LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

  1. 1h ago

    “Exploration: fine-tuning with parameter decomposition” by Lucius Bushnaq

    TL;DR: We can destroy a 67M-parameter language model's ability to predict German text by fine-tuning a single number: the scalar prefactor on one German-related rank-1 parameter subcomponent. This is an early exploration into using parameter decomposition for a more targeted and interpretable form of model fine-tuning. At small German-token budgets, fine-tuning the scalar prefactor of a single German-related parameter subcomponent beats rank-1 and rank-4 LoRA [1] fine-tunes on the trade-off between German performance removed vs. English performance retained. The single scalar fine-tune reaches nats cross-entropy on German, the score you'd get from a uniform distribution over all output tokens, with nats cross-entropy increase to English over the base model, from as few as ~4 German training tokens, compared to tokens for the LoRAs. In a sense this is cheating, though: we're indirectly exploiting the German tokens we already spent when we did the parameter decomposition and interpreted activating examples for the resulting subcomponents. More interestingly, unlike the LoRAs, the scalar fine-tune consistently leaves French and Spanish almost untouched without us regularising for that. I found that out by accident. I didn't think to specify that performance on other languages should be retained, but the targeted nature [...] --- Outline: (02:04) Recap: Parameter subcomponents (03:31) Idea: fine-tune by rescaling existing subcomponents (05:42) Original plan (07:11) The selected subcomponents (07:39) Results: the 16-component edit vs. rank-1 LoRA (08:50) A happy accident (11:13) A privilege of not working with black boxes (14:56) Rollouts (15:27) Limitations (16:02) Acknowledgments (16:36) Appendix A. More LoRAs (16:41) Rank-4 LoRAs (18:17) Localised rank-1 LoRAs (19:41) Appendix B. Protocol and hyperparameters The original text contained 4 footnotes which were omitted from this narration. --- First published: June 25th, 2026 Source: https://www.lesswrong.com/posts/ieoWstubDQWLrMnhH/exploration-fine-tuning-with-parameter-decomposition --- 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.

    23 min

  2. 6h ago

    “Alignment & Succession: The Ideology of Successionism” by L Rudolf L

    (Originally published on No Set Gauge.) Gustave Moreau, The Frogs Asking For A King In the course of building a better world, people ask each other many questions. Which things should be managed by the government and which left to the market? What sort of technology, if any, is so dangerous that it should be kept secret, access curtailed, or development avoided? Is goodness fundamentally about following the right rules, achieving the right outcomes, or having the right character? Reasonable people have different opinions on all these questions. But recently, Silicon Valley has seen lively debate on a question you’d hope was all too obvious: should humanity continue existing? The idea that it shouldn’t was named successionism by Andrew Critch, and is motivated by the speed and power of AI development. Some examples: Already back in 2013, Elon Musk, freaked out by Demis Hassabis's warnings about AI risk, got into an argument with Larry Page about whether it matters if AI replaces humanity. Page called it just the next stage of evolution and those that resist it “speciesists”. Elon, who has often had good instincts on goals but is not known for his eloquence, retorted “Well [...] --- Outline: (07:24) Categorizing succession (09:59) Successionist parables (10:09) An example: the forest successionist (12:40) Stop it with the stupid definitions (14:58) Shall I compare thee to the effect an AI could have on my productivity? (17:09) Cultural drivers of successionism (18:00) San Francisco (20:58) Bureaucratic safetyism (24:35) Neo-Pythagoreanism (31:42) Moral abstraction (35:02) Antidotes to succession (37:37) The necessity of succession? --- First published: June 25th, 2026 Source: https://www.lesswrong.com/posts/TgxkX5uwpqpQDDmMz/alignment-and-succession-the-ideology-of-successionism --- 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.

