JohnKemeny

> I have the same optimism as Prof Tao.

This is not written by Tao, but by Álvaro Lozano-Robledo.

When the dust settles, the housing bubble of '08 will look like chump change.

Retracted 3 results. Found errors in 14 others.

Beautifully written. Beautiful game. I never completed the game, but it has stuck with me since 2015.

They are not publishing Lean proofs. They are publishing proofs in natural language, and are not submitting to journals.

They are just putting out a bunch of weirdly written extremely long and technical papers and saying: Hey, here is the solution (we hope there are no mistakes).

10% is ridiculously high compared with published papers in math journals where the number of critical mistakes is near zero.

This doesn't seem to test the GC, nor important operations such as dict lookup/iteration, set add/membership, list iteration and updating.

Is really a recursive implementation of Fib an enlightening benchmark? I don't think so. And bubblesort is just messing around with two pointers.

Why not, if you're first doing benchmarking, find out what the most time consuming popular operations/algorithms are, and then use them?

Now it seems that you are just testing two obscure algorithms that are not really representative for Python coding in general.

Many people thought it could never be less than 2. They proved that it can. What is the true value? Nobody knows, now.

You can use Grammarly via an API, or simply submit text in a web interface.

It's a keylogger in the same way Vim is a keylogger.

Are you saying that modern LLMs cannot play chess now, or that LLMs (GPT architecture) cannot be trained to play chess well?

Or are you saying that neural networks in general cannot (practically) be trained to be an above-average chess player?

Or are you saying that it depends on the input? Would it be better if they were given a picture/drawing/ascii art of the board? If so, surely they can produce it at will?

Your link 404s

> The main branch of formal-conjectures does not contain the path `FormalConjectures/Millenium/NavierStokes.lean.`

> The second author, Elias Koutsoupias, dedicates this work to his constant friends Amos Fiat, Anna Karlin, and Christos Papadimitriou.

I wonder what Papadimitriou thinks about getting dedicated LLM generated proofs.

It's highly non-trivial to confirm that the theorems written in Lean are actually the same as the Navier–Stokes (non-)theorem.

It's one of the seven most important problems in mathematics.

Pick a better analogy.

Yes, that's a strange mistake to make.

> but only if Alpöge’s name was removed

This is blatant scientific misconduct.

Not trying to answer on OP's behalf, but perhaps they didn't mean blockchain in the literal sense, but more like a DAG of theorems that are shared globally?

I probably didn't understand what you meant by

> Hutter Prize being where you are paid if you can compress wikipedia small enough. LLMs do very well at that, if, big if, you ignore the cost of initial weights.

then.

If all you care about is compressing Wikipedia, but ignore the size of the actual data, what is it that you are actually trying to do?

> ignore the cost of initial weights

Well, then Wikipedia itself is a very good compression that only needs the title to perfectly predict the full article.

Maximum flow isn't NP-hard, though. It can be solved in near-linear time.

If you allow twice the solution, you can do it in O(m log n) time using MST.

You can't do it faster than O(n log n) for the simple reason that you need to sort the frequencies. If the symbols come sorted, then you can do it in O(n) time, yes, using two queues.

I think the phrase you're looking for is the phase transition.

Discussed here: https://cstheory.stackexchange.com/questions/33550

True, but what you can do is a one-sided guarantee. If it bears the mark, it is likely generated (or someone deliberately made it look generated).

Thus, if a news article, research article, book, student paper submission, blog post, HN comment, etc, bears the mark, it could be automatically flagged as such.

It helps detect low effort slop.

---

Caveat. If you write your own creative work and send it to Claude for "cleaning up grammar", it might insert the watermark.

You don't have to "revive" the device in order to use it.

The problem was: out-of-sync device clock prevented software updates which in turn prevented cloud syncing. You don't need software updates or cloud syncing for the device to work.

The fix is to set the clock and do the software updates.

OCR has definitely not "been solved long time ago", what are you talking about?

In your opinion, what is SOTA here?

Related: An Introduction to YOLO26

77 points, 24 comments

https://news.ycombinator.com/item?id=48639165

Follow-up. I in fact suspect that either she is a bot, or she is using an LLM to spew out papers. She uploads about one paper per month to Zenodo, and they all seem very AI generated.

The paper under discussion:

https://zenodo.org/records/17720178

Note that Zenodo is a DOI-provider, not a (scientific) journal. Anyone can upload anything to Zenodo. It's less strict than arXiv.

Edit: The "paper" is written by one Hiroko Konishi, an independent researcher (she is a voice actress).

yeah, only authorities are considered here at HN.