LessWrong (30+ Karma)

Summary Generalization is one lens on the alignment challenge. We'd like network-based AGI to generalize ethical judgments as well as some humans do. Broadening training is a classic and obvious approach to improving generalization in neural networks. Training sets might be broadened to include decisions like whether to evade human control, how to run the world if the opportunity arises, and how to think about one's self and one's goals. Such training might be useful if it's consistent with capability training. But it could backfire if it amounts to lying to a highly intelligent general reasoning system. Broader training sets on types of decisions Training sets for alignment could be broadened in two main ways: types of decisions, and the contexts in which those decisions occur. Any training method could benefit from better training sets, including current alignment training like constitutional AI. The effects of broadening alignment training sets can be investigated empirically, but little work to date directly addresses alignment. Broadening the training set won't solve alignment on its own. It doesn't directly address mesa-optimization concerns. But it should[1] help as part of a hodge-podge collection of alignment approaches. This is a brief take on [...] ---Outline:(00:10) Summary(01:40) Alignment generalization  is more nuanced than IID vs OOD(04:02) Generalization for visual and ethical judgments(07:57) Examples of broadening the training set for alignment(10:24) Broadened training for more human-like representations(12:22) Broadening the training set to include reasoning about goals(15:18) Provisional conclusions and next directions The original text contained 8 footnotes which were omitted from this narration. --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/oveSZmWHjFQbgR4Nn/broadening-the-training-set-should-help-with-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.

The problem of evaluation awareness I've taken on the task of making highly realistic alignment evaluations, and I'm now sure that the mainstream approach of creating such evals is a dead end and should change. When we run unrealistic alignment evals, models recognize that they are being evaluated. For example, when Anthropic evaluated Claude Sonnet 4.5 on their alignment tests, they found that the rate of misalignment had dropped to almost 0% compared to 10% for Claude Sonnet 4.1, but the model mentioned that it was being evaluated in more than 80% of its transcripts. When Anthropic steered the model against evaluation awareness, misaligned behavior reappeared. The evals were measuring evaluation awareness, not alignment because the model recognized that it's being evaluated. The natural solution is to make evals more realistic: Add context, remove artificial patterns, attempt to use feedback from the evaluated model on what makes these evals look fake, but I believe that it's a fundamentally flawed approach. In this post I'll explain why this is the case, and describe some directions that might actually help. The current paradigm and its problems The most common approach for agentic alignment evals follows a pattern: Give a model [...] ---Outline:(00:11) The problem of evaluation awareness(01:21) The current paradigm and its problems(03:11) Why just make it more realistic is unlikely to work(03:33) The problem with identifying eval features(04:52) The somethings off problem(06:37) Capability evals are different(07:10) Possible solutions(07:17) Production evaluations(08:22) Leveraging model internal values(09:05) Modify real conversations(09:29) Conclusion --- First published: January 6th, 2026 Source: https://www.lesswrong.com/posts/GctsnCDxr73G4WiTq/mainstream-approach-for-alignment-evals-is-a-dead-end --- Narrated by TYPE III AUDIO.

TL;DR: Simple inoculation prompts that prevent misalignment generalization in toy setups don't scale to more realistic reward hacking. When I fine-tuned models on realistic reward hacks, only prompts close to the dataset generation prompt were sufficient to prevent misalignment. This seems like a specification problem: the model needs enough information to correctly categorize what's being inoculated. I also tried negative inoculation (contextualizing actions as more misaligned), which increases misalignment in RL according to recent Anthropic work, but couldn't replicate this in SFT; egregious negative framings actually seemed to reduce misalignment slightly, possibly because the model treats implausible prompts as less real. Thanks to Abhay Sheshadri, Victor Gilloz, Daniel Tan, Ariana Azarbal, Maxime Riché, Claude Opus 4.5, and others for helpful conversations, comments and/or feedback. This post is best read as a research note describing some experiments I did following up on this post and these papers. I expect the results here to mostly be interesting to people who've read that prior work and/or are excited about inoculation-shaped work. Introduction In some recent work, my co-authors and I introduced a technique called inoculation prompting. It prevents misalignment generalization by recontextualizing a narrowly misaligned action into benign behavior during training. This [...] ---Outline:(01:23) Introduction(03:23) Inoculation(07:39) Why don't simple prompts work?(08:45) Takeaways(09:28) Negative Inoculation The original text contained 4 footnotes which were omitted from this narration. --- First published: January 6th, 2026 Source: https://www.lesswrong.com/posts/G4YXXbKt5cNSQbjXM/how-hard-is-it-to-inoculate-against-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.

