This is written with the idea that the exponential part keeps going forever.

It never does. The progress curve always looks sigmoidal.

- The beginning looks like a hockey stick, and people get excited. The assumption is that the growth party will never stop.

- You start to hit something that inherently limits the exponential growth and growth starts to be linear. It still kinda looks exponential and the people that want the party to keep growing will keep the hype up.

- Eventually you saturate something and the curve turns over. At this point it’s obvious to all but the most dedicated party-goers.

I don’t know where we are on the LLM curve, but I would guess we’re in the linear part. Which might keep going for a while. Or maybe it turns over this year. No one knows. But the party won’t go on forever; it never does.

I think Cal Newport’s piece [0] is far more realistic:

> But for now, I want to emphasize a broader point: I’m hoping 2026 will be the year we stop caring about what people believe AI might do, and instead start reacting to its real, present capabilities.

[0] Discussed here: https://news.ycombinator.com/item?id=46505735

Written by an anthropic employee, I mean how seriously can you take that piece? pure hype

I think it's coupled differential equations where each growth factor amplifies the others, I posted about it in 2024 - https://b.h4x.zip/ce/ - sent it around a bit but everyone thought I was nuts, look at that post from 2025 and think about what was happening IRL under the graphs line, then go look at where METR is today. I'm not trying to brag, I don't work for anthropic, but I do think I'm probably right.

Access blocked, suspicious site.

Firewall blocking it? Weird, dns is all fine.

I only take it partially seriously. I view it as a serious presentation that is misinformed. What I find unique is that people have become so interested in "the exponential" it's almost become like an axiom, or even a near religious belief in AI. It is a subtle admission that while current AI capabilities are impressive, it requires additional years of exponential growth for AI to reach the fantastic claims some people are making.

All glory to the exponential!

> Given consistent trends of exponential performance improvements over many years and across many industries, it would be extremely surprising if these improvements suddenly stopped.

This is the part I find very strange. Let's table the problems with METR [1], just noting that benchmarking AI is extremely hard and METR's methodology is not gospel just because METR's "sole purpose is to study AI capabilities". (That is not a good way to evaluate research!)

Taking whatever idealized metric you want, at some point it has to level off. That's almost trivially true: everyone should agree that unrestricted exponential growth forever is impossible, if only for the eventual heat death of the universe. That makes the question when, and not if. When do external forces dominate whatever positive feedback loops were causing the original growth? In AI, those positive feedback loops include increased funding, increased research attention and human capital, increased focus on AI-friendly hardware, and many others, including perhaps some small element of AI itself assisting the research process that could become more relevant in the future.

These positive feedback loops have happened many times, and they often do experience quite sharp level-offs as some external factor kicks in. Commercial aircraft speeds experienced a very sharp increase until they leveled off. Many companies grow very rapidly at first and then level off. Pandemics grow exponentially at first before revealing their logistic behavior. Scientific progress often follows a similar trajectory: a promising field emerges, significant increased attention brings a bevy of discoveries, and as the low-hanging fruit is picked the cost of additional breakthroughs surges and whatever fundamental limitations the approach has reveal themselves.

It's not "extremely surprising" that COVID did not infect a trillion people, even though there are some extremely sharp exponentials you can find looking at the first spread in new areas. It isn't extremely surprising that I don't book flights at Mach 3, or that Moore's Law was not an ironclad law of the universe.

Does that mean the entire field will stop making any sort of progress? Of course not. But any analysis that fundamentally boils down to taking a (deeply flawed) graph and drawing a line through it and simplifying the whole field of AI research to "line go up" is not going to give you well-founded predictions for the future.

A much more fruitful line of analysis, in my view, is to focus on the actual conditions and build a reasonable model of AI progress that includes current data while building in estimations of sigmoidal behavior. Does training scaling continue forever? Probably not, given the problems with e.g., GPT-4.5 and the limited amount of quality non-synthetic training data. It's reasonable to expect synthetic training data to work better over time, and it's also reasonable to expect the next generation of hardware to also enable an additional couple orders of magnitude. Beyond that, especially if the money runs out, it seems like scaling will hit a pretty hard wall barring exceptional progress. Is inference hardware going to get better enough that drastically increased token outputs and parallelism won't matter? Probably not, but you can definitely forecast continued hardware improvements to some degree. What might a new architectural paradigm be for AI, and would that have significant improvements over current methodology? To what degree is existing AI deployment increasing the amount of useful data for AI training? What parts of the AI improvement cycle rely on real-world tasks that might fundamentally limit progress?

That's what the discussion should be, not reposting METR for the millionth time and saying "line go up" the way people do about Bitcoin.

[1] https://www.transformernews.ai/p/against-the-metr-graph-codi...

"everyone should agree that unrestricted exponential growth forever is impossible, if only for the eventual heat death of the universe." - why is this a good/useful framing?

All models are wrong; some are useful. Cognizance of that is even more critical for a model like exponential growth that often leads to extremely poor predictions quickly if uncritically extrapolated.

I think "are the failures of a simple linear regression on the METR graph relevant" is a much better framing than "does seeing a line if you squint extrapolate forever." As I said, I'd much rather frame the discussion around the actual material conditions of AI progress, but if you are going to be drawing lines I'd at least want to start by acknowledging that no such model will be perfect.

Are they failing to understand, or are they manipulating their audience?

They are manipulating and fear mongering gullible people, it's their new business model.

I'm friends with one of the Anthropic founders for over 15 years now, and I just find this line of thinking so sad. They are not manipulative fear mongering people, they're actually very decent people who you might consider listening to.

If that were true, they wouldn't publish hype results that then turn out to be completely unsubstantiated. Remember the "agents built a web browser"? I can't personally judge your friend as I don't know him. But the company is consistently lying about how good their product is in order to hype it up.

I don't talk to said friend about their work, so I genuinely have no insight here, but if I were a betting man, I'd bet what they have internally is considerably disparate from what is currently available in their consumer product.

The stuff they have internally might be slightly better than what they have now lmao. You have to super dense to believe otherwise.

Also I don't need the Anthropic ghouls telling me what I can or can not ask their stupid bot. At least Elon doesn't play this sad censorship game where you cannot say "boob" to it without it locking down.

Yeah and my dad works at Nintendo. If they want us to listen they really need to stop with releasing all the bullshit and over exaggeration what their chat bot does. And stop freaking whining about "MUHHH CHINA". Those ghouls stole almost all the books in the world, I hope China steals all from them and keeps releasing the free models.

Well, if your dad isn't one of the founder of Nintendo your point is moot. Given I was on the founding team of digitalocean as head of strategy till the IPO, and one of the founders of anthropic is a former tech journalist who covered my startups, maybe my friend is a founder of Anthropic?! Sorry your dad didn't work anywhere cool tho. :(

Who cares about these journalists. My point is that Amodei is a complete ghoul and loves fear mongering normies and being racist towards Chinese while HIS company stole all the damn books in the world. I hope the Chinese steal all their data and keep doing public models. This guy can't even figure a solution for his balding head let alone making an AGI lmao. But let Anthropic keep tricking midwits along. Elon has 100x the backbone that these fraudsters have btw.

For God's sake these guys are selling the doubling of human life span to some desperate elderly investors. Really going for people's deepest fears there. Oh yes just invest in us so you can get double the life span and don't have to die!

Alright well I can tell you're grumpy about this so how about we agree to disagree? I don't know Dario so I couldn't say, but I do trust Jack a lot. That aside: this is the 3rd time I've heard the racist towards Chinese thing, what exactly is that all about if you'd be willing to save me a google?