Despite this election, I remain an optimist about America and the world. Humanity will continue to produce new ideas and new innovations to improve our lives. Good people will continue to come together to improve the world. And the political tide will turn. We'll make it so.

Begun, the robot wars have.

I think pacing the frontier is good as an idea in principle but dead on arrival as something you can actually do. What will we, decide which benchmarks we can improve on? Who is involved? If we stop pushing "capabilities" we'll do a ton of research on swarms + efficiency which will bring on other risks. It distracts from how so much of AI risks is simply coming from diffusing what we have, and we should invest in preparedness, pushing the labs to be more careful, etc. Change my mind?

The goal on climate is to get to zero (& then negative) as fast as we can, globally. Supply-side restrictions often don’t work (demand is met, often dirtier) + they ignore problem is global. If you want stranded fossil assets, you beat them on cost and deploy them everywhere.

Gen Z boss and a mini

Craziest thing I saw yet: GLM-5.3 *flash* can exceed Mythos Preview's scores on ExploitBench. Not exactly surprising though. I've been saying for a while that ultimately, TTC scaling destroys categorical capability tiers. Models are now AGI enough to Just keep making progress.

Jason's read too much science fiction.

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks: Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better: Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better: Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better: Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work! In summary: - As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding. - Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

Actual climate scientist here. The permitting reform bill was introduced by Senate climate champions like @SenWhitehouse. It will reduce US emissions. We don't have the luxury to make the perfect the enemy of the good – which time and time again has led to doing nothing at all.

To be quite frank, a lot (not all!) of these “rogue agent hacks” on government statistical websites are things that think-tank interns and research assistants have done for many years. I have known many a think tank paper that was enriched by an enterprising RA using eg urlquery to find CSVs and PDFs on government agency websites that were publicly accessible and non-sensitive but nonetheless not things the agency *intended* to be on their website. Is it “hacking” to find those things? Is it hacking in the even more innocuous examples where agents are literally just accessing tabular data available on a government statistical website, just a version of that data that is easier to access and analyze programmatically? This constant trickle of examples (which tbc I have personally seen many models do, before I joined OpenAI; it is even something I have discussed with models before!) will probably have the effect of diminishing the importance of “an AI hack” on the eyes of the observing public by creating an oversupply of “AI hack” examples. OAI-HF was an example of AI hacking. Many of these more recent examples, very candidly, are straining the definition of the word “hack” and, I worry, cheapen the non-technical public’s understanding of that concept.

Absolutely fantastic to see. Transparency and openness will do more to advance AI safety, equality, and democracy than secrecy. Thrilled to see @natolambert and @thesezickbeats launching this.

Elasticity of demand predicts enthusiasm for AI cost cuts. Art has inelastic demand, so artists are opposed. High math elasticity is neutral, so those mathematicians feel middling. Software demand is very elastic, so software folks are enthusiastic.

RT @curiouswavefn: Just in the last six months I’ve used AI to: - Design antibodies with better predicted binding affinity - Design protei…

