skinner_

7 ms·

31.
by skinner_ 2y ago
> in such a way that could be construed as reneging on your promise Nothing could be construed in such a way, because such a reneging is not possible.
32.
by skinner_ 2y ago
Here is my prediction, FWIW: the hard part of the problem has already been solved, in the following technical sense: there is a few 1000 lines program that has not been invented yet, but it will be invented soon, that loads a current LLM mo
33.
by skinner_ 2y ago
Interesting, as a very frequent HN reader, I do not associate this pattern with Python zealots at all. Rust, Swift, Go, more like.
34.
by skinner_ 2y ago
Bingo. That was not obvious to me before your one-word reply, and now, in retrospect, it is. And I say this as a person who usually just rolls his eyes seeing the typical cynical HN hot takes. I believe Altman voices his deepest convictions
35.
by skinner_ 2y ago
Neither the author of the GPT from scratch post, nor eclectic29 who recommended it above did ever promise that the post is about building LLMs from the ground up. That was the original post. The GPT from scratch post explains, from the grou
36.
by skinner_ 2y ago
> Your resource is really bad. What a bad take. That resource is awesome. Sure, it is about inference, not training, but why is that a bad thing?
37.
by skinner_ 2y ago
Thank you, that's a perfect follow-up piece, makes the whole thing much clearer.
38.
by skinner_ 2y ago
That's my understanding as well, FWIW. This is how I would phrase it: Shrinking helps. In R^d there's no such thing as shrinking generally, only shrinking in the direction of some point. (The point that's the fixed point of t
39.
by skinner_ 2y ago
Awesome, thank you! A seal of approval!
40.
by skinner_ 2y ago
Very good question! Yes, the final value is somewhere in there. I expect that it's the lowest point, meaning Croft's Tortoise is optimal. Since then we have an unpublished new 0.241 upper bound with the same proof but wider comput
41.
by skinner_ 2y ago
But if you pack the same atoms in a cube, that's 17 centimeters. ((8e9*7e17)^(1/3)*1e-10=0.17)
42.
by skinner_ 2y ago
The question of what is an Erdős problem and what is not is not black and white. My team has managed to solve a problem that's sort of an Erdős problem. In the early sixties, Leo Moser asked about the value of some quantity. In a 1985
43.
by skinner_ 3y ago
> For instance, an autonomous driving system that doesn't know the first thing about the zombie apocalypse could put its passengers' lives at risk by refusing to run over "pedestrians". I'd prefer my self-driving
44.
by skinner_ 3y ago
There's this nice new paper ("Simplifying Transformer Blocks" https://arxiv.org/abs/2311.01906 ) about getting rid of all the unnecessary parts of a transformer. On their dataset, they manage to get rid o
45.
by skinner_ 3y ago
Very nice explanation! I think it's worth mentioning that this is called universal search (you mention this in passing) or Levin Search or Levin's universal search algorithm. It was invented by Leonid Levin in 1972. https:/&
46.
by skinner_ 3y ago
The fractional chromatic number of the plane is by definition less than or equal to the regular chromatic number of the plane, but already we know for sure that they are not equal. We know that 5 <= X(R^2) <= 7, and we know that 3.899
47.
by skinner_ 3y ago
Sorry, I've just spotted your question 3 days late. The Moser-Erdős question is about how large the density of a unit distance avoiding set can be. The Hadwiger-Nelson question is about how many unit distance avoiding sets are necessar
48.
by skinner_ 3y ago
(Submitter here, but also one of the authors of the presented work.) The Quanta article rushed through some aspects of our work, so I'll try to use this opportunity to elaborate. This has a very small target audience: those who have sk
49.
by skinner_ 3y ago
Not really. That is concerned with much less rigid planar structures. What this is directly related to is the famous unsolved https://en.wikipedia.org/wiki/Hadwiger%E2%80%93Nelson_proble... , the number of colors neede
50.
Mathematicians Solve Long-Standing Coloring Problem(quantamagazine.org)
41 points by skinner_ 3y ago|8 comments
51.
by skinner_ 3y ago
I understand that the people who claim this don't provide any evidence. But do you have any pointers for the claim that it is not true?
52.
by skinner_ 3y ago
Current LLMs can go crazy just by seeing a single well-chosen word, see https://www.lesswrong.com/posts/aPeJE8bSo6rAFoLqg/solidgoldm... for an amusing exanple. I think we need some breakthrough in adversarial trai
53.
by skinner_ 3y ago
My Macbook Pro's webcam went wrong in 2020. I haven't had it fixed yet, because I hate Zoom video. Nobody ever complains.
54.
xAI, Elon Musk's new AI company announced(wired.com)
2 points by skinner_ 3y ago|0 comments
55.
by skinner_ 3y ago
I was taught that choosing a QM interpretation is a matter of taste. Am I understanding correctly that we do have proposed quantum gravity experiments that can falsify various QM interpretations, it's just that they are all very hard t
56.
by skinner_ 3y ago
I believe you are misparsing that sentence. The dogma and the tradition are the same thing.
57.
by skinner_ 3y ago
This effort is led by Ilya Sutskever. Listening to a bunch of interviews with him, and talking to a bunch of people who know him personally, I don't think he cares at all about AIs saying naughty words.
58.
by skinner_ 3y ago
Tegmark has his name on a bunch of competently written AI papers: https://arxiv.org/search/cs?searchtype=author&query=Tegmark%... In each case, the first author is Ziming Liu, which is more impressive than Tegmark
59.
by skinner_ 3y ago
Thanks! What a self-own, ICLR Spotlight paper, that's close to the top of the reputability food chain!
60.
by skinner_ 3y ago
Alphafold uses something they call Evoformer, it is an attention mechanism. Our group has tried, and so far failed to utilize transformers for a very very specific search problem in geometry ( https://bit.ly/unit-distances ).

More ›