Scott Alexander famously warned us to Beware Trivial Inconveniences. When you make a thing easy to do, people often do vastly more of it. When you put up barriers, even highly solvable ones, people often do vastly less. Let us take this seriously, and carefully choose what inconveniences to put where. Let us also take seriously that when AI or othe…

There's this popular trope in fiction about a character being mind controlled without losing awareness of what's happening. Think Jessica Jones, The Manchurian Candidate or Bioshock. The villain uses some magical technology to take control of your brain - but only the part of your brain that's responsible for motor control. You remain conscious and…

This research was conducted at AE Studio and supported by the AI Safety Grants programme administered by Foresight Institute with additional support from AE Studio. Summary In this post, we summarise the main experimental results from our new paper, "Towards Safe and Honest AI Agents with Neural Self-Other Overlap", which we presented orally at the…

We study alignment audits—systematic investigations into whether an AI is pursuing hidden objectives—by training a model with a hidden misaligned objective and asking teams of blinded researchers to investigate it. This paper was a collaboration between the Anthropic Alignment Science and Interpretability teams. Abstract We study the feasibility of…

The Most Forbidden Technique is training an AI using interpretability techniques. An AI produces a final output [X] via some method [M]. You can analyze [M] using technique [T], to learn what the AI is up to. You could train on that. Never do that. You train on [X]. Only [X]. Never [M], never [T]. Why? Because [T] is how you figure out when the mod…

You learn the rules as soon as you’re old enough to speak. Don’t talk to jabberjays. You recite them as soon as you wake up every morning. Keep your eyes off screensnakes. Your mother chooses a dozen to quiz you on each day before you’re allowed lunch. Glitchers aren’t human any more; if you see one, run. Before you sleep, you run through the whole…

Exciting Update: OpenAI has released this blog post and paper which makes me very happy. It's basically the first steps along the research agenda I sketched out here. tl;dr: 1.) They notice that their flagship reasoning models do sometimes intentionally reward hack, e.g. literally say "Let's hack" in the CoT and then proceed to hack the evaluation …

LLM-based coding-assistance tools have been out for ~2 years now. Many developers have been reporting that this is dramatically increasing their productivity, up to 5x'ing/10x'ing it. It seems clear that this multiplier isn't field-wide, at least. There's no corresponding increase in output, after all. This would make sense. If you're doing anythin…

Background: After the release of Claude 3.7 Sonnet,[1] an Anthropic employee started livestreaming Claude trying to play through Pokémon Red. The livestream is still going right now. TL:DR: So, how's it doing? Well, pretty badly. Worse than a 6-year-old would, definitely not PhD-level. Digging in But wait! you say. Didn't Anthropic publish a benchm…

Note: an audio narration is not available for this article. Please see the original text. The original text contained 169 footnotes which were omitted from this narration. The original text contained 79 images which were described by AI. --- First published: March 3rd, 2025 Source: https://www.lesswrong.com/posts/2w6hjptanQ3cDyDw7/methods-for-stron…

In a recent post, Cole Wyeth makes a bold claim: . . . there is one crucial test (yes this is a crux) that LLMs have not passed. They have never done anything important. They haven't proven any theorems that anyone cares about. They haven't written anything that anyone will want to read in ten years (or even one year). Despite apparently memorizing…

This isn't really a "timeline", as such – I don't know the timings – but this is my current, fairly optimistic take on where we're heading. I'm not fully committed to this model yet: I'm still on the lookout for more agents and inference-time scaling later this year. But Deep Research, Claude 3.7, Claude Code, Grok 3, and GPT-4.5 have turned out la…

This is a critique of How to Make Superbabies on LessWrong. Disclaimer: I am not a geneticist[1], and I've tried to use as little jargon as possible. so I used the word mutation as a stand in for SNP (single nucleotide polymorphism, a common type of genetic variation). Background The Superbabies article has 3 sections, where they show: Why: We shou…

This is a link post.Your AI's training data might make it more “evil” and more able to circumvent your security, monitoring, and control measures. Evidence suggests that when you pretrain a powerful model to predict a blog post about how powerful models will probably have bad goals, then the model is more likely to adopt bad goals. I discuss ways t…

I recently wrote about complete feedback, an idea which I think is quite important for AI safety. However, my note was quite brief, explaining the idea only to my closest research-friends. This post aims to bridge one of the inferential gaps to that idea. I also expect that the perspective-shift described here has some value on its own. In classica…

First, let me quote my previous ancient post on the topic: Effective Strategies for Changing Public Opinion The titular paper is very relevant here. I'll summarize a few points. The main two forms of intervention are persuasion and framing. Persuasion is, to wit, an attempt to change someone's set of beliefs, either by introducing new ones or by ch…

In a previous book review I described exclusive nightclubs as the particle colliders of sociology—places where you can reliably observe extreme forces collide. If so, military coups are the supernovae of sociology. They’re huge, rare, sudden events that, if studied carefully, provide deep insight about what lies underneath the veneer of normality a…

This is the abstract and introduction of our new paper. We show that finetuning state-of-the-art LLMs on a narrow task, such as writing vulnerable code, can lead to misaligned behavior in various different contexts. We don't fully understand that phenomenon. Authors: Jan Betley*, Daniel Tan*, Niels Warncke*, Anna Sztyber-Betley, Martín Soto, Xuchan…

It doesn’t look good. What used to be the AI Safety Summits were perhaps the most promising thing happening towards international coordination for AI Safety. This one was centrally coordination against AI Safety. In November 2023, the UK Bletchley Summit on AI Safety set out to let nations coordinate in the hopes that AI might not kill everyone. Ch…

Note: this is a static copy of this wiki page. We are also publishing it as a post to ensure visibility. Circa 2015-2017, a lot of high quality content was written on Arbital by Eliezer Yudkowsky, Nate Soares, Paul Christiano, and others. Perhaps because the platform didn't take off, most of this content has not been as widely read as warranted by …

Arbital was envisioned as a successor to Wikipedia. The project was discontinued in 2017, but not before many new features had been built and a substantial amount of writing about AI alignment and mathematics had been published on the website. If you've tried using Arbital.com the last few years, you might have noticed that it was on its last legs …

We’ve spent the better part of the last two decades unravelling exactly how the human genome works and which specific letter changes in our DNA affect things like diabetes risk or college graduation rates. Our knowledge has advanced to the point where, if we had a safe and reliable means of modifying genes in embryos, we could literally create supe…

Audio note: this article contains 134 uses of latex notation, so the narration may be difficult to follow. There's a link to the original text in the episode description. In a recent paper in Annals of Mathematics and Philosophy, Fields medalist Timothy Gowers asks why mathematicians sometimes believe that unproved statements are likely to be true.…

This is an all-in-one crosspost of a scenario I originally published in three parts on my blog (No Set Gauge). Links to the originals: A History of the Future, 2025-2027 A History of the Future, 2027-2030 A History of the Future, 2030-2040 Thanks to Luke Drago, Duncan McClements, and Theo Horsley for comments on all three parts. 2025-2027 Below is …

On March 14th, 2015, Harry Potter and the Methods of Rationality made its final post. Wrap parties were held all across the world to read the ending and talk about the story, in some cases sparking groups that would continue to meet for years. It's been ten years, and think that's a good reason for a round of parties. If you were there a decade ago…