rspeer

64 ms·

571.
by rspeer 9y ago
The decision to try to use FreeBSD or Linux would probably leave them unable to type in their own language. I have tried what passes for an IME for Linux desktop applications (formerly SCIM, now IBus). It doesn't work with some applica
572.
by rspeer 9y ago
Okay, but a lot of these changes don't sound like the final word, they sound like "hey let's try this new thing and see how people like it".
573.
by rspeer 9y ago
This is exactly the sort of vague CS-flavored hype that the article is criticizing. You've said nothing that distinguishes Sanskrit from any other language.
574.
by rspeer 9y ago
Yeah, I understand that. I'm re-purposing the data and it's my job to decide how that works. But this could be easier. What I hate about Wikimedia's format is templates. They are not very human-editable (try editing a templat
575.
by rspeer 9y ago
The structure I need for my particular project (ConceptNet) is: * The definitions from each Wiktionary entry. * The links between those definitions, whether they are explicit templated links in a Translation or Etymology section, or vaguer
576.
by rspeer 9y ago
The GP said Wikidata isn't suitable for many purposes, different from any . It's a nice agreed-upon vocabulary for linked data. But you still need the data that the vocabulary refers to. The information you can get without ever
577.
by rspeer 9y ago
That seems to just be kinds of IO errors, though.
578.
by rspeer 9y ago
That gives you Wikitext encapsulated in XML. How do you get at the content of the Wikitext? I work on a Wikitext parser [1]. So do many other people, in different ways. Wikitext syntax is horrible and it mixes content and presentation ind
579.
by rspeer 9y ago
Sorry, I'm trying very hard to learn Rust, and I'm willing to accept a lot of its restrictions, but this didn't clear anything up. Why does every library implement its own Error type? It feels like reinventing the wheel, an
580.
by rspeer 9y ago
Eh. It's not too much of a strawman. I was under the apparently incorrect impression that there's some Fortran in NumPy, as the NumPy codebase is where f2py comes from. But it could very well not be used in any NumPy operations. I
581.
by rspeer 9y ago
These are some pretty naive suggestions for how to uproot an entire domain of computing. Julia is a JIT language. Whatever edge you get in pretty code is going to be paid for in speed. Also, it's pre-1.0 and breaks compatibility freque
582.
by rspeer 9y ago
And then nobody ever receives your mail because the big mail servers don't trust you.
583.
by rspeer 9y ago
Okay... this turns out to have been a false accusation. I'm very sorry. The company that does this is ResearchGate, not Academia. Academia's e-mails have appropriate From: lines, and don't appear to be sent from beyond the gr
584.
by rspeer 9y ago
Hi Ben. Will Academia ever stop impersonating people in e-mail? I receive e-mails all the time purporting to be actually sent by academic colleagues, which are instead form messages sent by Academia.edu. I know that you're not sendin
585.
by rspeer 9y ago
"Galaxies" is the one I keep coming back to. It has the property that computer-generated instances of it are often satisfying, with distinct "aha" moments where everything falls into place. That's not true for all l
586.
by rspeer 9y ago
Do you have any evidence that PayPal's sign-up process 10 years ago included verifying that you were over 18, or are you just making things up?
587.
by rspeer 9y ago
You don't need to. You're never forced to solve really thorny problems to make your code run faster by a constant.
588.
by rspeer 9y ago
Many graph algorithms benefit hugely from locality of reference, because graphs have an inherently relevant concept of things being affected by other things they are near. Taking advantage of locality of reference can make a difference of m
589.
by rspeer 9y ago
Such as graph algorithms, the current topic of discussion.
590.
by rspeer 10y ago
I think deep learning is both useful and overhyped. If we got in the habit of calling it something like "large-scale parameter fitting by gradient descent", we could talk about it more sensibly. I have not met any actual deep lear
591.
by rspeer 10y ago
What comparison are you even making here? I know that Deep Blue and Deep Mind are different. There is 25 years (edit: sorry, 20 years) between them! Deep Blue is not deep learning. Did the word "deep", used in two unrelated ways,
592.
by rspeer 10y ago
Any actual chess knowledge in Deep Blue was an optimization. The fundamental algorithm is alpha-beta search, which can learn to play many, many games of strategy. Not Go, as we know now, but the exceptions are few. This is an algorithm that
593.
by rspeer 10y ago
I think there's a heavy dose of press release to what Google is saying. Most people wouldn't call PR puff "lying", but only because standards are low. I don't think Google has fundamentally different deep-learning t
594.
by rspeer 10y ago
The Atari AI's learning phase is a specifically designed deep neural network. The network did not design itself. It was designed by human programmers. There are probably numerous variants of it that did not learn to play Atari games, a
595.
by rspeer 10y ago
I wouldn't say that DeepMind learns Atari games "from scratch" any more than Deep Blue learned chess from scratch. It learns to play Atari games because it's a machine designed to learn to play Atari games.
596.
by rspeer 10y ago
I haven't read Bostrom's book. I don't think I would enjoy it. Maybe I need to grudgingly read it to be able to respond to what Bostromites say. Here's the thing. If Bostrom's claims about AI are so strong, why does
597.
by rspeer 10y ago
No, the current limit is not physical, it's that nobody has any idea how general intelligence works. You do not, in fact, get general intelligence by accident by throwing a lot of connections in one place, just like you do not get a hu
598.
ConceptNet Numberbatch 17.04: better, less-stereotyped word vectors(blog.conceptnet.io)
6 points by rspeer 10y ago|0 comments
599.
by rspeer 10y ago
Let's suppose the machines are spying on you and you want to learn a secret language of humans, a language with the highest ratio of human speakers to NLP technology. Which language should it be? I think it would be a language spoken i
600.
by rspeer 10y ago
They are. There was a SemEval competition about having vectors that are both accurate for 5 individual languages and aligned between all the pairs of languages [1], where Numberbatch was the first-place system [2]. The fastText_multilingual

More ›