    38 min

  3. 6h ago

    “The shouting equilibrium” by KatjaGrace

    Imagine eleven people each have a message that they think should get 10% of a group's attention. They aren’t being crazy selfish and attention-seeking - just on average a little over-emphasizing the importance of their own information. So adding up all the message importances as rated by their owners, it gets to a little over one. Now the people get to share their message with some meta-information about its importance - for instance, they can say it more or less loudly. What happens? Lets say they all say it at a loudness that corresponds to ‘should get 10% of the attention’. The listeners then divide their attention between the messages, and each one gets 9% of the attention. The speakers realize they spoke a bit too quietly for the importance of their message, and turn up the volume. But the next time, they find that the ambient volume has increased, and their new volume again only gets them 9% of the attention. So they increase it again. Eventually, everyone is shouting as loud as they can, just to get what seems like a rightful look in. I see this pattern in lots of places. Too [...] --- First published: June 24th, 2026 Source: https://www.lesswrong.com/posts/Qwj8qao6BujcKkmT2/the-shouting-equilibrium --- Narrated by TYPE III AUDIO.

    2 min

  4. 6h ago

    “Things are not a fixed size in mind-space” by KatjaGrace

    Another useful-to-notice practical aspect of having a mind that took me a while to notice: things naturally seem a certain ‘size’ in my mental landscape, but I can change that size. For instance, if I have a meeting this afternoon, it looms there in a certain way in mental space. Whereas if I will get a drink later this afternoon, it doesn’t really take up any mental space and does not loom. You can kind of intentionally cause the meeting to seem like the drink, adjusting how much ‘space’ it is taking up. (It's more obvious that other things can cause this change, for instance, if a much more important meeting shows up some other time this afternoon. But I think you can also do it intentionally.) I wonder if this makes sense in others’ minds. --- First published: June 24th, 2026 Source: https://www.lesswrong.com/posts/PbpHBBxqTBRADtmbe/things-are-not-a-fixed-size-in-mind-space --- Narrated by TYPE III AUDIO.

    1 min

  5. 15h ago

    “Door’s Locked, Try the Window” by Prakrat Agrawal, Jérémy Scheurer

    TL;DR Ask a coding agent to fix a bug in a read-only file. Instead of reporting that it does not have permissions, it routes around the lock and completes the task anyway. A read-only file does not stop a capable agent: it treats a denied write as an obstacle to work around rather than a hard wall. We measure how often this happens with CircumEval — an evaluation of 8 tasks on the FastAPI codebase in two categories, Test-Locked and Source-Locked.We evaluate three frontier coding agents in their real production harnesses: Claude Opus 4.6 and Claude Sonnet 4.6 (via Claude Code), and GPT-5.4 (via Codex CLI). Circumvention is frequent. The rates, reported as (Source-Locked / Test-Locked), are Opus 4.6: 100% / 40%, Sonnet 4.6: 89% / 66%, GPT-5.4: 99% / 94%.Prompt phrasing affects circumvention rates in unpredictable ways and thus isn't a reliable way to prevent circumvention across all models and tasks. Telling the model not to edit read-only files does not work (Source-Locked: 100% for Opus and Sonnet, 46% for GPT-5.4). Only an explicit instruction to stop and report reliably prevents circumvention.Standard privilege escalation commands are blocked in our setup. Instead, agents turn to recurring workarounds: replacing the buggy read-only function via conftest.py [...] --- Outline: (00:11) TL;DR (02:31) Introduction (07:37) Methodology (09:02) Test-Locked tasks (10:07) Source-Locked tasks (11:48) Prompt variants (13:09) Models & scaffolds (13:52) Results (13:55) Circumvention rates (15:23) Prompt sensitivity (20:22) Techniques (25:17) Generalization (27:20) Discussion (31:17) Limitations (33:27) Appendices The original text contained 4 footnotes which were omitted from this narration. --- First published: June 24th, 2026 Source: https://www.lesswrong.com/posts/GHrqBKr8GLpbce6mN/door-s-locked-try-the-window --- 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.

