Leopold Aschenbrenner

A well reasoned article that fundamentally downplays both the pace of innovation and the exponential increase in capabilities per dollar happening over time. AI is rapidly accelerating its improvement rate, and I do believe its capabilities will continue growing exponentially. In particular, GPT-2 to GPT-4 spans an increase from 'well read toddler' to 'average high school student' in just a few years, while simultaneously the computational cost of training less capable models goes down similarly. Also worth noting: the article claims Stripe, another huge money raiser, had an obviously useful product. gdb, sometime-CTO of stripe and its fourth employee, is now president of OpenAI. And, most of all, the author doesn't remember how nonobvious Stripe's utility was in its early days, even in the tech scene: there were established ways to take people's money and it wasn't clear why Stripe had an offering worth switching to. For an alternate take, I think https://situational-awareness.ai provides a well reasoned argument for the current status of AI innovation and growth rate, and addresses all of the points here on a general (though not OpenAI specific) way.

Well it's not a random 25 year old. He became valedictorian at Columbia and graduated at 19. He was then briefly a part of OpenAI's superalignment team. After he got fired, he wrote a 165 page thesis on AI in early 2024 that all ended up becoming true in 2026; you can read it for yourself and be the judge: https://situational-awareness.ai That's what he based the fund on. It's generational foresight not available to 99.999999999999999999% of 25 year olds. I want to get ahead of the default cynicism and "oh it's just dumb luck woe is me" aspect of HN.

These types of posts are tediously regular. Aschenbrenner’s “Situational Awareness” in 2024 [1]. Fraudster Matt Schumer’s “Something Big is Happening” in February [2]. Next viral doom post in September-ish? [1] https://situational-awareness.ai/ [2] https://lowendbox.com/blog/ai-fraudster-matt-shumer-wrote-something-bi…

(For reference I’m talking about the DFT post from the same blog.) I love that ML is still in the “gentleman researcher” stage where relatively small amounts of startup capital can buy a ticket into frontier research. For a lot of research questions 6 GPUs is even overkill. It’s one of the reasons I’m skeptical of the “trillion dollar supercluster” idea [0]. I think what we need is more reasonably smart people investigating medium-sized problems. A “GPU middle class” you might say. [0] https://situational-awareness.ai/racing-to-the-trillion-dollar-cluster…

It's their own decisions they made long before the controls and presure. Besides being in bed with the US gov, people that run big AI shops tend to be fervently nationalistic and politically ambitious on their own. Leopold Aschenbrenner's dystopian rant [1] or Dario Amodei's [2] [3] are pretty representative. [1] https://situational-awareness.ai/ [2] https://www.darioamodei.com/essay/machines-of-loving-grace [3] https://www.darioamodei.com/post/on-deepseek-and-export-controls

I think ground-zero for that line of thought is with Leopold Aschenbrenner[0]. Who I believe now runs an AI focused hedge fund. [0] https://situational-awareness.ai

I wish I could agree with you. But have you read Aschenbrenner's "Situational Awareness" [1]? I am very much afraid that the big decision makers in AI do in fact think in those terms, and do not in any way frame this as fair competition for the benefit of all. 1. https://situational-awareness.ai/

This website by a former OpenAI employee was arguing that a combination of hardware scaling, algorithmic improvements, etc would all combine to yield AGI in the near future: https://situational-awareness.ai/

This article[1] does a good job of laying out the argument from trends and arguing that taking the trends seriously suggests reaching AGI in a few years. Admittedly, my own argument in favor of near-term AGI is much less sophisticated but is also largely based on extrapolating trends and noting the leaps in capabilities that can come from singular architectural innovations. The GPT family has been amazingly productive, and they result from largely a single architectural innovation. They have glaring deficiencies, but their failures are fairly well understood and overcoming them are basically engineering problems at this point. Whereas 10 years ago no one had any idea how to get from the SOTA at the time to AGI, I see a clear path to it from here. [1] https://situational-awareness.ai/from-gpt-4-to-agi/

For an opposing view, see Leopold Aschenbrenner’s recently released “Situational Awareness: The Decade Ahead”: https://situational-awareness.ai He did a very long interview with Dwarkesh Patel, too: https://www.youtube.com/watch?v=zdbVtZIn9IM

Read https://situational-awareness.ai/ If you can't develop a counter-argument to that, there's no reason to believe another AI winter is coming.

Good article, the METR metric is very interesting. See also Leopold Aschenbrenner's work in the same vein: https://situational-awareness.ai/from-gpt-4-to-agi/ IMO this approach ultimately asks the wrong question. Every exponential trend in history has eventually flattened out. Every. single. one. Two rabbits would create a population with a mass greater than the Earth in a couple of years if that trend continues indefinitely. The left hand side of a sigmoid curve looks exactly like exponential growth to the naked eye... until it nears the inflection point at t=0. The two curves can't be distinguished when you only have noisy data from t<0. A better question is, "When will the curve flatten out?" and that can only be addressed by looking outside the dataset for which constraints will eventually make growth impossible. For example, for Moore's law, we could examine as the quantum limits on how small a single transistor can be. You have to analyze the context, not just do the line fitting exercise. The only really interesting question in the long term is if it will level off at a level near, below, or above human intelligence. It doesn't matter much if that takes five years or fifty. Simply looking at lines that are currently going up and extending them off the right side of the page doesn't really get us any closer to answering that. We have to look at the fundamental constraints of our understanding and algorithms, independent of hardware. For example, hallucinations may…

