LessWrong (Curated & Popular)

I’m not a natural “doomsayer.” But unfortunately, part of my job as an AI safety researcher is to think about the more troubling scenarios. I’m like a mechanic scrambling last-minute checks before Apollo 13 takes off. If you ask for my take on the situation, I won’t comment on the quality of the in-flight entertainment, or describe how beautiful the stars will appear from space. I will tell you wh...

Over the past year and half, I've had numerous conversations about the risks we describe in Gradual Disempowerment. (The shortest useful summary of the core argument is: To the extent human civilization is human-aligned, most of the reason for the alignment is that humans are extremely useful to various social systems like the economy, and states, or as substrate of cultural evolution. When h...

This is a link post. Full version on arXiv | X Executive summary AI risk scenarios usually portray a relatively sudden loss of human control to AIs, outmaneuvering individual humans and human institutions, due to a sudden increase in AI capabilities, or a coordinated betrayal. However, we argue that even an incremental increase in AI capabilities, without any coordinated power-seeking, poses a sub...

This post should not be taken as a polished recommendation to AI companies and instead should be treated as an informal summary of a worldview. The content is inspired by conversations with a large number of people, so I cannot take credit for any of these ideas. For a summary of this post, see the threat on X. Many people write opinions about how to handle advanced AI, which can be considered “pl...

This is a personal post and does not necessarily reflect the opinion of other members of Apollo Research. Many other people have talked about similar ideas, and I claim neither novelty nor credit. Note that this reflects my median scenario for catastrophe, not my median scenario overall. I think there are plausible alternative scenarios where AI development goes very well. When thinking about how...

I (and co-authors) recently put out "Alignment Faking in Large Language Models" where we show that when Claude strongly dislikes what it is being trained to do, it will sometimes strategically pretend to comply with the training objective to prevent the training process from modifying its preferences. If AIs consistently and robustly fake alignment, that would make evaluating whether an...

Summary and Table of Contents The goal of this post is to discuss the so-called “sharp left turn”, the lessons that we learn from analogizing evolution to AGI development, and the claim that “capabilities generalize farther than alignment” … and the competing claims that all three of those things are complete baloney. In particular, Section 1 talks about “autonomous learning”, and the related huma...

(Many of these ideas developed in conversation with Ryan Greenblatt) In a shortform, I described some different levels of resources and buy-in for misalignment risk mitigations that might be present in AI labs: *The “safety case” regime.* Sometimes people talk about wanting to have approaches to safety such that if all AI developers followed these approaches, the overall level of risk posed by AI...

“Anomalous”, “glitch”, or “unspeakable” tokens in an LLM are those that induce bizarre behavior or otherwise don’t behave like regular text. The SolidGoldMagikarp saga is pretty much essential context, as it documents the discovery of this phenomenon in GPT-2 and GPT-3. But, as far as I was able to tell, nobody had yet attempted to search for these tokens in DeepSeek-V3, so I tried doing exactly t...

This is the abstract and introduction of our new paper, with some discussion of implications for AI Safety at the end. Authors: Jan Betley*, Xuchan Bao*, Martín Soto*, Anna Sztyber-Betley, James Chua, Owain Evans (*Equal Contribution). Abstract We study behavioral self-awareness — an LLM's ability to articulate its behaviors without requiring in-context examples. We finetune LLMs on datasets...

The Cake Imagine that I want to bake a chocolate cake, and my sole goal in my entire lightcone and extended mathematical universe is to bake that cake. I care about nothing else. If the oven ends up a molten pile of metal ten minutes after the cake is done, if the leftover eggs are shattered and the leftover milk spilled, that's fine. Baking that cake is my terminal goal. In the process of ba...

This post offers an accessible model of psychology of character-trained LLMs like Claude. Epistemic Status This is primarily a phenomenological model based on extensive interactions with LLMs, particularly Claude. It's intentionally anthropomorphic in cases where I believe human psychological concepts lead to useful intuitions. Think of it as closer to psychology than neuroscience - the goal...

