Comment by killerstorm
11 hours ago
This is literally an opinion of one dude which is not backed by any kind of quantitative evidence.
It's actually possible to answer this question rigorously:
1. Define a scientific result which qualifies as a "jump". They should be frequent enough that they happen every year - otherwise one might say humans can't jump either.
2. Identify all such "jumps" in articles published in 2026, and use LLM with 2025 knowledge cut-off to re-derive these results with minimal amount of information.
It really irks me that people boost these low-effort articles just because they confirm pre-conceived notion that LLMs are limited
You don't need any reproduction. Assuming researchers are using LLMs, you should just see the number of jumps increase as models get better.
I boost it so someone else will see it and prove it wrong.
LLMs most definitely are limited. What’s your position, that LLMs have no limits? That’s obviously wrong, and the fact you hold such an unreasonable position may be why you react so strongly to pieces like this one.
What's the obvious limitation?
To Yann LeCun it was "obvious" that you can't build a world model from a text, so he predicted that even "GPT 5000" can't predict that an object sitting on the table would move together with the table. But GPT-4 could solve this task and provided correct explanation. GPT-5 can solve much more complex tasks, code physics simulation, etc.
Now people talk about invention of General Relativity as something LLMs can't do... something almost no human was able to do either.
You're hallucinating facts again.