jhrmnn

10 ms·

91.
by jhrmnn 4y ago
Ok, so when people say that they use ChatGPT for coding, they use it as a code snippet library? Here, a function that has been probably written million times before in some similar form, and a template for a simple web app.
92.
by jhrmnn 4y ago
I’ve been thinking about this. My current theory is that molecular simulation is a much more heterogeneous activity than language modeling. Language is always the same kind of data. Molecular simulations span orders of magnitude in space
93.
by jhrmnn 4y ago
Right, correct. I was already thinking ahead about the recent TensorFloat precision, which is further directly targeting matrix multiplication.
94.
by jhrmnn 4y ago
Computation-wise, transformers are really just a bunch of matrix multiplications, nothing more to it. (Which is partially why they’re so efficient and scalable.) Also, Nvidia’s GPU architectures are moving in the TPU direction ( https:&#x2F
95.
by jhrmnn 4y ago
https://en.wikipedia.org/wiki/Tensor_Processing_Unit
96.
by jhrmnn 4y ago
> you don't get to decide what "social good" is, especially in this case. People get to choose what is good for them He or she doesn’t get to, but the society does. Collectively. Not individually. That’s the premise that m
97.
by jhrmnn 4y ago
I understood this as $/FLOP, I think it's plausible that that has been happening.
98.
by jhrmnn 4y ago
Brilliant work.
99.
by jhrmnn 4y ago
Indeed. It’s hard to beat the IDE concept here. Especially given the unique features of VS Code like remote development and dev containers.
100.
by jhrmnn 4y ago
Are there 10k commute bikes? I doubt people take 10k pro road bikes to work
101.
by jhrmnn 4y ago
The same with Czech (and I assume all Slavic languages?). Would be interesting to track the etymological development of this meaning.
102.
by jhrmnn 4y ago
Apple, Microsoft, Amazon?
103.
by jhrmnn 4y ago
The boundaries are getting fuzzier. Implement this in Python _with JAX_ with little if any extra effort, run it on GPU, and you get performance you’d alternatively get only by writing custom CUDA kernels in C++.
104.
by jhrmnn 4y ago
How do you deal with the 9 hour difference between Europe and West Coast?
105.
by jhrmnn 4y ago
I think the fear of black swans, that is, unknown risks. I think most people don’t understand that the experts can really rule out something like a meltdown in modern reactors
106.
by jhrmnn 4y ago
What I'd love is a program that automatically downloads and archives any article-form website on which I spend more than a minute in a browser, and makes everything full-text searchable. Does something like that exist? Perhaps it could
107.
by jhrmnn 4y ago
I found it amusing how it just took them a few chapters to make me naturally think of 17th century Europeans as barbarians in light of native North Americans
108.
by jhrmnn 4y ago
--update-refs is great, resetting dependent branches after rebasing has been a constant source of annoyance
109.
by jhrmnn 4y ago
I rather think that we're past our ability to make useful yet comprehensible (to us) models in physics. The only avenue is to make useful and incomprehensible models. Using machine learning for that is one way. String theory might be o
110.
by jhrmnn 4y ago
I like that most of that protocol is multi-scale—it applies to one-day, one-person efforts as well as to tasks requiring dozens of people for many years.
111.
by jhrmnn 4y ago
It's interesting that this is closer to how human memory operates—we're quite good in unconsciously fabricating false yet strong memories.
112.
by jhrmnn 4y ago
I think of Jupyter Notebooks as scratch paper on my desk. It's not to archive things, it's for developing ideas. Once ideas are developed, I transfer them to a long-term medium (LaTeX or Markdown document, Python source file, etc)
113.
by jhrmnn 4y ago
This excellent book discusses a lot of what you touch on—D. Golash: The Case against Punishment https://books.google.com/books/about/The_Case_Against_Punish...
114.
by jhrmnn 4y ago
Interesting! Is this US-specific?
115.
AI4Science to empower the fifth paradigm of scientific discovery(microsoft.com)
4 points by jhrmnn 4y ago|0 comments
116.
by jhrmnn 4y ago
The document appears to have 80 contributors, that's hard to do with a blog post. It could have been a wiki page. But then I'm not sure if hosting a single repo on Github is harder than hosting a wiki. And of course Github provide
117.
by jhrmnn 4y ago
Well, yes, if you take the whole Bible literally, it’s not very nice and not even self-consistent. But I don’t think many reasonable people do that
118.
by jhrmnn 4y ago
He doesn’t need anything. The statement is that there’s a way to live a good (satisfying) life, the God being the creator knows that way, and he reveals it to people, who should follow it (to live satisfying lives)
119.
by jhrmnn 4y ago
Are you suggesting that all those people doing fusion research don’t know about this, or that they make false promises to the public to fund their research?
120.
by jhrmnn 4y ago
Never thought of that, good point. I guess in the end if we wanted to find the boundary for harmful/tolerable addiction, we might not be able to find it, it’s a continuous spectrum. I guess in the end intent would need to decide the

More ›