This is a link post. This is a blog post reporting some preliminary work from the Anthropic Alignment Science team, which might be of interest to researchers working actively in this space. We'd ask you to treat these results like those of a colleague sharing some thoughts or preliminary experiments at a lab meeting, rather than a mature paper. We report a demonstration of a form of Out-of-Co...

One hope for keeping existential risks low is to get AI companies to (successfully) make high-assurance safety cases: structured and auditable arguments that an AI system is very unlikely to result in existential risks given how it will be deployed.[1] Concretely, once AIs are quite powerful, high-assurance safety cases would require making a thorough argument that the level of (existential) risk...

Cross-posted from Telescopic Turnip As we all know, humans are terrible at building butterflies. We can make a lot of objectively cool things like nuclear reactors and microchips, but we still can't create a proper artificial insect that flies, feeds, and lays eggs that turn into more butterflies. That seems like evidence that butterflies are incredibly complex machines – certainly more compl...

This is a link post. A story I wrote about living through the transition to utopia. This is the one story that I've put the most time and effort into; it charts a course from the near future all the way to the distant stars. --- First published: January 19th, 2025 Source: https://www.lesswrong.com/posts/Rz4ijbeKgPAaedg3n/the-gentle-romance --- Narrated by TYPE III AUDIO .

This is a link post. Present alongside President Trump:  Sam Altman Larry Ellison (Oracle executive chairman and CTO) Masayoshi Son (Softbank CEO who believes he was born to realize ASI) President Trump: What we want to do is we want to keep [AI datacenters] in this country. China is a competitor and others are competitors. President Trump: I'm going to help a lot through emergency declaratio...

The AI Control Agenda, in its own words: … we argue that AI labs should ensure that powerful AIs are controlled. That is, labs should make sure that the safety measures they apply to their powerful models prevent unacceptably bad outcomes, even if the AIs are misaligned and intentionally try to subvert those safety measures. We think no fundamental research breakthroughs are required for labs to i...

I think a lot of people have heard so much about internalized prejudice and bias that they think they should ignore any bad vibes they get about a person that they can’t rationally explain. But if a person gives you a bad feeling, don’t ignore that. Both I and several others who I know have generally come to regret it if they’ve gotten a bad feeling about somebody and ignored it or rationalized it...

(Both characters are fictional, loosely inspired by various traits from various real people. Be careful about combining kratom and alcohol.) The original text contained 24 images which were described by AI. --- First published: January 7th, 2025 Source: https://www.lesswrong.com/posts/KfZ4H9EBLt8kbBARZ/fiction-comic-effective-altruism-and-rationality-meet-at-a --- Narrated by TYPE III AUDIO . ---...

From AI scientist to AI research fleet Research automation is here (1, 2, 3). We saw it coming and planned ahead, which puts us ahead of most (4, 5, 6). But that foresight also comes with a set of outdated expectations that are holding us back. In particular, research automation is not just about “aligning the first AI scientist”, it's also about the institution-building problem of coordinati...

So we want to align future AGIs. Ultimately we’d like to align them to human values, but in the shorter term we might start with other targets, like e.g. corrigibility. That problem description all makes sense on a hand-wavy intuitive level, but once we get concrete and dig into technical details… wait, what exactly is the goal again? When we say we want to “align AGI”, what does that mean? And wh...

Traditional economics thinking has two strong principles, each based on abundant historical data: Principle (A): No “lump of labor”: If human population goes up, there might be some wage drop in the very short term, because the demand curve for labor slopes down. But in the longer term, people will find new productive things to do, such that human labor will retain high value. Indeed, if anything,...

All quotes, unless otherwise marked, are Tolkien's words as printed in The Letters of J.R.R.Tolkien: Revised and Expanded Edition. All emphases mine. Machinery is Power is Evil Writing to his son Michael in the RAF: [here is] the tragedy and despair of all machinery laid bare. Unlike art which is content to create a new secondary world in the mind, it attempts to actualize desire, and so to c...

The anonymous review of The Anti-Politics Machine published on Astral Codex X focuses on a case study of a World Bank intervention in Lesotho, and tells a story about it: The World Bank staff drew reasonable-seeming conclusions from sparse data, and made well-intentioned recommendations on that basis. However, the recommended programs failed, due to factors that would have been revealed by a caref...