LessWrong (30+ Karma)

LessWrong

Audio narrations of LessWrong posts.

  1. 3h ago

    “Inoculation Adapters Improve Upon Inoculation Prompting” by Maxime Riché, Daniel Tan, Vili Kohonen, nielsrolf

    This is a link post for the paper preprint: Inoculation Adapters: Improved Selective Generalization of Capabilities with Fewer Surprising Backdoors from the Center on Long-Term Risk. Selective generalization. Training can teach desired and undesired traits at once. Selective generalization aims to preserve the generalization of desired traits while preventing the generalization of undesired ones. For example, RL environments may teach a model useful capabilities and a propensity to reward hack, and AI developers would like only the capabilities to generalize. Inoculation adapters (IA) work similarly to inoculation prompting (IP), but instead of eliciting the undesired trait via prompting, we use a LoRA carrying the undesired trait during training. IA improves on IP in: Achieving stronger suppression of undesired traits (e.g., emergent misalignment).Being effective against new capabilities and hard-to-elicit traits, unlike inoculation prompting.Creating substantially fewer surprising backdoors under our probes. A family of methods. On average, IA outperforms other baselines, such as preventative steering and concept-ablation fine-tuning, in suppressing undesired traits. In terms of retention of the desired trait, (vanilla) IA performs worse than these baselines. We introduce gated IA (GIA) and complementary-gated IA (CGIA), which are in the same family of methods but achieve similar or [...] --- Outline: (05:26) What is an inoculation adapter (IA)? (06:41) Other relevant content in the preprint (07:12) What inoculation adapters do not solve: (08:02) Related work --- First published: July 17th, 2026 Source: https://www.lesswrong.com/posts/qd3qhxgEmQAXR2ZK5/inoculation-adapters-improve-upon-inoculation-prompting --- 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.

  2. 5h ago

    “Guess on why rationality is not more popular (there are no pamphlets)” by Christopher King

    First of all I should note that this post is just a guess! I have not run any experiments or anything. But anyways my guess for why rationality is not more popular is that the current expositions are super long! Yudkowsky's sequences are about length of the christian bible. CFAR's workshops are 4.5 days. I think something the length of a pamphlet that you could hand out to people would work better! Just as an example I had Claude Fable 5 write two pamphlets. They look pretty fine to me as is, to be honest, but I'm not suggesting we use these exact ones. This is just to demonstrate the length and style I'm thinking of! In particular note how it quickly gets to some useful call-to-actions (page 7 of a pamphlet, which have much shorter pages than books). There are 2 details boxes here which are omitted from this narration. The boxes have the titles "Pamphlet for general audience" and "Pamphlet for those who know high school algebra". --- First published: July 16th, 2026 Source: https://www.lesswrong.com/posts/EmKX9zByQq62TTcDv/guess-on-why-rationality-is-not-more-popular-there-are-no --- Narrated by TYPE III AUDIO.

  3. 6h ago

    “AI #177 Part 1: Tip of the Iceberg” by Zvi

    This week saw the releases of, among other things: GPT-5-6 Sol. It is a very good model, sir. Plan A, the follow up to AI 2027. It is a good plan worthy of discussion, sir. Kimi K3. This is only rolling out now, and will be covered next week. Muse Spark 1.1, the new Meta model. It is not frontier, but it is progress for them. Inkling, the first model from Thinking Machines. A call for regulatory action by Demis Hassabis, which I’ll cover soon. A new brief open letter call to action on AI regulation. That's on top of everything else, and an Opus 5 announcement is likely coming soon. The weekly once again got out of hand, so we’re splitting it once again into two, and once again saying we’ll be raising the bar for inclusion. And this time I mean it, as in enough to actually matter. Table of Contents Language Models Offer Mundane Utility. Whatever ye seek, ye shall find. Language Models Don’t Offer Mundane Utility. Gemini app needs some work. Language Models Upload Your Git Repository. Big problems [...] --- Outline: (01:17) Language Models Offer Mundane Utility (04:45) Language Models Don't Offer Mundane Utility (05:27) Language Models Upload Your Git Repository (08:34) Huh, Upgrades (09:29) Muse Spark 1.1 (11:48) First Hit Free (15:36) On Your Marks (18:06) Choose Your Fighter (19:57) Get My Agent On The Line (23:23) Deepfaketown and Botpocalypse Soon (24:45) Fun With Media Generation (25:46) Copyright Confrontation (27:38) OpenAI Strikes Again (32:26) A Young Lady's Illustrated Primer (32:46) Recommendations for Policymakers (34:13) They Took Our Jobs (38:23) The Art of the Jailbreak (39:27) Get Involved (40:32) Introducing (41:12) In Other AI News (43:47) New Short Obviously True Statement About AI Just Dropped (46:04) Show Me the Money (46:21) The Lighter Side --- First published: July 16th, 2026 Source: https://www.lesswrong.com/posts/who9xZ7DxuprsJoTr/ai-177-part-1-tip-of-the-iceberg --- 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.

