rspeer

66 ms·

691.
by rspeer 10y ago
Can an article be honorably retired? Joel's article was important 14 years ago, but that was a time when it was not clear that UTF-8 was going to become the default encoding of the Internet, and a time before emoji. In 2003, codepoints
692.
by rspeer 10y ago
Indeed, I seem to recall from the XKCD color survey that the correct spelling of "fuchsia" was not in the top five. Which makes it a messy name for a project. Though maybe Google can change the spelling by brand-name fiat, just li
693.
by rspeer 10y ago
I know Hoogle exists, but that just searches one kind of documentation. Despite the cute name, it's not Google. You can't Hoogle an error message and see if anyone else got it.
694.
How Luminoso made ConceptNet into the best word vectors, and won at SemEval(blog.conceptnet.io)
3 points by rspeer 10y ago|0 comments
695.
by rspeer 10y ago
I think there have been some attempts to explain, none of them fully satisfying. But there have also been systems that create vectors with that property more directly instead of as a side-effect of a neural net. Examples include Stanford&#x
696.
by rspeer 10y ago
This will be extremely helpful next time I have to use a Haskell library that decides to implement everything as infix operators named "~<$>" and ".~=" and stuff.
697.
by rspeer 10y ago
Very good question! Our OOV strategy was pretty important in SemEval. The first line of defense -- so fundamental to Numberbatch that I don't even think of it as OOV -- is to see if the term exists in ConceptNet but with too low a degr
698.
by rspeer 10y ago
Words with similar vectors have similar meanings. You use this in search, sentiment analysis, topic detection, finding similar text, and classification. Of course there are tests for this "words with similar vectors have similar meanin
699.
by rspeer 10y ago
OVH doesn't even want to take my money to keep my server running. Their auto-billing process is busted and when it goes wrong they just delete your server.
700.
by rspeer 10y ago
> Of course, norms would have to change. Sure. How? Changing the norms of academia will take a generation, and would require some actual incentive to change them in the correct direction.
701.
by rspeer 10y ago
Note the bonus assumption that legal immigrants don't matter and the US isn't "really" their home.
702.
by rspeer 10y ago
Publishing poorly-written crap will make you look worse than publishing nothing at all. There's already a terrible stigma around negative results. Grant committees and tenure committees would subconsciously like to believe that you are
703.
by rspeer 10y ago
> The closest thing I can think of is Watson from IBM. What are you imagining Watson is? Watson is not a singular entity that decides whether to play Jeopardy or analyze medical records or whatever. It's a brand. The comparison that
704.
by rspeer 10y ago
Duck typing is not the solution to every problem. Pylint has nothing to do with catching type errors. It catches problems with code style. I don't find the code in this blog post convincing, but there are reasons to write type-safe c
705.
by rspeer 10y ago
There are cases where you could justify writing "except Exception as e" and having it return a value, like if you're the author of Flask or Raven or something that includes handling arbitrary errors in other people's cod
706.
by rspeer 10y ago
Checking 2^80 hashes is indeed faster than I thought, thanks for the stats. But checking 2^80 hashes and writing them to long term storage is still ridiculous. That budget should still go to hiring cryptographers, not buying hard drives.
707.
by rspeer 10y ago
I think the more important collision to worry about in 2^80 time is the Earth colliding with the Sun. Our interstellar successors, if any, will probably have found something better than Git to use. EDIT: I should be clear that I'm not
708.
by rspeer 10y ago
Would it be possible to mirror just the data somewhere else, such as S3? I don't need the R code, but this sounds like it would make good companion data to my own wordfreq [1]. It would be interesting to see which words are learned ear
709.
by rspeer 10y ago
I doubt we have any container technology today that anyone would want to use in 10 years, just like nobody would want to use a Java applet today. And I say this having used Docker myself to make one piece of computational research reproduci
710.
by rspeer 10y ago
Can you set up Facebook to require 2FA when you log in from a mysterious new device? And configure 2FA so it'll fail to reach you while you're traveling? Edit: Actually, a sufficiently robust Suck Mode would make 2FA impossible. A
711.
by rspeer 10y ago
What if the phone can be put into "suck mode" that just makes it completely infuriating to get any information from for a while after you've flown a long distance in a plane? Sure, you can open Facebook and stare at a loading
712.
by rspeer 10y ago
If someone else made a speech synthesizer named "Char2wav", wouldn't the author of this project feel like their branding was being stolen from them?
713.
by rspeer 10y ago
AlphaGo is interesting because it's the world's best Go player, not because its input is a grid and its output is a move. A terrible Go player has the same format of inputs and outputs. Describing everything as a "universal f
714.
by rspeer 10y ago
I am getting tired of people implementing "deep learning to convert foo into bar" and staking a claim on the name "foo2bar". It leads to "AI hallucination", where even if "foo2bar" doesn't work,
715.
by rspeer 10y ago
Hackathons seem to be where companies can dump an untested implementation of a platform without ruining the experience of actual customers. They can either find out what's wrong with their platform or be pleasantly surprised when someo
716.
by rspeer 10y ago
Yeah, he has a point in the abstract, but this would have been better timed before Steve Bannon took over the executive branch and basically broke the Overton window. It's intellectually lazy to dismiss views via ad-hominem attacks, bu
717.
by rspeer 10y ago
I can't imagine a language even starting to displace Python without a good REPL. Not just a hacked-together "lol does this work" REPL, something you can rely on, and preferably something you can use as a Jupyter backend.
718.
by rspeer 10y ago
Yes, I'm glad that on top of the cultural problems of Uber, at least they don't support tipping culture. I use Lyft when I have the choice now, but the tipping makes me uncomfortable. Tipping is why taxis drive past people of the
719.
by rspeer 10y ago
Most recruiters are comically bad at targeting -- apparently there isn't enough downside to just carpet-bombing as many e-mail addresses as they can find -- but this does not mean that recruiting is inherently wrong.
720.
by rspeer 10y ago
So does this mean Google is going to stop indexing the text of WSJ articles?

More ›