viscanti

10 ms·

61.
by viscanti 3y ago
I guess it depends what you mean by LK-99. They didn't have any luck with their early batches and it wasn't until after an extended break in adjunct professorship and industry that they got back together. A happy accident happened
62.
by viscanti 3y ago
It wasn't discovered in 99. That's when two of the researchers were in university and had a professor who had a theory that something like it could work. They took a good amount of time off from it while they worked at other jobs,
63.
by viscanti 3y ago
Well the whole argument feels kind of clickbait and premature. Arguing that replication efforts fall short, after less than a week of attempts, feels pretty weird. Why even weigh in on things right now if you're Nature? How many of the
64.
by viscanti 3y ago
> Humanity doesn’t hate itself, we like ourselves, which is why we are so concerned about what we are doing to ourselves. Yeah. The argument that single cellular life or some small animals might survive and eventually evolve into somethi
65.
by viscanti 3y ago
The problem here would be that if mask sets wear out at an unprecedented rate, it might not make sense to use these cutting edge nodes in the future (if/when we're several nodes ahead by then) for the things that currently use old
66.
by viscanti 3y ago
Yeah. Having a job seems like it could keep you from regularly being able to stop everything for 5 months of high paid work. Maybe the money is enough from the few months that they're fine with it (and maybe it's easy for them to
67.
by viscanti 3y ago
They need to be able to have 5 months where they can clear the calendars and just work on that. It's still a lot for 5 months, but I imagine there's a lot of downtime too. Are they getting 5+ months every year?
68.
by viscanti 3y ago
>Legacy admissions discriminate by class (if you are well off to be part of a donor family, you get a leg up) "Nearly three quarters of them would have been rejected if they had been subjected to the same standards as other white ap
69.
by viscanti 3y ago
They're very familiar with TikTok and how many people are moving over to use that. Their bet is that they can figure out how to highlight the most interesting content for each user and that content would be better than the vast majorit
70.
by viscanti 3y ago
> younger people don't like the idea that in order to have freedom to pursue the things you love, you might need to marry someone who supports you financially. How many people have ever liked it? Hasn't it been one of the few a
71.
by viscanti 3y ago
Some of the abstractions seem a bit forced, but it seems nice to be abstracted away from how you'd handle vector storage or the LLM that get used (maybe someone launches something that does better than GPT4 on some sets of tasks and a
72.
by viscanti 3y ago
Yeah, it sounds like you prefer plot or character development. I don't know that Bloom is supposed to be super interesting (not much interesting happens because it's a relatively mundane regular day). My take was that it was more
73.
by viscanti 3y ago
> but what do you, the reader, get out of it? Some deeper understanding of the human condition? Please. Certainly there are some chapters/styles where he's playing for the critics. But I thought it did a brilliant job of creati
74.
by viscanti 3y ago
From everything I've come across, hunter-gatherer people are substantially deficient in their Cheerio intake. Modern diets have multiple subspecies to choose from that would likely be unavailable to hunter-gatherers.
75.
by viscanti 3y ago
Training is unlikely for most things but you could probably run a trained model on it that needs a lot of RAM.
76.
by viscanti 3y ago
Not that weird of a spot. That's what it costs for the most premium experience they can launch with. They'd rather it be good than something everyone buys on day one. They can get costs down in future models as they scale and prog
77.
by viscanti 3y ago
> Honestly I blame the community. Sure. The point is that a community can make things go poorly, with or without a BDFL.
78.
by viscanti 3y ago
The py2 to py3 transition seems to have gone very poorly, although most of the pain from that is behind us now.
79.
by viscanti 3y ago
I'm not sure they needed to wait until there was some enterprise data solution before enabling it. It looks like a useful feature and if someone has assets that they're fine using it with, do they really need to wait until some en
80.
by viscanti 3y ago
It can be a Bayes rule thing where they're statistically less likely to participate in white collar work, but it could also be biases in companies who believe a candidate is less educated if they have a strong regional accent. The arti
81.
by viscanti 3y ago
Are the people gaming the system being rewarded for innate talent? Or is it hard work at gaming the system (presumably if it was easy to game the system everyone would do it)?
82.
by viscanti 3y ago
> People have been conditioned on expecting the next big thing, FOMO on trillion dollar companies, winner-takes-all etc. as if that is the new normal. People have been FOMO investing since the start of investment opportunities. Tulip Man
83.
by viscanti 4y ago
Regardless of how much time you have for your literature review, you likely are doing multiple passes through each paper. On the first pass you'll likely just read the summary and skim through the sections to see what kind of argument
84.
by viscanti 4y ago
A monad is just a monoid in the category of endofunctors. Would most people know what "brrrdogcogfog" means? Couldn't we make that argument about literally any word? I don't see why it applies more here than elsewhere. F
85.
by viscanti 4y ago
If an LLM hallucinates lines of code that can't even compile, I suppose it could also hallucinate logic issues which are more difficult to track down.
86.
by viscanti 4y ago
That's still under construction, which we'll need another marquee or images to let everyone know about.
87.
by viscanti 4y ago
> GPT-4's tokenizer is already far more efficient though still weighted to English. Right. It's a general question. Should they be allowed to take the kinds of optimizations they can with tokenization when it's a function
88.
by viscanti 4y ago
Because there's so much more English language for them to train on relative to most other languages, they're able to do some optimizations for English that they can't elsewhere. Should they not be able to implement optimizati
89.
by viscanti 4y ago
Sure. I think GR was probably heavily influenced by Joyce and Ulysses specifically. I don't think it's a hot take. I just don't know that I personally would characterize it as a re-telling. GR even goes as far as having mixed
90.
by viscanti 4y ago
Ulysses parallels The Odyssey, which is also a father and son story. I understand how those books are connected. I'm not sure that I understand how GR is a post-modernist retelling of either. Seems like I'm missing something so I&

More ›