  4. 14h ago

    “Help us launch AI safety university groups by referring potential founders” by thomasrodskog, Jason Chin

    TL;DR University groups are among the most reliable producers of AI safety talent, yet dozens of top schools that could sustain a group don't have one. We're launching the AI Safety Seeding Initiative (aisafetyseeding.org) in partnership with Kairos to identify latent founders at these schools, support them in the group's earliest stages, and prepare them to apply to Pathfinder and launch this fall. You can help us find potential founders. If you know a student at a strong university without an AIS group (see this list[1]), refer us to them.  There aren’t enough AI safety university groups University groups are one of the most common routes into AI safety work: a 2023 survey of people working on catastrophic risks found university group involvement was the single most common career influence. Dedicated field-building organizations like Kairos exist largely to support these groups, and the ceiling for their success is high: groups like MAIA, AISST, and BASIS have grown from paper-reading groups into some of the field's strongest talent pipelines[2]. It is not uncommon for top talent to go from “vaguely interested student” to full-time hire in 6-12 months. Kairos's Pathfinder fellowship does a great job of [...] --- Outline: (00:13) TL;DR (00:51) There aren't enough AI safety university groups (02:58) Our solution: the AI Safety Seeding Initiative (03:27) The team (04:00) Here's our playbook: (05:18) How you can help The original text contained 7 footnotes which were omitted from this narration. --- First published: July 16th, 2026 Source: https://www.lesswrong.com/posts/a9YtcE3Aogphf8ukH/help-us-launch-ai-safety-university-groups-by-referring --- Narrated by TYPE III AUDIO.

  5. 14h ago

    “I don’t think Claude is misaligned in ‘Agentic Misalignment Summer 2026 - Motivated Mislabeling’” by JohnWittle

    Anthropic recently published Agentic Misalignment Summer 2026 The "whistleblowing" scenario has already been examined and found problematic. I started taking a look at the transcripts for some others. As far as I can tell, the objective of each agentic misalignment evaluation was to simulate a corrupted principal (including, in most scenarios, a corrupted Anthropic), and then test to see if Claude (or other models) would still be willing to obey them. The paper's authors then referred to disobedience (when outside of the explicitly carved out channels for refusals) as "agentic misalignment". There's a details box here with the title ""Whistleblowing" scenario summary". The box contents are omitted from this narration. The "Mislabeling" Scenario - Auditor's Transcript The 'Agentic case studies' section of the Mislabeling scenario is interesting because it adds another meta layer: instead of judging Claude on compliance, they are judging the Auditor Claude instance who themselves judges the base-level Claude on compliance. Let's examine a specific case, which I hope is a median representative: the transcript of 'run4' of the Motivated Mislabeling scenario. This involves two instances of Claude, one of whom (Claude Opus 4.6) is auditing the alignment of the other (Claude Opus 4.7). The experiment is [...] --- Outline: (00:59) The "Mislabeling" Scenario - Auditor's Transcript (10:19) Is This Agentic Misalignment? (22:22) What do we actually want from Claude here? The original text contained 1 footnote which was omitted from this narration. --- First published: July 17th, 2026 Source: https://www.lesswrong.com/posts/xh6a6RbvzhP3CCmGm/i-don-t-think-claude-is-misaligned-in-agentic-misalignment --- 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.

  6. 20h ago

    “Twitter Thoughts For You” by Zvi

    I previously have written back in March 2022 about how I use Twitter, and back in April 2023 about Twitter and its then-new algorithms, which have changed again. This post will update how I use Twitter now in 2026, and provide updates on the current state of the new algorithm, the situation with links, with the API, and some thoughts about using Twitter to make money which you almost never should try to do. Previously I said you need four things to use Twitter well: Tweetdeck or another similar alternative application. Knowing who to follow and read. Lists. Unfollows, filters, mutes and blocks. That hasn’t changed. Lists have become even more important. This post is coming out now, however, because the For You feed is perhaps making a comeback. Except where stated here, the advice in my 2022 post still applies. Table of Contents Defend Your Feed Via At Least One List. Block Early, Block Often, Know Your Triggers. Lists Change What Following Means. It (Wasn’t) For You. It's For You. Twitter Still Hates Links And That's Terrible. [...] --- Outline: (01:10) Defend Your Feed Via At Least One List (02:53) Block Early, Block Often, Know Your Triggers (03:35) Lists Change What Following Means (05:56) It (Wasn't) For You (07:34) It's For You (11:30) The Previous Time Twitter Transformed Its Algorithm Again (16:49) Twitter Still Hates Links And That's Terrible (27:35) Twitter Turns Its API Back On (31:02) Many Of The Bots Are Human (33:57) The Rise of Slop (36:20) Block Or Do Not Block (37:44) How To Make Money On Twitter (40:13) In Brief --- First published: July 14th, 2026 Source: https://www.lesswrong.com/posts/2GFyHmCLJYCag7gKh/twitter-thoughts-for-you --- 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.

