Hacker News Reader

Comment by lccerina | Hacker News Reader

> If such people were capable of being impressed by anything that happens in the empirical world, of updating on anything, they would’ve already been impressed and already updated several years ago, long before things had reached the point of an actual Mathocalypse.

I am honestly impressed by how many arguably highly intelligent humans are happy to throw out millennia of scientific methodology just because the world-burning probabilistic machine produced an output that tickles their dopamine brain.

Also glancing at the repo. Such an advanced machine, yet failing to output results with a common template. Every pdf is entirely different from the one before, making the work of the poor humans trying to find a sense in it even more difficult.

View on HN

> Also glancing at the repo. Such an advanced machine, yet failing to output results with a common template. Every pdf is entirely different from the one before, making the work of the poor humans trying to find a sense in it even more difficult.

I don’t think it helps your argument if this is half of your post. The fact that the pdfs don’t follow a common template? That’s the complaint? Human papers also don’t follow a template.

But human papers follow templates...? (Both literal templates based on editorial and org guidelines and structural and presentation templates in the form of traditions of the research area.)

They follow journal templates once they are accepted. Not for preprints. Which these essentially are.

> structural and presentation templates in the form of traditions of the research area

AFAIK, people are complaining about the lack of this.

"This ruins my life, so I'm going to be really petty!"

Plus, you know, you could use the AI to reformat it into any template you want. You could turn them all into anime music videos if you wanted. This is not a lack of capability, they didn't prompt it to use a template.

You sound like some useless bureaucrat. Oh noes, they didn't publish everything using the form XZC-152 template! And what scientific methodology? You want a scientific proof of when a threshold of where you can be impressed has been crossed? Ignoring the shifting goalposts, for me personally, AI is plenty impressive already by the mere fact that it can write a program that solves my problems from a natural language description. And "world-burning" how? Because water evaporates to create clouds? Using excess water treatment system capacity? Or maybe because they are using gas turbines to power it short-term instead of nuclear/renewables? That doesn't sound to me like an inherent flaw of the technology.

I just don't find particularly impressive that a probabilistic machine designed to write code, wrote code, based on the likely distillation of whatever input it got from math researchers blindly trusting OpenAI.

If the only way to power a "technology" would be to kill an elephant every time you use it, then it's a flawed technology even if it solved fundamental problems of humanity. This is currently how GenAI sprawling data centers operate AND without solving any fundamental problem of humanity.

Also I don't care about A specific template, I just find it ridiculous that some trillion dollar company cannot put a line like "please keep papers' structure consistent" in whatever markdown file controls these things. It's sloppy, like everything OpenAI does.

I guess the point of publishing breakthrough research is that by definition, it’s not just regurgitating human output, because there is no human output to these unsolved problems.

There was a bunch public literature on these topics. They were major unsolved problems. Lots of humans would want to have solved them and tried. But here’s it an llm that has solved them.

This comment reeks of the rapidly dwindling AI doomer rationalizing.

How much more scientific methodology do you want than formalized lean proofs?

What is this nonsensical complaining about templates and formats? You know that’s just like one additional ask of that particular LLM away.

https://news.ycombinator.com/item?id=50003107

It's already falling apart.

Reply on news.ycombinator.com