currymj

38 ms·

511.
by currymj 6y ago
apart from the benefit of the humanities in an undergraduate education, there are big positive externalities from the scholarly work as well. for instance, even if most people never read any of the written work of historians, the existence
512.
by currymj 6y ago
A bill, the Endless Frontiers Act, has been introduced in both houses of Congress to make a section of the National Science Foundation that works more like DARPA, while massively increasing its funding. The idea is to extend the DARPA model
513.
by currymj 6y ago
for classifiers, often people want error rates not to differ too much on different groups, but this is fraught with difficulties. it seems to me that it is really situation dependent, complex, and you hit tough philosophical issues about th
514.
by currymj 6y ago
i understand your point perfectly well, but I think we've gotten to the crux of the disagreement, at least. I'm thinking about this pragmatically: a model is explainable if you can explain the model to some people you need to expl
515.
by currymj 6y ago
i think "explainable model" literally means "a model you can explain to people". You may have to explain your decision to a judge or customer after the fact, or you might even be asking a layperson to actually compute pr
516.
by currymj 6y ago
it is very strange to read someone claiming with extreme confidence that all fairness and explainability research is completely bogus. of course there are a lot of papers of questionable value, but that's a problem with machine learnin
517.
by currymj 6y ago
in the US, at this point, advocating for metrification is associated with a sort of odd, fringe type of political figure. See the recent presidental campaign of Lincoln Chafee. Or the very humorous "Stonecutters' Song" from t
518.
by currymj 6y ago
most machine learning research does not require millions of dollars... if you’re thinking about GPT-3, it uses the transformer architecture. But the paper that introduced the transformer architecture trained for a few days on 8 GPUs — surel
519.
by currymj 6y ago
Habitat 67 in Montreal is a Brutalist concrete building (many cubes) that is generally well-liked and has been a desirable place to live for 50 years. https://en.wikipedia.org/wiki/Habitat_67#/media/File:Habit
520.
by currymj 6y ago
i tend to trust Nielsen on this one, who literally wrote the book on quantum computing and has spent years teaching it to people. his philosophy is that the biggest obstacle to understanding is memorization of basic concepts and notation.
521.
by currymj 6y ago
just use PyTorch, it’s generally going to be the nicest. TF2 is much nicer than TF1 but most of the changes were just to make it more like PyTorch.
522.
by currymj 6y ago
thank you very much for the detailed response! much to chew on here, it really does seem like a very interesting class of models. from the papers it sounds like in practice clipping weights to a small set works okay, so the constant factors
523.
by currymj 6y ago
Nintendo’s portable systems (going back to the Game Boy and including the Switch) have never had region locks, on the logic that somebody might want to buy a new game while traveling. The consoles (except for Switch) do, though.
524.
by currymj 6y ago
Do you know of any other kinds of universal function approximators that also have a good regret bound like this, or is yours the first one? global convergence to any arbitrarily weird function at rate O(sqrt(T)) seems amazing, almost too go
525.
by currymj 6y ago
If you look into the news reporting on the individual cases, some look bad (emails documenting attempts to cover up, smuggling lab materials out of the country) and some don't (despite not including it on the grant form, researchers ac
526.
by currymj 6y ago
The policies proposed in this editorial seem pretty tame, but they aren’t all the policies being put forward on this issue. A recent executive order does target individual students who come from universities with military ties, even though
527.
by currymj 6y ago
While it seems reasonable to try to stop the Chinese government from unduly influencing US institutions, I am sincerely worried this is just going to turn into a witch hunt against Chinese nationals in the US on student and work visas.
528.
by currymj 6y ago
You know throwing a neural network at things doesn't just magically make them better... lot of advantages to linear techniques... not even clear that there's enough structure in a big 5 questionnaire to benefit from a fancy NN...
529.
by currymj 6y ago
wow, the rockafellar book is fantastic. i can't believe i'd never run into it before. thanks!
530.
by currymj 6y ago
if you want to be rigorous about it, you can often talk in terms of elements of the set of subgradients rather than gradients, and the convergence proofs for many of the popular optimization algorithms still go through.
531.
by currymj 6y ago
of course it can’t be avoided completely, but you’re supposed to try not to know who the authors are.
532.
by currymj 6y ago
peer review at CS conferences is normally anonymous and double-blind to reviewers and authors. however, the conference organizers themselves know who the reviewers and authors are, and the conference management websites are set up to avoid
533.
by currymj 6y ago
they can, although so can SSRIs, surprisingly, by sometimes triggering mania. it's good for experienced professionals to be involved who can look for warning signs beforehand, evaluate the risks, and intervene quickly if things go badl
534.
by currymj 6y ago
tons of elected sheriffs in the US, usually at the county rather than the city level. for example in the rest of New York State outside of the city, there are elected sheriffs. but there may also be police departments associated with towns
535.
by currymj 6y ago
the website itself is, but it aggregates journal articles which are not. there are a lot of academic journals with very low standards for peer review, but PubMed doesn't make any editorial choices.
536.
by currymj 6y ago
One example is England up until the 19th century, which had a well-established court system but no formal police departments, just unpaid constables, local militias, etc. I think it was probably similar in many other parts of Europe. I&#x27
537.
by currymj 6y ago
the argument is that there shouldn't be police forces as we know them today -- that the whole issue of crime and punishment should be handled completely differently, with communities organizing themselves to prevent malicious activity.
538.
by currymj 6y ago
This domain name issue also comes up with PubMed, which uses the nih.gov domain, making people think it's actually endorsed by NIH. I think it's not just a problem on this website; social news link previews also show the domain an
539.
by currymj 6y ago
take linear algebra ASAP. it's probably the single most useful course you can take at a university.
540.
by currymj 6y ago
i am honestly slightly worried that in certain social circles, things will get to the point where not taking psychedelics carries a stigma. i am guessing this is already the case in certain subcultures but I think there is a chance it could

More ›