There is a common argument that AI development is dangerous that goes something like: The “goal” of evolution is to make animals which replicate their genes as much as possible; humans do not want to replicate their genes as much as possible; we have some goal which we want AIs to accomplish, and we develop them in a similar way to the way evolution developed humans; therefore, they will not share this goal, just as humans do not share the goal of evolution. This argument sucks. It has serious, fundamental, and, to my thinking, irreparable flaws. This is not to say that its conclusions are incorrect; to some extent, I agree with the point which is typically being made. People should still avoid bad arguments. Problems Why an Analogy? Consider the following argument: Most wars have fewer than one million casualties. Therefore, we should expect that the next war (operationalised in some way) which starts will end with fewer than one million casualties. There are some problems with this. For instance, we might think that modern wars have more or fewer casualties than the reference class of all wars; we might think that some particular conflict is [...] ---Outline:(00:58) Problems(01:01) Why an Analogy?(03:10) Evolution Does Not Have Goals(04:06) Mendel, Not Crick(06:51) Humans Are Not Misaligned(09:19) We Do Not Train AIs Using Evolution(09:36) Evolution Does Not Produce Individual Brains(11:42) ...And So On(14:13) Further Reading The original text contained 8 footnotes which were omitted from this narration. --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/HhkjohHCmJzwWoBmz/the-evolution-argument-sucks --- Narrated by TYPE III AUDIO.

On Thinkish, Neuralese, and the End of Readable Reasoning In September 2025, researchers published the internal monologue of OpenAI's GPT-o3 as it decided to lie about scientific data. This is what it thought: Pardon? This looks like someone had a stroke during a meeting they didn’t want to be in, but their hand kept taking notes. That transcript comes from a recent paper published by researchers at Apollo Research and OpenAI on catching AI systems scheming. To understand what's happening here - and why one of the most sophisticated AI systems in the world is babbling about “synergy customizing illusions” - it first helps to know how we ended up being able to read AI thinking in the first place. That story starts, of all places, on 4chan. In late 2020, anonymous posters on 4chan started describing a prompting trick that would change the course of AI development. It was almost embarrassingly simple: instead of just asking GPT-3 for an answer, ask it instead to show its work before giving its final answer. Suddenly, it started solving math problems that had stumped it moments before. To see why, try multiplying 8,734 × 6,892 in your head. If you’re like [...] The original text contained 3 footnotes which were omitted from this narration. --- First published: January 6th, 2026 Source: https://www.lesswrong.com/posts/gpyqWzWYADWmLYLeX/how-ai-is-learning-to-think-in-secret --- 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.

It seems to be a real view held by serious people that your OpenAI shares will soon be tradable for moons and galaxies. This includes eminent thinkers like Dwarkesh Patel, Leopold Aschenbrenner, perhaps Scott Alexander and many more. According to them, property rights will survive an AI singularity event and soon economic growth is going to make it possible for individuals to own entire galaxies in exchange for some AI stocks. It follows that we should now seriously think through how we can equally distribute those galaxies and make sure that most humans will not end up as the UBI underclass owning mere continents or major planets. I don't think this is a particularly intelligent view. It comes from a huge lack of imagination for the future. Property rights are weird, but humanity dying isn't People may think that AI causing human extinction is something really strange and specific to happen. But it's the opposite: humans existing is a very brittle and strange state of affairs. Many specific things have to be true for us to be here, and when we build ASI there are many preferences and goals that would see us wiped out. It's actually hard to [...] ---Outline:(01:06) Property rights are weird, but humanity dying isnt(01:57) Why property rights wont survive(03:10) Property rights arent enough(03:36) What if there are many unaligned AIs?(04:18) Why would they be rewarded?(04:48) Conclusion --- First published: January 6th, 2026 Source: https://www.lesswrong.com/posts/SYyBB23G3yF2v59i8/on-owning-galaxies --- 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.

