Amateur armed with ChatGPT solves an Erdős problem

It is wildly not true.

The request is for some reasonable math problem a model like GPT or Claude will fail at. I'm not going to set up a local model or some harness for it; I'm just going to copy/paste it into ChatGPT and watch it solve it.

Propose a problem, if you think I'm wrong about this. Seems simple.

Why be such an absolutist.

How about I caveat it the way you want:

AI equalizes intelligence in the sense that it closes the gap. Not perfectly, not infinitely, but directionally. The distribution compresses. The floor rises faster than the ceiling, so people who used to be far apart end up operating much closer together.

You can already see it in the Erdős example. The person who wrote that prompt wasn’t some random idiot. It took real cleverness to even set it up that way. But the fact that they could get that far, with assistance, is exactly the point. The distance between “amateur” and “expert” shrinks when the tool fills in large parts of the path.

Now extend that forward. Today it’s one clever person, one problem, one careful interaction. As the tooling improves, that same pattern scales. Better reasoning, better search, better guidance. The amount of lift the tool provides increases, which means the gap continues to narrow.

All the supposed “counterpoints” people bring up are already implied in the claim. “Equalize” here obviously means moving closer to equality. Is it NOT obvious that LLMs don't actually equalize intelligence to a level of 100%? Do I actually need to spell that out? If there was nothing at stake, I wouldn't need to.

But instead people latch onto the most absurd version possible, knock that down, and act like they’ve said something meaningful. It’s the same mindset as that guy demanding a formal paper or citation for an observation you can see unfolding in real time. Not because it’s unclear, but because engaging with the actual claim is uncomfortable. It’s easier to distort it into something extreme and dismiss it than to admit the gap is closing.

None of this makes any sense. I'm not in school and I'm not a teacher. It's just a random attempted drive-by dunk that faceplanted.

That's not my point. I'm talking about something far more mundane - transformers do inference over raw tokens and perform an n^2 loop over tokens, but tokens are itself the context. So it's better to have more raw tokens in your input that all nudge it to the right idea space, even if technically it doesn't need all those tokens. ICL and CoT have a lot of study into them at this point, these are well known phenomena.

This applies to any transformer-based architecture including JEPA which tries to make the tokens predict some kind of latent space (in which I've separately heard arguments as to why the two are equivalent, but that's a different discussion.)

His core issue is jealousy and fear. I don't think these types of people are at the top of the intelligence curve (more closer to bottom) but that is orthogonal to my point. What I'm saying is his personality archetype makes him think (keyword) he's at the top of the intelligence curve and an equalization means, personally to him, that he's losing his edge.

More specific to HN is the archetype of: "I have spent years honing my craft as a expert programmer, my identity is predicated on being an expert programmer in which high intelligence is causal and associated positively with my identity" That's why ironically most of HN was completely wrong about AI. They were wrong about driverless cars, they claimed vibe coding was trash. It's the people who think (keyword) their stupid/average (aka general public) who got it right... because perceptually they stand to gain from the equalization.

Anyway.. this fear and jealousy is not something most humans can admit to themselves. Nobody will actually be able to realize that these emotions drive there thinking. They have to lie to themselves and rationalize a different reality. That's why you get absurdist takes like this.

To everyone reading. It is obviously that chatGPT does not equalize intelligence to the point of 100%. That statement is obviously not saying that. Everyone knows this. You want proof?:

Look at the declaration of independence... without getting to pedantic: "All Men are created equal" is not saying all Males are 100% equal. Everyone knows this. First off no one is 100% equal.. and second the statement in a modern context is obviously not referring to only men. It is referring to women&men and clearly men and women are nowhere near equal.

So if you all know this about the declaration of independence... how can you not see the same nuance for: "ChatGPT equalizes intelligence."? First ask yourself... do you think you're smart? If you do, then the self delusion I just described is likely happening with you.