nerdponx

45 ms·

811.
by nerdponx 3y ago
Intent aside, there's certainly a striking difference in the actual outcomes for civilians.
812.
by nerdponx 3y ago
I really enjoyed these 3 together: https://www.youtube.com/watch?v=lOSfM2V8EkI https://www.youtube.com/watch?v=HB7NxVpskI0 https://www.youtube.com/watch?v=PmKWkiDAm5k Great atmosphere. Make
813.
by nerdponx 3y ago
Same reason we can't trust people to drive 35 MPH when the road is straight and wide, no matter how many signs are posted to declare the speed limit. It's just too tempting and easy to become complacent. That, and these companies
814.
by nerdponx 3y ago
The alternative is the Kik NPM problem. It's bad both ways.
815.
by nerdponx 3y ago
The R language has this feature as well. It's a whole lot of fun to work with.
816.
by nerdponx 3y ago
If it takes any inspiration from APL, it would be mostly indirect, via Matlab.
817.
by nerdponx 3y ago
The whole magic is that expressions are in fact just objects in the language. And no, there aren't any immutable bindings here.
818.
by nerdponx 3y ago
It's often said that the cruelty is the point.
819.
by nerdponx 3y ago
These aren't financial hobbyists. At $48,560, these are mostly people who don't have a lot of money relative to their living expenses, and are doing what they need to do in order to not feel totally broke. < $50k in a lot of pl
820.
by nerdponx 3y ago
My point was that this particular inductive bias doesn't inherently beget a "language model".
821.
by nerdponx 3y ago
I wish. My savings are all going to go into a house, kids' college, and my retirement. Not everyone on HN is making $300-500k TC for a decade before having their first child.
822.
by nerdponx 3y ago
The entire innovation (discovery?) of LLMs is that a good formula for the task of sequence completion turns out to also be a good formula for a wide range of AI tasks. That emergent property is why language models are called language models
823.
by nerdponx 3y ago
The article makes it sound like the only parser for URLS in the entire Python ecosystem is urllib.parse, regardless of which spec it supports. Hyperlink and Yarl are absolutely prior art here IMO, and at least deserve a mention in an articl
824.
by nerdponx 3y ago
I posted this elsewhere in the thread, but there absolutely is prior art here. Check out Yarl (urllib.parse wrapper with the nicer interface) and Hyperlink (green field, immutable OO style, focus on correctness). Both on PyPI for many years
825.
by nerdponx 3y ago
No mention of prior art? The Hyperlink library has stated correctness as its goal for a long time: https://pypi.org/project/hyperlink Of course there is always room for new projects, but it still feels weird to act as
826.
by nerdponx 3y ago
I think the really disturbing aspect to me is that no editor or reviewer flagged this. It's literally the first sentence of the paper! Did they just read the abstract and say "meh, good enough"? Were they all using AI to revi
827.
by nerdponx 3y ago
As if it wasn't bad enough that most consumer goods have completely bifurcated into "junk" and "luxury", now it's hard to even tell which products fall into which category, because there is so much junk now bei
828.
by nerdponx 3y ago
The difference is that she is alleging in court that she was not acting.
829.
by nerdponx 3y ago
It depends on what you mean by traditional. If you're talking about a "tube" site, then you're probably right. But no namebrand porn studio would do something like that.
830.
by nerdponx 3y ago
I think that's kind of the point of the paper. The model is based on un-normalized dot products, and wasn't deliberately designed to produce meaningful cosine similarities. They are showing that, in that case, cosine similarities
831.
by nerdponx 3y ago
It's not even a distance metric, it doesn't obey the triangle inequality (hence the not-technically-meaningful name "similarity", like "collection" as opposed to "set").
832.
by nerdponx 3y ago
It's their point, and it's a good one, but I think they're somewhat overstating how common power-law data is; it probably varies a lot by field of study. And at least the logarithm of a power-law variable can help bring it ba
833.
by nerdponx 3y ago
The T distribution arises as the ratio between a Gaussian r.v. and the square root of a Chi-square r.v.: X ~ Gaussian(m, v) S ~ ChiSquare(n) T = X / sqrt(S / n) The sampling distribution of the sample mean is Gaussian w
834.
by nerdponx 3y ago
Great little demo. > It is only when the statistical evidence against the null is overwhelming — “six sigma” overwhelming or more —that you’re also getting tight confidence intervals in relative terms. Among other things, this highlights
835.
by nerdponx 3y ago
Fortunately that's exactly what the program covers: replacing crusty old supply pipes running from the main to your building.
836.
by nerdponx 3y ago
Isn't OnlyOffice browser-based? Or is that something else?
837.
by nerdponx 3y ago
Oh, so the Boston Lead Replacement Incentive program is based on benefits that may be entirely speculative? Might I encourage you to sample our finest tap water, from our finest century-old lead pipes? Oh, maybe another day then.
838.
by nerdponx 3y ago
The implication is not that low IQ causes lead exposure, but that low IQ and lead exposure are both caused by a third factor, presumably poverty.
839.
by nerdponx 3y ago
It's possible that you experienced the end of the 2->3 transition. My Python scripts from 2015 all still work, except the ones that used the then-experimental async framework, which was widely documented as experimental and subject
840.
by nerdponx 3y ago
The grocery stores generally aren't the ones increasing prices.

More ›