405error

12 ms·

1.
by 405error 7d ago
But consider they could decide that they want to scoop more regular research work too. They could automate it with just a few LoC. Even if you opted out in the ToS, you'd have to file a massive lawsuit just to enforce it. And the act
2.
by 405error 7d ago
Personally this is a watershed moment for researchers and grad students I know. All of them are close sourcing WIP repos, not putting their progress in LLMs, or have lab level initiatives to self host models.
3.
by 405error 8d ago
I can give you some context. 1. Terence Tao's mastodon explains the way this problem was solved does not in itself contribute much. LLMs (and in this case) produce massive, often unintelligible proofs that do not further understanding.
4.
by 405error 8d ago
We know they are training on private chats. It's listed in the ToS.
5.
by 405error 8d ago
Tristian's allegations are much more serious than academic slap-fighting. If what he suggests is true, every academic using AI is going to get scooped. Yes AI can do non-trivial work, but the situation is that you could be a PhD stud
6.
by 405error 8d ago
Many academics and grad students I know have closed source their in progress work, and started being really careful about what they chat with LLMs (or using local ones) because of the drama around this. No one wants four years of their life
7.
by 405error 8d ago
It would not be difficult to write a pipeline to remove 99% of low quality posts, especially about specific subjects. It would be very easy to identify accounts as researchers based on their chat logs.
8.
by 405error 13d ago
It's probably AI coded and hallucinated many things. That drumming up the importance of a minor thing is a real tell. Another hallucination - it hasn't found any video APIs (despite statements that it has and uploaded it). It has
9.
by 405error 13d ago
More directly, there simply aren't any video files uploaded. It's just parquet files, which contain no video columns (I'm not even sure if it supports it).
10.
by 405error 13d ago
It's that mix of dense, impressive sounding jargon, but even scanning across it raises glaring problems. Like, if you have 4.5 billion videos on HF, and it's 289GB, it's about 60 bytes per video. Checking the column fields as
11.
by 405error 13d ago
The first immediate smell is that if you have 4.5B rows and 289GB in data, you have ~60 bytes per row.
12.
by 405error 13d ago
I cannot verify whether it is technically correct, but it's about how to defeat Tiktok's bot filters to scrape it.
13.
by 405error 23d ago
I don't agree with some things Anthropic does but that is a cool thing.
14.
by 405error 29d ago
I have a vague recollection of a point someone made before that most ideas are usually wrong or incorrect at the start. For example, Nvidia was built on the thesis that non-triangular polygons would dominate, which was 100% wrong (today all
15.
by 405error 29d ago
Aren't several cryptocurrencies built on ZKPs? Their business model aside, ZKPs do look like one of the rare examples of theoretical elegance and real world use (even if not widespread).