LessWrong (Curated & Popular)

This is a link post. It is unbearable to not be consuming. All through the house is nothing but silence. The need inside of me is not an ache, it is caustic, sour, the burning desire to be distracted, to be listening, watching, scrolling. Some of the time I think I’m happy. I think this is very good. I go to the park and lie on a blanket in a sun with a book and a notebook. I watch the blades of g...

I spent a few hundred dollars on Anthropic API credits and let Claude individually research every current US congressperson's position on AI. This is a summary of my findings. Disclaimer: Summarizing people's beliefs is hard and inherently subjective and noisy. Likewise, US politicians change their opinions on things constantly so it's hard to know what's up-to-date. Also, I vi...

Since artificial superintelligence has never existed, claims that it poses a serious risk of global catastrophe can be easy to dismiss as fearmongering. Yet many of the specific worries about such systems are not free-floating fantasies but extensions of patterns we already see. This essay examines thirteen distinct ways artificial superintelligence could go wrong and, for each, pairs the abstract...

Boaz Barak, Gabriel Wu, Jeremy Chen, Manas Joglekar [Linkposting from the OpenAI alignment blog, where we post more speculative/technical/informal results and thoughts on safety and alignment.] TL;DR We go into more details and some follow up results from our paper on confessions (see the original blog post). We give deeper analysis of the impact of training, as well as some preliminary comparison...

Two cats fighting for control over my backyard appear to have settled on a particular chain-link fence as the delineation between their territories. This suggests that: Animals are capable of recognizing Schelling points Therefore, Schelling points do not depend on language for their Schelling-ness Therefore, tacit bargaining should be understood not as a special case of bargaining where communica...

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 researc...

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 i...

We’ve significantly upgraded our timelines and takeoff models! It predicts when AIs will reach key capability milestones: for example, Automated Coder / AC (full automation of coding) and superintelligence / ASI (much better than the best humans at virtually all cognitive tasks). This post will briefly explain how the model works, present our timelines and takeoff forecasts, and compare it to our...

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 empowe...

Past years: 2023 2024 Continuing a yearly tradition, I evaluate AI predictions from past years, and collect a convenience sample of AI predictions made this year. In terms of selection, I prefer selecting specific predictions, especially ones made about the near term, enabling faster evaluation. Evaluated predictions made about 2025 in 2023, 2024, or 2025 mostly overestimate AI capabilities advanc...

They say you’re supposed to choose your prior in advance. That's why it's called a “prior”. First, you’re supposed to say say how plausible different things are, and then you update your beliefs based on what you see in the world. For example, currently you are—I assume—trying to decide if you should stop reading this post and do something else with your life. If you’ve read this blog be...

A key risk factor for scheming (and misalignment more generally) is opaque reasoning ability. One proxy for this is how good AIs are at solving math problems immediately without any chain-of-thought (CoT) (as in, in a single forward pass).I've measured this on a dataset of easy math problems and used this to estimate 50% reliability no-CoT time horizon using the same methodology introduced in...

Prior results have shown that LLMs released before 2024 can't leverage 'filler tokens'—unrelated tokens prior to the model's final answer—to perform additional computation and improve performance.[1]I did an investigation on more recent models (e.g. Opus 4.5) and found that many recent LLMs improve substantially on math problems when given filler tokens. That is, I force the LL...

Master version of this on https://parvmahajan.com/2025/12/21/turning-20.html I turn 20 in January, and the world looks very strange. Probably, things will change very quickly. Maybe, one of those things is whether or not we’re still here. This moment seems very fragile, and perhaps more than most moments will never happen again. I want to capture a little bit of what it feels like to be alive righ...

TL;DR LLMs pretrained on data about misaligned AIs themselves become less aligned. Luckily, pretraining LLMs with synthetic data about good AIs helps them become more aligned. These alignment priors persist through post-training, providing alignment-in-depth. We recommend labs pretrain for alignment, just as they do for capabilities. Website: alignmentpretraining.ai Us: geodesicresearch.org | x.co...

Commercial airplane tickets are divided up into coach, business class, and first class. In 2014, Etihad introduced The Residence, a premium experience above first class. The Residence isn't very popular. The reason The Residence isn't very popular is because of economics. A Residence flight is almost as expensive as a private charter jet. Private jets aren't just a little bit better...

At many points now, I've been asked in private for a critique of EA / EA's history / EA's impact and I have ad-libbed statements that I feel guilty about because they have not been subjected to EA critique and refutation. I need to write up my take and let you all try to shoot it down. Before I can or should try to write up that take, I need to fact-check one of my take-central beli...

Screwtape, as the global ACX meetups czar, has to be reasonable and responsible in his advice giving for running meetups. And the advice is great! It is unobjectionably great. I am here to give you more objectionable advice, as another organizer who's run two weekend retreats and a cool hundred rationality meetups over the last two years. As the advice is objectionable (in that, I can see rea...

TL;DR: In 2025, we were in the 1-4 hour range, which has only 14 samples in METR's underlying data. The topic of each sample is public, making it easy to game METR horizon length measurements for a frontier lab, sometimes inadvertently. Finally, the “horizon length” under METR's assumptions might be adding little information beyond benchmark accuracy. None of this is to criticize METR—in...

TL;DR: We train LLMs to accept LLM neural activations as inputs and answer arbitrary questions about them in natural language. These Activation Oracles generalize far beyond their training distribution, for example uncovering misalignment or secret knowledge introduced via fine-tuning. Activation Oracles can be improved simply by scaling training data quantity and diversity. The below is a reprodu...

A couple of years ago, Gavin became frustrated with science journalism. No one was pulling together results across fields; the articles usually didn’t link to the original source; they didn't use probabilities (or even report the sample size); they were usually credulous about preliminary findings (“...which species was it tested on?”); and they essentially never gave any sense of the magnitu...

I have goals that can only be reached via a powerful political machine. Probably a lot of other people around here share them. (Goals include “ensure no powerful dangerous AI get built”, “ensure governance of the US and world are broadly good / not decaying”, “have good civic discourse that plugs into said governance.”) I think it’d be good if there was a powerful rationalist political machine to...

How it started I used to think that anything that LLMs said about having something like subjective experience or what it felt like on the inside was necessarily just a confabulated story. And there were several good reasons for this. First, something that Peter Watts mentioned in an early blog post about LaMDa stuck with me, back when Blake Lemoine got convinced that LaMDa was conscious. Watts not...

Previous: 2024, 2022 “Our greatest fear should not be of failure, but of succeeding at something that doesn't really matter.” –attributed to DL Moody[1] 1. Background & threat model The main threat model I’m working to address is the same as it's been since I was hobby-blogging about AGI safety in 2019. Basically, I think that: The “secret sauce” of human intelligence is a big unifor...

This is the abstract and introduction of our new paper. Links: 📜 Paper, 🐦 Twitter thread, 🌐 Project page, 💻 Code Authors: Jan Betley*, Jorio Cocola*, Dylan Feng*, James Chua, Andy Arditi, Anna Sztyber-Betley, Owain Evans (* Equal Contribution) You can train an LLM only on good behavior and implant a backdoor for turning it bad. How? Recall that the Terminator is bad in the original film but go...