deepfaketown and Botpocalypse Soon (15:57) Fun With Media Generation (16:36) A Young Ladys Illustrated Primer (16:48) Cyber Lack of Security (17:05) They Took Our Jobs (17:39) The Art of the Jailbreak (17:49) Introducing (18:09) Musk v OpenAI (21:14) Show Me the Money (23:03) Peace In Our Time (26:18) Quiet Speculations (28:24) Quickly, Theres No Time (30:35) The Quest for Sane Regulations (34:04) People Really Hate AI (34:53) Chip City (35:05) The Week in Audio (36:49) People Just Say Things (40:22) People Just Publish Things (41:05) Google Sells Out (42:08) Greetings From Project Glasswing (44:57) The Prior Restraint Era Begins (56:42) Is This Even Legal? (59:49) Pick Up The Phone (01:03:27) Rhetorical Innovation (01:04:05) People On The Internet Sometimes Lie (01:07:14) Goblin Mode (01:08:34) The Mask Comes Off (01:16:51) Aligning a Smarter Than Human Intelligence is Difficult (01:20:00) Some Penalties May Apply (01:22:28) Messages From Janusworld (01:22:41) Good Advice (01:23:42) The Lighter Side --- First published: May 7th, 2026 Source: https://www.lesswrong.com/posts/rn3iKuDcE4SiSg4DW/ai-167-the-prior-restraint-era-begins --- Narrated by TYPE III AUDIO. --- Images from the article: Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

    1h 27m

  8. 1D AGO

    “Mechanistic estimation for wide random MLPs” by Jacob_Hilton

    This post covers joint work with Wilson Wu, George Robinson, Mike Winer, Victor Lecomte and Paul Christiano. Thanks to Geoffrey Irving and Jess Riedel for comments on the post. In ARC's latest paper, we study the following problem: given a randomly initialized multilayer perceptron (MLP), produce an estimate for the expected output of the model under Gaussian input. The usual approach to this problem is to sample many possible inputs, run them all through the model, and take the average. Instead, we produce an estimate "mechanistically", without running the model even once. For wide models, our approach produces more accurate estimates, both in theory and in practice. Paper: Estimating the expected output of wide random MLPs more efficiently than sampling Code: mlp_cumulant_propagation GitHub repo We are excited about this result as an early step towards our goal of producing mechanistic estimates that outperform random sampling for any trained neural network. Drawing an analogy between this goal and a proof by induction, we see this result as (part of) the "base case": handling networks at initialization. We have a vision for the "inductive step", although we expect that to be much more difficult. Summary of results [...] --- Outline: (01:29) Summary of results (04:39) Significance of results (07:18) Extending to trained networks (08:36) Conclusion The original text contained 18 footnotes which were omitted from this narration. --- First published: May 7th, 2026 Source: https://www.lesswrong.com/posts/fsG4m6sRMpomd7Rk6/mechanistic-estimation-for-wide-random-mlps --- Narrated by TYPE III AUDIO. --- Images from the article: Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

    9 min
See All (250)

About

Audio narrations of LessWrong posts.

Information

  • Creator
    LessWrong
  • Years Active
    2023 - 2026
  • Episodes
    250
  • Rating
    Explicit
  • Copyright
    © 2026 All rights reserved
  • Show Website
    LessWrong (30+ Karma)

You Might Also Like