This project was conducted as part of the SPAR Fall 2025 cohort. TL;DR Chain-of-thought (CoT) monitoring may serve as a core pillar for AI safety if further advancements in AI capabilities do not significantly degrade the monitorability of LLM serial reasoning.  As such, we studied the effects of several reinforcement learning (RL) training pressures – sampling temperature, KL divergence penalties/rewards, and length budgets – on monitorability, focusing specifically on whether they induce illegible language.  We train R1-distills on math datasets.  Several training setups, especially those using high temperatures, resulted in high accuracy alongside strange reasoning traces containing nonsensical tokens. We noticed some of the same strange tokens being used across different responses by the same model and across different training runs. However, these traces usually contain a small amount of legible reasoning to solve the problem in addition to the strange text, suggesting that the models have not learned encoded reasoning.   We encountered other intriguing phenomena that we highlight, but we are not currently prioritizing studying them because they don’t seem very analogous to realistic monitorability failures. Motivation Models trained with large amounts of outcome-based RL are already starting to exhibit weird language in their reasoning traces (e.g. [...] ---Outline:(00:19) TL;DR(01:36) Motivation(02:54) Methods(05:44) Results(05:47) Sampling temperature + entropy bonus(10:46) Length penalty(16:02) KL-Divergence(21:30) Discussion/Summary/Conclusion The original text contained 6 footnotes which were omitted from this narration. --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/9z6TuKEgZNsmqdfy6/exploring-reinforcement-learning-effects-on-chain-of-thought --- 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.

Currently, we primarily oversee AI with human supervision and human-run experiments, possibly augmented by off-the-shelf AI assistants like ChatGPT or Claude. At training time, we run RLHF, where humans (and/or chat assistants) label behaviors with whether they are good or not. Afterwards, human researchers do additional testing to surface and evaluate unwanted behaviors, possibly assisted by a scaffolded chat agent. The problem with primarily human-driven oversight is that it is not scalable: as AI systems keep getting smarter, errors become harder to detect: The behaviors we care about become more complex, moving from simple classification tasks to open-ended reasoning tasks to long-horizon agentic tasks. Human labels become less reliable due to reward hacking: AI systems become expert at producing answers that look good, regardless of whether they are good. Simple benchmarks become less reliable due to evaluation awareness: AI systems can often tell that they are being evaluated as part of a benchmark and explicitly reason about this. For all these reasons, we need oversight mechanisms that scale beyond human overseers and that can grapple with the increasing sophistication of AI agents. Augmenting humans with current-generation chatbots does not resolve these issues [...] ---Outline:(03:56) Superhuman Oversight from Specialized Assistants(08:30) A Taxonomy of Oversight Questions(13:51) Vision: End-to-End Oversight Assistants The original text contained 8 footnotes which were omitted from this narration. --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/oZuJvSNuYk6busjqf/oversight-assistants-turning-compute-into-understanding --- 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.

Epistemic Status: I wrote the bones of this on August 1st, 2022. I re-read and edited it and added an (unnecessary?) section or three at the end very recently. Possibly useful as a reference. Funny to pair with "semantic stopsigns" (which are an old piece of LW jargon that people rarely use these days). You might be able to get the idea just from the title of the post <3 I'll say this fast, and then offer extended examples, and then go on at length with pointers into deep literatures which I have not read completely because my life is finite. If that sounds valuable, keep reading. If not, not <3 The word "value" is a VERB, and no verb should be performed forever with all of your mind, in this finite world, full of finite beings, with finite brains, running on a finite amounts of energy. However, if any verb was tempting to try to perform infinitely, "valuing" is a good candidate! The problem is that if time runs out a billion years into the future, and you want to be VNM rational about this timescale, you need to link the choice this morning of what to have [...] --- First published: January 4th, 2026 Source: https://www.lesswrong.com/posts/z2uqkmohWKvgytsy6/axiological-stopsigns --- 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.