See situational-awareness[1], see the "algorithmic efficiencies" section. He shows many examples of how models are getting cheaper. With many citations. Costs are not just down on a specific service. Even though I don't see the problem in that, as long as you get the promised level of performance, without being subsidised. See the deepseek model I linked above. It's an open model and you can run it yourself. > At best, older models are getting cheaper to run. What's your definition of old here? If you compare the literal bleeding edge model (o3) to 2 years ago best model (GPT-4)? Not only is this a ridiculously misleading comparison, it's not even valid! o3 is a reasoning model. It can spend money at test time to improve results. Previous models don't even have this capability. You can't look at one example of where they just threw a lot of money and say this is the cost. The cost is unbounded! If they want, they can just not let the model think for ages and have basically "0-thinking" outputs. This is what you use to compare models. If you compare _todays_ cost for training and inference of a model as good as GPT-4 when it was released, this cost has massively gone down on both counts. [1] - https://situational-awareness.ai/from-gpt-4-to-agi/#The_trends_in_deep…

> Who would have thought that vectorized linear algebra will be at the center of so much financial speculation? "vectorized linear algebra" is at the root of most of modern Physics. Specifically, the laws of Physics are represented by the Lie groups U(1), SU(2), SU(3) and SO(3,1). While the manifolds that Physics act on are curved, they're "locally flat". That is why local operations are tensor operations. Or linear algebra, if you prefer. It's not all that surprising to me that "intelligence" is represented by similar math. In fact, there is active work being done on making sense of deep learning using Lie algebra [1] (and Algebraic Topology, which generalizes the Lie algebra). This math can be a bit hard, though, so the learning curve can be steep. However, when we're creating AI models to be ML scientists, I suspect that this kind of math may be a source of "unhobbling", as meant in Situational Awareness [2]. Because if we can understand the symmetries at play in a problem domain, it's generally a lot easier to find a mathematical architecture (like in the algebras above) that effectively describe the domain, which allows us to potentially reduce the degrees of freedom by many OOM. > Heck, maybe we will even eventually see the famously absent productivity gains from digital tech? I think it's a mistake to think of AI as "digital tech", especially so to assume that the development of the Internet, Social Media or crypto that we've seen over the last…

If by “several” orders of magnitude, you mean 3-5, then we might be there by 2030 or earlier. https://situational-awareness.ai/from-gpt-4-to-agi/

Leopold Aschenbrenner predicted it last June. https://situational-awareness.ai/racing-to-the-trillion-dollar-cluster…

> What we have right now is impressive, but no one has shown anything close to a possible path from where we are right now to AGI[0]. [0]: From GPT-4 to AGI: Counting the OOMs https://situational-awareness.ai/from-gpt-4-to-agi/

This ex-OpenAI engineer’s recent essay makes the case that LLMs are progressing at 5x the rate of Moore’s Law across a number of dimensions: compute, efficiency, and capability: https://situational-awareness.ai/from-gpt-4-to-agi/

Important context, this appears to be a brief intro+table of contents for a 165-page pdf/html book.[0] It’ll probably be easy for HN to take potshots at snippets of the intro, but the writing intrigued me enough that I want to at least skim over the whole thing. [0] pdf link for reference, but the html navigation seems pretty clean in the OP link as well. https://situational-awareness.ai/wp-content/uploads/2024/06/situationa…

26 Jun 2025 18:15:45 UTC Situational Awareness: A One-Year Retrospective https://www.lesswrong.com/posts/EGGruXRxGQx6RQt8x/situational-awarenes… https://news.ycombinator.com/item?id=44389862 [ok] 30 Aug 2025 07:36:18 UTC Situational Awareness: The Decade Ahead (2024) https://situational-awareness.ai/ https://news.ycombinator.com/item?id=45072686 [ok] 12 Oct 2025 15:02:14 UTC Situational Awareness: The Decade Ahead https://situational-awareness.ai/ https://news.ycombinator.com/item?id=45558688 [ok] 29 Dec 2025 12:33:40 UTC Situational Awarness The Decade Ahead (2024) [pdf] https://situational-awareness.ai/wp-content/uploads/2024/06/situationa… https://news.ycombinator.com/item?id=46420030 [ok] 03 Mar 2026 13:19:26 UTC Situational Awareness: the decade ahead (2024) https://situational-awareness.ai/ https://news.ycombinator.com/item?id=47231897 [ok] 05 Mar 2026 15:19:17 UTC Situational Awareness (2024) https://situational-awareness.ai/ https://news.ycombinator.com/item?id=47262572 [ok] 10 May 2026 02:47:06 UTC Situational Awareness Kindle eBook ePub https://blog.cahillanelabs.com/kindle/epub/llm/2026/05/10/situational-… https://news.ycombinator.com/item?id=48080509 [ok] 18 May 2026 19:05:28 UTC Situational Awareness LP Q1 13F Filings Are Public https://kiankyars.github.io/investing/2026/05/18/situational-awareness… https://news.ycombinator.com/item?id=48184085 [ok] 22 May 2026…

This one ? https://situational-awareness.ai/parting-thoughts/ There are some big differences with the atom bomb though : A lot of people have been talking about AGI / the Singularity for decades, and that essay is yet another example. Not sure if it's the hindsight talking, but the situation with (un)controlling fission was much clearer ? We knew it was at least theoretically possible. And the first test was very precisely engineered (in fact switching at the last minute to a more complex design because the calculations showed that the material wasn't pure enough to work). And the very first test worked. Meanwhile, despite for AGI it (almost) all being out in the open, nobody can define what AGI exactly is, and it's all extremely empirical : throw compute at the wall, try another algorithm, see if the black box seems smarter, or not. Nobody knows even in theory when (or even if) AGI will be reached, and for how long scaling up compute will keep providing the very impressive results we have seen so far.