Concerns about AI fall, roughly, into two classes. The first comprises conventional risks: disinformation, discrimination, surveillance, fraud, cyberattacks, job displacement and concentration of economic power. AI can indeed generate convincing disinformation, but, as the philosopher @danwilliamsphil has argued, political persuasion is extremely hard; and insofar as AI does influence public attitudes, its overall effect may be to strengthen the influence of expert opinion. @AndyMasley has shown in a detailed analysis that data-centre water use is minuscule compared with that of other sectors. The second comprises existential risks: the possibility that superintelligent AI systems, pursuing goals that diverge from ours, might act in ways that lead to human extinction or permanent disempowerment. Once associated with @allTheYud, Nick Bostrom and rationalist blogs, these fears are now taken seriously by many leading AI figures. They attracted global media attention in early September when AI researcher Jacob Coxon sensationally resigned from Anthropic, saying that the people building AI “earnestly believe that it could kill us all by the end of the decade”. Anthropic’s alignment science lead, @EvanHub, agreed, putting his own estimate of the chance of AI killing all humans at more than 10 per cent. In a recent article, @clairlemon compared AI-doom fears to past apocalyptic scenarios involving overpopulation, nuclear annihilation, and environmental collapse. She rightly noted that technophobic alarmism can prompt restrictions that do great damage to human progress. However, unlike past doomsday prophets, AI-doomers are highly technophilic, politically diverse and genuine experts in the systems they fear; several have also forecast AI’s development with unusual accuracy. They should be taken seriously. The detailed AI 2027 scenario, published in April 2025 by @slatestarcodex and leading AI thinkers from the AI Futures Project, describes a rapid rise in AI capabilities ending either in human extinction or a technofeudal regime controlled by oligarchs. Yet both outcomes depend on rapid diffusion: AI spreading through government, the military and society, controlling factories that manufacture robots and drones, while powerful political figures cede judgment and decision-making to it. The last point matters because an AI cannot exterminate an embodied creature by pure thought: it needs direct or indirect control over physical objects. Alexander has also written elsewhere about the “diffusion gap”: the time between AI becoming capable of doing 90 per cent of knowledge-work jobs and AI actually doing even half of them. Diffusion is hard. But there may also be a deeper source of friction: the powerful human urge for meaning and mattering. @platobooktour argues that the need to matter is our deepest longing and the crux of the human experience. This instinct is likely to work against surrendering control of meaningful parts of our lives to machines. The reaction of a section of the mathematical community to recent AI breakthroughs is suggestive. Nor should we assume that persuasive power scales indefinitely with intelligence: being vastly better at reasoning does not guarantee an ability to persuade humans to give up control. I suspect many AI-doom scenarios underestimate the social and institutional resistance that large-scale displacement of human agency would provoke. Handing AI control of military systems, drone factories or critical infrastructure is likely to meet far greater opposition than adopting systems to help run them. Halting development could mean forgoing extraordinary benefits. In AI 2040: Plan A, the AI Futures Project sketches a path to preserve those benefits while sharply reducing existential risk. Scott Alexander envisages triple-digit GDP growth in the 2030s, putting the stakes starkly: “It’s nanobots eating the solar system in 2033 vs. cancer cures in 2035”. Nor would GDP hypergrowth or spectacular scientific breakthroughs capture all of AI’s benefits. In a recent post on X, @DavidDeutschOxf described how AI helped him fix a dishwasher he was about to replace. Widespread access to knowledge and practical competence may reduce measured GDP while still making people better off. Some AI-safety researchers favour pausing frontier development until we have “solved alignment”; others such as @sebkrier, @mboudry and Simon Friederich instead emphasize iterative feedback, selection and pressures favouring more controllable AI. Recently, Dario Amodei, Anthropic’s CEO, argued for pacing the frontier, rather than pausing AI development or racing ahead. This is a sensible approach. There are also pitfalls to avoid. In "If Anyone Builds It, Everyone Dies", @ESYudkowsky and @So8res advocate a global prohibition on publishing new AI research. This would eviscerate freedom of speech and academic freedom. Yudkowsky has elsewhere advocated airstrikes on rogue data centres to enforce an AI moratorium, even at the risk of nuclear war. More broadly, some of the more sweeping conclusions of AI-doomerism seem to pay insufficient attention to the epistemic and political norms that help societies avoid mistakes, the realities of human psychology, and the relationship between intelligence and power. AI is a revolutionary technology. In some fields, including my own, it may soon become an extraordinarily intelligent oracle. But intelligence is not sovereignty. A machine’s ability to think more powerfully than we can does not require us to surrender what we choose to pursue, build, control or understand. On the contrary, AI could enormously expand what humans can know and do. To realize this promise, we need to protect free inquiry, remain open to empirical evidence, and be willing to solve difficult problems. Read my whole piece at Café Américain: cafeamericainmag.com/against…

desperately hoping my gf becomes powerful enough one day that I, too, am featured in a tiny little "husband" box in an article

If you say you care about climate change, but are against tools to effectively build transmission infrastructure, you’re either ignorant or don’t actually care about climate action.

We are launching a new newsletter today: Understanding Robots. Kai wrote the inaugural piece about Astra's shockingly good performance controlling robots. understandingrobots.org/p/op…

This is a huge problem for aging countries, i.e. all of the West. via @cowenconvos

One fifth of US counties (17% of land mass) now restrict clean energy development via local ordinance. heatmap.news/politics/laws-r…

This is one of several startups using robotics and standardized designs to vastly accelerate installation of utility-scale solar. Modules are so cheap, they're only a fraction of solar capex now. Attacking installation costs is a huge lever. Innovation to make clean energy cheaper hasn't stopped.

This graph should sink home for all of us. And keep in mind: 2010-2018 was a period when solar and wind growth was far slower than it is now. Solar was a miniscule part of new electricity capacity in this period, for example, yet a major share of all federal studies under the National Environmental Policy Act. That's insane.