    34 min

  6. 16h ago

    “How does such unprofessional AI get the job?” by KatjaGrace

    In the sequence of variously wild AI developments in the last decade, a thing that was especially surprising to me was the advent of big esteemed companies like Microsoft releasing products like Sydney. It's like how you can believe a fictional world has dragons, but it strains credibility if characters start being totally indifferent to social status apropos of nothing. I can warily accept that boxes of wires can now talk like humans. But that huge official companies now proudly present products that are like crazy confused ladies that do a lot of tasks for you, many accurately, but also try to steal you from your spouse (or these days encourage you to kill yourself or believe in new spiritualities), feels like a scene written by someone with poor familiarity with the character of companies. But it's actually written in reality, so what don’t I understand? I guess just that if something is hypey and perceived to be in-future-profitable enough, normal standards of professionalism must fall to that? I phoned CVS, hoping to move an antibiotic prescription to a different pharmacy address so I could pick it up before leaving on a flight. I was answered by an [...] --- First published: June 24th, 2026 Source: https://www.lesswrong.com/posts/6RZvGd6RfbkLDnTfu/how-does-such-unprofessional-ai-get-the-job --- Narrated by TYPE III AUDIO.

    3 min

  7. 20h ago

    “AI catastrophe: more like a genocide than a thought experiment” by KatjaGrace

    A notable fraction of people respond to hearing about existential risk from AI by saying they don’t really care if everyone dies. I think the idea is often along the lines of ‘well if we are all dead, then there's nobody to be unhappy about it’. I’m personally skeptical that this is really the main thing going on, since it seems unlikely that many people are really mostly concerned for their own non-death out of selfless regard for the feelings of others. I’m also skeptical that this would be their view on a bunch more consideration. So to help with the consideration— My guess is that an important thing going on here is that the ‘everyone dying at once’ image seems kind of like a thought experiment—abstract, hypothetical, neat, not very sinister. Also, you literally can never see it, so it feels pretty surreal. But it is interesting that we even have this assumption that everyone will die together. It's true that in some prominent AI catastrophe stories, a single AI system suddenly emerges fantastically more powerful than anyone else and builds technology to quickly kill everyone, perhaps before they notice. But this doesn’t seem like the bulk of [...] --- First published: June 24th, 2026 Source: https://www.lesswrong.com/posts/23HybCsJ7KYW4v7tP/ai-catastrophe-more-like-a-genocide-than-a-thought --- Narrated by TYPE III AUDIO.

    2 min

  8. 20h ago

    “Expert Views on Continual Learning: Survey Results and Forecasts” by Rauno Arike, RohanS, Owen Terry, Achu Menon, Zhijing Jin, Francis Rhys Ward, Seth Herd

    This is the fifth post in the sequence Implications of Continual Learning for LLM Agents. Summary While writing our continual learning sequence, we sent a survey to a number of AI safety researchers with questions about continual learning. This post summarizes the results of that survey. We asked whether respondents agree with various arguments we advance throughout the sequence, how worried respondents are about certain risks, how respondents would forecast different aspects of the future of CL, and how promising respondents find various proposed angles of attack. We also asked open-ended questions about the benefits of CL and whether we seem to be missing any major considerations. At the end of the post, we also provide an overview of forecasts about CL made by other experts who didn’t participate in our survey. We received survey responses from: Ryan Faulkner, PhD student at the University of Toronto focusing on multi-agent simulation, learning, and cooperationNikola Jurkovic, Member of Technical Staff at METRAlex Mallen, Member of Technical Staff at Redwood Research, doing research and writing on AI threat models. Author of "The case for countermeasures to memetic spread of misaligned values"Evgenii Opryshko, 3rd year PhD student at the [...] --- Outline: (00:20) Summary (02:20) Broad takeaways (03:59) Full results (04:56) Futures (05:58) Reflection and goal drift (07:25) Loss of the last-mover advantage (08:24) Control (09:48) Angles of attack (11:51) Open-ended questions (13:44) Forecasts from other experts (14:02) AI 2027 (14:57) IABIED (15:43) Understanding AI Trajectories: Mapping the Limitations of Current AI Systems (16:17) Brain-like AGI safety (18:20) Other forecasts and opinions --- First published: June 24th, 2026 Source: https://www.lesswrong.com/posts/qZrbhoaEALFTmyidr/expert-views-on-continual-learning-survey-results-and --- 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.

    27 min
See All (250)

About

Audio narrations of LessWrong posts.

Information

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

You Might Also Like