In the last few weeks, I've been playing around with the newest version of Claude Code, which wrote me a read-it-later service including RSS, email newsletters and an Android app. Software engineering experience was useful, since I did plan out a lot of the high-level design and data model and sometimes push for simpler designs. Overall though, I mostly felt like a product manager trying to specify features as quickly as possible. While software engineering is more than coding, I'm starting to think Claude is already superhuman at this part.Narrating an article from an RSS feed in the web app. The Android app can do this in the background and supports media controls from my car. This was a major change from earlier this year (coding agents were fun but not very useful) and a few months ago (coding agents were good if you held their hands constantly). Claude Opus 4.5 (and supposedly some of the other new models) generally writes reasonable code by default. And while some features had pretty detailed designs, some of my prompts were very minimal. After the first day of this, I mostly just merged PRs without looking at them and assumed they'd work [...] ---Outline:(01:48) Selected Features(01:51) Android App(02:18) Narration(03:01) Selected Problems(03:05) Not Invented Here Syndrome(03:23) Bugs in Dependencies(03:58) Other Observations --- First published: January 4th, 2026 Source: https://www.lesswrong.com/posts/vzaZwZgifypbnSiuf/claude-wrote-me-a-400-commit-rss-reader-app --- 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.

This is a link post. Book review: The Thinking Machine: Jensen Huang, Nvidia, and the World's Most Coveted Microchip, by Stephen Witt. This is a well-written book about the rise of deep learning, and the man who is the most responsible for building the hardware that it needs. Building the Foundations Nvidia was founded in 1993 by three engineers. They failed to articulate much of a business plan, but got funding anyway due to the reputation that Jensen had developed while working for LSI Logic. For 20 years, Nvidia was somewhat successful, but was an erratic performer in a small industry. Around 2004, Nvidia increased the resources it devoted to parallel processing. There was probably some foresight involved in this strategy, but for around a decade Nvidia's results created doubts about its wisdom. The market for such GPUs was tiny. Most experts believed it was too hard to make parallel software work. But Jensen excels at solving problems that most others would reject as impossible. The Most Important Decision Nvidia was sometimes less interested in neural networks than neural network researchers were interested in Nvidia. Geoffrey Hinton sometimes couldn't get Nvidia to [...] ---Outline:(00:27) Building the Foundations(01:17) The Most Important Decision(02:16) Jensens Character(03:41) AI Risks --- First published: January 4th, 2026 Source: https://www.lesswrong.com/posts/NXW2QrKvhZmiGvkcd/the-thinking-machine Linkpost URL:https://bayesianinvestor.com/blog/index.php/2026/01/04/the-thinking-machine/ --- Narrated by TYPE III AUDIO.

Summary: Analysis claiming that automating X% of the economy can only boost GDP by 1/(1-X) assumes all sectors must scale proportionally. The economy is a graph of processes, not a pipeline. Subgraphs can grow independently if they don't bottleneck on inputs from non-growing sectors. AI-driven automation of physical production could create a nearly self-contained subgraph that grows at rates bounded only by raw material availability and speed of production equipment. Models being challenged: This post is a response to Thoughts (by a non-economist) on AI and economics and the broader framing it represents. Related claims appear across LessWrong discussions of AI economic impact: Amdahl's Law for economics: Automating a sector that represents X% of GDP can boost output by at most 1/(1-X). Automating software (2% of GDP) gives ~2% boost; automating all cognitive labor (30% of GDP) gives ~42%. Bottleneck tasks determine growth rate: "Suppose there are three stages in the production process for making a cheese sandwich: make the bread, make the cheese, combine the two together. If the first two stages are automated and can proceed much more quickly, the third stage can still bottleneck the speed of sandwich [...] ---Outline:(03:13) A better model(03:53) An illustrative case(05:24) A hypothetical timeline --- First published: January 4th, 2026 Source: https://www.lesswrong.com/posts/bBmaDRG8XkoHpkrwx/the-economy-is-a-graph-not-a-pipeline --- Narrated by TYPE III AUDIO.

