currymj

50 ms·

631.
by currymj 7y ago
she was a very good writer. "Against Interpretation" is a reasonable collection of essays to look at and they are mostly pretty accessible.
632.
by currymj 7y ago
tectonic is great! I find it works as a "daily driver" now, it's not just another experimental Rust project. it downloads everything as it's used, and automatically decides when to rerun compilation to deal with citation
633.
by currymj 7y ago
right now such a parallel track doesn't really exist, at least not on an even footing. as the article shows, there's still a heavy focus on beating accuracy benchmarks, at huge computational cost, in terms of what actually gets ac
634.
by currymj 7y ago
being in Europe makes a big difference here I think. in the US i know several people who did an MS, worked in industry, then went back for a PhD. it’s extremely possible. i suspect this is less true in Europe, where the degree process is fa
635.
by currymj 7y ago
redlining and similar practices are still usually illegal, although it is probably easier to cover up.
636.
by currymj 7y ago
Misleading article and headline; a few celebrity researchers in high-up positions make this much. You can probably find some chemists who make $1M too if they have made it to executive positions. Most AI researchers in industry make a lot o
637.
by currymj 7y ago
this is fantastic, it could be a parody of olden times technology but it’s real!
638.
by currymj 7y ago
it's conventional to leave these out to preserve anonymity during the peer review process; if one is available, it will be added in the final published version if it's accepted. this article is still unpublished and was only submi
639.
by currymj 7y ago
This paper was literally submitted to the venue yesterday! I generally think it's good how fast-moving ML is, but this seems excessive.
640.
by currymj 7y ago
There hasn't been any kind of paradigm shift but there have been a bunch of real although incremental improvements. Improved optimization algorithms, interesting and novel neural network layers. Some of this is even motivated by mathem
641.
by currymj 7y ago
I use conda because it’s the default way to install PyTorch. That’s pretty much it. Pretty sure any NumPy install should use reasonable BLAS nowadays but I believe conda includes Intel MKL which maybe makes a difference. Every couple months
642.
by currymj 7y ago
i think the idea of reducing emissions is to reach a new equilibrium as quickly as possible, and one as close as possible to our current climate which we know is pretty good. climate change is likely not going to destroy all civilization (n
643.
by currymj 7y ago
even if we miss a target, more emissions beyond that target still make things much worse. it's not like we can just declare bankruptcy or something. ultimately we have to reduce emissions as much as possible while preparing for their
644.
by currymj 7y ago
getting things like this to transfer well is not necessarily easy, it’s a major active area of research. even if you somehow had a perfect physics model, there are problems like the fact that motors and joints behave differently as they wea
645.
by currymj 7y ago
I sort of agree with the principle -- for instance, releasing easy-to-use pretrained SOTA models for "deepfakes" type tasks can cause real harm. There are existing repositories out there which are already morally dubious in my vie
646.
by currymj 7y ago
1-5 with half increments has 9 possible scores which is basically the same as 1-10. I think this is what is meant, not "people say 4.5/5 but nobody is really that stressed, they must be thinking 4.5/10".
647.
by currymj 7y ago
here's a story, originally published in 2010, about Jeffrey Epstein: https://web.archive.org/web/20170606013136/http://www.thedai... People knew, or should have known, who they were dealing with.
648.
by currymj 7y ago
that account in particular eventually admits to not having read much about the case. i honestly am pretty sure the vast majority of the contrarians really are just not familiar with the whole Epstein story, and may only be hearing about it
649.
by currymj 7y ago
the HN comments on all the recent Epstein stories have been very strange. I have to assume at least some of the more contrarian views are coming from people who haven't been following the full story in the US news, and so don't qu
650.
The Secret History of Dune(lareviewofbooks.org)
319 points by currymj 7y ago|77 comments
651.
by currymj 7y ago
that statistic, unlike some of the others, isn't a space comparison between parking spots and apartment square footage. it is instead a legal requirement. if somebody wants to build a building with 25 2-bedroom apartments, they have to
652.
by currymj 7y ago
this is the cricket approach. baseball typically plays multi-game series instead. i agree in spirit though. i have long thought it would make sense to abolish the playoffs. give the pennants to the teams with the best records. then, have th
653.
by currymj 7y ago
The Michael Nielsen online text is very well-regarded and mostly fits what you describe for 1, although it may have too much code for your taste in addition to the mathematical basics. However the code is not "how to use TensorFlow&quo
654.
by currymj 7y ago
with arXiv the TeX source is available for download (under the "Other formats" link), which I think is what arXiv-vanity uses. I imagine it would be really tough to go reliably from arbitrary PDF to HTML without screwing up the ma
655.
by currymj 7y ago
regarding 1, even papers with high quality published code fall prey to bit rot very quickly -- go dig through some git repos from ML papers from 2015 and try to run them. distributing Docker containers works better but even that won't
656.
by currymj 7y ago
this is an incredible reference for a machine learning researcher who wants to fill in some gaps in their existing mathematical knowledge. But I would be shocked if this would be of any use for someone trying to learn a little linear algebr
657.
by currymj 7y ago
this is precisely what’s cool about speed running, boring computer bugs and exploits become a kind of magic when you put them in the context of a game people already know and can understand.
658.
by currymj 7y ago
The excellent book "Paradoxes in Probability Theory" by William Eckhardt has some good arguments against this, the Simulation Hypothesis, and similar things. One simple way to summarize the counterargument is that in a lot of case
659.
by currymj 7y ago
Naming people after flowers is actually incredibly common if you think about it.
660.
by currymj 7y ago
I don’t know but Moms Across America is 100% not a credible source for anything. If they’re right it’s a “stopped clock” situation.

More ›