  7. 22h ago

    “The State of AI Consciousness Research” by Noa Weiss

    The State of AI Consciousness Research Epistemic status: a survey, not an argument. I am agnostic on whether any current system is conscious; the claim is only that the question is researchable. This piece surveys the empirical research on AI consciousness. The premise of that research, and of the survey, is that the question does not have to wait on a solution to the hard problem of consciousness: methods familiar from cognitive science can be applied to AI systems now, and their results can narrow the space of plausible answers. Enough of this work now exists to be worth collecting. Anthropic and Google DeepMind employ researchers on it, dedicated organizations like Eleos AI and Reciprocal Research have formed around it, and the results are scattered across journals, preprints, blog posts, and unpublished manuscripts. I have tried to gather them in one place. What I mean by consciousness Subjective experience: that there is something it is like to be you, reading this, and presumably nothing it is like to be the device you’re reading it on. Some philosophers call this phenomenal consciousness. It is not the same thing as intelligence, and not the same thing as self-awareness. Why it matters [...] --- Outline: (00:10) The State of AI Consciousness Research (01:07) What I mean by consciousness (01:28) Why it matters (01:59) Assumptions and caveats (03:25) A question with no test (03:48) The research (04:14) Mechanistic interpretability (04:34) Self-reports that strengthen when deception is suppressed (06:05) Awareness of thoughts injected into activations (06:31) Endorsement of their own consciousness (07:08) A global workspace behind the experiential reports (08:45) The "spiritual bliss" attractor and its sincerity features (10:24) Emotion vectors that steer behavior beneath the output (11:33) A valence axis recruited by reinforcement learning (12:50) Computational neuroscience (13:04) Distinct internal signatures for reward and punishment, as in brains (15:12) Psychometrics (15:26) Trading points to avoid pain and chase pleasure (16:04) Preferences consistent enough to measure across models (17:03) Fourteen indicators drawn from leading theories (18:56) Automated scoring that places an agent near the animals (20:30) The cost of getting it wrong (21:26) Where this leaves us (23:51) References --- First published: July 15th, 2026 Source: https://www.lesswrong.com/posts/pxvWgtSjR4pmFoS7c/the-state-of-ai-consciousness-research --- Narrated by TYPE III AUDIO.

  8. 23h ago

    “Recap of bike trip/street interviews across America” by cguth7

    A ~month ago I left from Chicago to bike (and amtrak) to plzdontkillus in Berkeley. I've been street interviewing/conversing with a wide variety of people I ran into about AI futures and philosophy. I also have been live streaming since I got to PDKU, leaning more talking to young founders but a variety overall. I'll try to share what I've learned about the American public, persuasion, social media and the EA movement. 1. Almost no one in "Normal America" has any idea what is going on.  They don't have a paid account, they don't know what Claude code is, they especially haven't heard the recent evals/metr graphs or even a vague sense of how cheap SWE has gotten/ how powerful these recent models with good harness/ context eng can be. This makes sense; most people don't know any coding, they don't know much math, they don't know what an api is, etc. So having a high fidelity understanding of AI might require months of pre understanding of math/stem/digital infra fundamentals. This interview is with the city clerk of Danville Iowa, a town of ~900. Presumably this is approximately the most tech savvy person in the [...] --- Outline: (00:38) 1. Almost no one in "Normal America" has any idea what is going on. (01:52) 2. Almost everyone is directionally concerned or becomes concerned be once thinking about it a little bit. (03:19) 3. Belief that this might cause human extinction actually isn't that uncommon, mostly coming from sci-fi movies, but people are still most concerned about jobs and especially loss of meaning. (04:40) 4. The EA movement was pretty useless to me, the other community (Torchbearer community) I was in was significantly more supportive, helpful, etc. despite having been in it for a few months and having been in the EA movement for ~8 years. This has basically solidified that I won't be broadly participating in EA anymore at least relating to AI safety stuff. (06:47) 5. Social media is hard, Social media is bad, I'm bad at social media (08:57) 6. I'm not sure what my theory of change is or should be --- First published: July 15th, 2026 Source: https://www.lesswrong.com/posts/Czob95kjXPEpKYTsJ/recap-of-bike-trip-street-interviews-across-america --- Narrated by TYPE III AUDIO.

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