This week, Philip Trammell and Dwarkesh Patel wrote Capital in the 22nd Century. One of my goals for Q1 2026 is to write unified explainer posts for all the standard economic debates around potential AI futures in a systematic fashion. These debates tend to repeatedly cover the same points, and those making economic arguments continuously assume you must be misunderstanding elementary economic principles, or failing to apply them for no good reason. Key assumptions are often unstated and even unrealized, and also false or even absurd. Reference posts are needed. That will take longer, so instead this post covers the specific discussions and questions around the post by Trammell and Patel. My goal is to both meet that post on its own terms, and also point out the central ways its own terms are absurd, and the often implicit assumptions they make that are unlikely to hold. What Trammell and Patel Are Centrally Claiming As A Default Outcome They affirm, as do I, that Piketty was centrally wrong about capital accumulation in the past, for many well understood reasons, many of which they lay out. They then posit that Piketty could have been unintentionally [...] ---Outline:(01:02) What Trammell and Patel Are Centrally Claiming As A Default Outcome(03:15) Does The Above Conclusion Follow From The Above Premises?(03:30) Sounds Like This Is Not Our Main Problem In This Scenario?(07:31) To Be Clear This Scenario Doesn't Make Sense(13:51) Ad Argumento(14:44) The Baseline Scenario(16:31) Would We Have An Inequality Problem?(20:57) No You Can't Simply Let Markets Handle Everything(22:22) Proposed Solutions(25:55) Wealth Taxes Today Are Grade-A Stupid(28:27) Tyler Cowen Responds With Econ Equations(30:29) Brian Albrecht Responds With More Equations --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/pQwNgB7ytwqTxxYue/dos-capital --- 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.

Originally published in No Set Gauge.Dido Building Carthage, J. M. W. Turner In every technological revolution, we face a choice: build for freedom or watch as others build for control. -Brendan McCord There are two moral frames that explain modern moral advances: utilitarianism and liberalism. Utilitarianism says: “the greatest good for the greatest number”. It says we should maximize welfare, whatever that takes. Liberalism says: “to each their own sphere of freedom”. We should grant everyone some boundary that others can’t violate, for example over their physical bodies and their property, and then let whatever happen as long as those boundaries aren’t violated. (Before the philosophers show up and cancel me: “utilitarianism” and “liberalism” here are labels for two views, that correspond somewhat but not exactly to the normal uses of the phrases. The particular axis I talk about comes from my reading of Joe Carlsmith, who I’ll quote at length later in this post. Also, for the Americans: “liberalism” is not a synonym for “left-wing”.) Some of the great moral advances of modernity are: women's rights abolition of slavery equality before the law for all citizens of a country war as an aberration to be avoided, rather [...] ---Outline:(05:50) The real Clippy is the friends we made along the way(17:20) The limits and lights of liberalism(26:08) Weaving the rope(35:05) Luxury space communism or annihilation?(39:43) Technologies of liberalism(49:51) Build for freedom --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/J363Nns4n99TBKZMx/the-technology-of-liberalism --- 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.

Artificial General Intelligence (AGI) poses an extinction risk to all known biological life. Given the stakes involved -- the whole world -- we should be looking at 10% chance-of-AGI-by timelines as the deadline for catastrophe prevention (a global treaty banning superintelligent AI), rather than 50% (median) chance-of-AGI-by timelines, which seem to be the default[1].It's way past crunch time already: 10% chance of AGI this year![2] Given alignment/control is not going to be solved in 2026, and if anyone builds it, everyone dies (or at the very least, the risk of doom is uncomfortably high by most estimates), a global Pause of AGI development is an urgent immediate priority. This is an emergency. Thinking that we have years to prevent catastrophe is gambling a huge amount of current human lives, let alone all future generations and animals. To borrow from Stuart Russell's analogy: if there was a 10% chance of aliens landing this year[3], humanity would be doing a lot more than we are currently doing[4]. AGI is akin to an alien species more intelligent than us that is unlikely to share our values. The original text contained 4 footnotes which were omitted from this narration. --- First published: January 5th, 2026 Source: https://www.lesswrong.com/posts/xEjxQ4txmxTybQLmQ/ai-risk-timelines-10-chance-by-year-x-should-be-the-headline --- Narrated by TYPE III AUDIO.

After five months of me (Buck) being slow at finishing up the editing on this, we’re finally putting out our inaugural Redwood Research podcast. I think it came out pretty well—we discussed a bunch of interesting and underdiscussed topics and I’m glad to have a public record of a bunch of stuff about our history. Tell your friends! Whether we do another one depends on how useful people find this one. You can watch on Youtube here, or as a Substack podcast. Notes on editing the podcast with Claude Code (Buck wrote this section) After the recording, we faced a problem. We had four hours of footage from our three cameras. We wanted it to snazzily cut between shots depending on who was talking. But I don’t truly in my heart believe that it's that important for the video editing to be that good, and I don’t really like the idea of paying a video editor. But I also don’t want to edit the four hours of video myself. And it seemed to me that video editing software was generally not optimized for the kind of editing I wanted to do here (especially automatically cutting between different shots according [...] ---Outline:(00:43) Notes on editing the podcast with Claude Code(03:11) Podcast transcript --- First published: January 4th, 2026 Source: https://www.lesswrong.com/posts/p4iJpumHt6Ay9KnXT/the-inaugural-redwood-research-podcast --- Narrated by TYPE III AUDIO.

For the past year I've been sinking into the Great Books via the Penguin Great Ideas series, because I wanted to be conversant in the Great Conversation. I am occasionally frustrated by this endeavour, but overall, it's been fun! I'm learning a lot about my civilization and the various curmudgeons that shaped it. But one dismaying side effect is that it's also been quite empowering for my inner 13 year old edgelord. Did you know that before we invented woke, you were just allowed to be openly contemptuous of people? Here's Schopenhauer on the common man: They take an objective interest in nothing whatever. Their attention, not to speak of their mind, is engaged by nothing that does not bear some relation, or at least some possible relation, to their own person: otherwise their interest is not aroused. They are not noticeably stimulated even by wit or humour; they hate rather everything that demands the slightest thought. Coarse buffooneries at most excite them to laughter: apart from that they are earnest brutes – and all because they are capable of only subjective interest. It is precisely this which makes card-playing the most appropriate amusement for them – card-playing for [...] The original text contained 3 footnotes which were omitted from this narration. --- First published: January 4th, 2026 Source: https://www.lesswrong.com/posts/otgrxjbWLsrDjbC2w/in-my-misanthropy-era --- 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.

The Rationalist Project was our last best hope that we might not try to build it. It failed. But in the year of the Coding Agent, it became something greater: our last, best hope – for everyone not dying. This is what 2026 looks like. The place is Lighthaven. Table of Contents Language Models Offer Mundane Utility. 2026 is an age of wonders. Claude Code. The age of humans writing code may be coming to an end. Language Models Don’t Offer Mundane Utility. Your dog's dead, Jimmy. Deepfaketown and Botpocalypse Soon. Keep your nonsense simple. Fun With Media Generation. YouTube facing less AI slop than I’d expect. You Drive Me Crazy. Another lawsuit against OpenAI. This one is a murder. They Took Our Jobs. Yet another round of ‘oh but comparative advantage.’ Doctor Doctor. Yes a lot of people still want a human doctor, on principle. Jevons Paradox Strikes Again. It holds until it doesn’t. Unprompted Attention. Concepts, not prompts. The Art of the Jailbreak. Love, Pliny. Get Involved. CAISI wants an intern, OpenAI hiring a head of preparedness. Introducing. [...] ---Outline:(00:30) Language Models Offer Mundane Utility(01:34) Claude Code(08:49) Language Models Don't Offer Mundane Utility(10:04) Deepfaketown and Botpocalypse Soon(12:17) Fun With Media Generation(12:47) You Drive Me Crazy(13:39) They Took Our Jobs(17:46) Doctor Doctor(18:46) Jevons Paradox Strikes Again(20:52) Unprompted Attention(22:42) The Art of the Jailbreak(23:22) Get Involved(24:09) Introducing(24:30) In Other AI News(25:56) Show Me the Money(26:14) Quiet Speculations(29:38) People Really Do Not Like AI(32:23) Americans Remain Optimistic About AI?(33:40) Thank You, Next(36:36) The Quest for Sane Regulations(39:42) Chip City(42:54) Rhetorical Innovation(43:29) Aligning a Smarter Than Human Intelligence is Difficult(44:25) People Are Worried About AI Killing Everyone(44:45) The Lighter Side --- First published: January 1st, 2026 Source: https://www.lesswrong.com/posts/qp7kEfd2MnGRR8evZ/ai-149-3 --- 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.

How to lose weight, infringe patents, and possibly poison yourself for 22 Euros a month. Introduction In March 2025, Scott Alexander wrote: Others are turning amateur chemist. You can order GLP-1 peptides from China for cheap. Once you have the peptide, all you have to do is put it in the right amount of bacteriostatic water. In theory this is no harder than any other mix-powder-with-water task. But this time if you do anything wrong, or are insufficiently clean, you can give yourself a horrible infection, or inactivate the drug, or accidentally take 100x too much of the drug and end up with negative weight and float up into the sky and be lost forever. ACX cannot in good conscience recommend this cheap, common, and awesome solution. With a BMI of about 28, low executive function, a bit of sleep apnea and no willpower to spend on either dieting or dealing with the medical priesthood, I thought I would give it a try. This is a summary of my journey. Please do not expect any great revelations here beyond "you can buy semaglutide from China, duh". All of the details here can also be [...] ---Outline:(00:17) Introduction(01:58) Picking a substance and route of administration(03:03) Finding a vendor(04:52) Sourcing bacteriostatic water(05:47) Other equipment(06:14) My current procedure(08:00) Outcome (_N=1_) The original text contained 9 footnotes which were omitted from this narration. --- First published: January 1st, 2026 Source: https://www.lesswrong.com/posts/coLiSHpP338Xwibbp/the-bio-pirate-s-guide-to-glp-1-agonists --- Narrated by TYPE III AUDIO.

Every time I see someone mention statistics on nonconsent kink online, someone else is surprised by how common it is. So let's start with some statistics from Lehmiller[1]: roughly two thirds of women and half of men have some fantasy of being raped. A lot of these are more of a rapeplay fantasy than an actual rape fantasy, but for purposes of this post we don’t need to get into those particular weeds. The important point is: the appeal of nonconsent is the baseline, not the exception, especially for women. But this post isn’t really about rape fantasies. I claim that the preference for nonconsent typically runs a lot deeper than a sex fantasy, mostly showing up in ways less extreme and emotionally loaded. I also claim that “deep nonconsent preference”, specifically among women, is the main thing driving the apparent “weirdness” of dating/mating practices compared to other human matching practices (like e.g. employer/employee matching). Let's go through a few examples, to illustrate what I mean by “deep nonconsent preference”, specifically for (typical) women. Generalizing just a little bit beyond rape fantasies: AFAICT, being verbally asked for consent is super-duper a turn off for most women. Same with having [...] ---Outline:(01:57) Alternative Hypotheses(03:10) Hypothesis: Being Asked Out Is A Turn Off(05:19) Subtle Signals and Blindspots(08:44) ... But Then Reality Hits Back(10:31) The Weirdness of Dating/Mating The original text contained 2 footnotes which were omitted from this narration. --- First published: January 2nd, 2026 Source: https://www.lesswrong.com/posts/e4TyoEfTeW7FFcwYy/the-weirdness-of-dating-mating-deep-nonconsent-preference --- Narrated by TYPE III AUDIO.