@antolius I thought it was hardware that helped accelerate execution of pre-trained models. The idea being you don't need to round-trip to the cloud and back for some things. Possibly for privacy, but also for performance reasons?
@Moss The sun was definitely an orange ball as it came up this morning, low on the horizon. But that changed after it was a bit higher. I think the smoke is still well east of Portland. For now. :/
It looks pretty bad on the eastern half of Washington right now. :(
The Hubble Ultra Deep Field in Light and Sound
Image Credit: NASA, ESA, Hubble; Sonification: G. Salvesen (UCSB); Data: M. Rafelski et al.
I do things with words. And the things that lie behind words.
I'm a #Linguistics professor, working mainly on Formal #Semantics and its interfaces. Much of my research also involves Historical Linguistics (/#Philology) to a certain extent. My central tool for analysis is a predicate logic-style system augmented with a typed #LambdaCalculus.
I frequently research languages of #SouthAsia (e.g. #Hindi, #Nepali, #Sinhala, #Sanskrit) & the #WestIndies (#Patois, #RastaTalk & the role of #African languages in the #Caribbean), as well as a bit of #OldEnglish (https://heorot.dk ).
I have a certain wary enthusiasm about particular types of technology. #LaTeX and fine #Typography more generally. #Lisp things, including #CommonLisp, and #Scheme (#Guile, #Racket), and #Emacs, as a sort of incarnation of a minor Lisp Machine. I'm #Haskell-curious, but don't care much for significant whitespace languages. (Software I've written mainly resides (for the moment) at: https://gitlab.com/emacsomancer/ )
Within linguistics, I work on a fairly wide range of things, from the use of delimited continuations in natural language semantics to the morphology of #Rastafari language to #Cyberpunk neologisms to #IndoEuropean myth-preserving formulaic language (slaying of dragons and worms) to algorithmic detection of different focus types and constituents in unannotated plaintext. (you can see more about these things here: https://lambda-y.net )
Recently, I've been exploring the use of devices borrowed from computer science (e.g. delimited continuations, monads) for working out computational issues in natural language semantic formalism.
Safari isn't killing the web
The web as we thought we knew it is dead for most people. The only way *companies* have influence is to control the underlying platform the browser runs on. Apple's iOS, and Microsoft's Windows (where they heavily market Edge and decided to let Google's team do a lot the work for them.) It's a war of platforms. iOS vs Windows vs Chrome (nee the Web).
At least that's what it looks like from here.
(Atlernate CW: Unimportant opinion about the web)
Safari isn't killing the web
Safari is supposedly killing the web by not shipping features, which in the absence of total control of the web platform on iOS, will result in Google's Chrome dominating everything. So the suggested alternative is to go along with everything Google wants, but really *lean in* to the process to somehow convince Google to do something it probably doesn't want to do in the face of its overwhelming market power on the web.
@akkartik Also, I don't believe the distros provide an absolute guarantee, but I'm generally not worried that my next apg-get upgrade or dnf update is going to end up with Inkscape secreting off my ssh private keys or something. Whereas anytime I install something on a Windows machine from Totally Trustworthy Inc, I have my doubts. 😆
@akkartik Well, I don't have a fixed definition of "trust". I think it will depend on threat models and the like. So some will be stricter than others. I agree that a flatpak from same volunteers can be just as trustworthy as a .deb, but my impression was that the point of those packaging solutions was primarily to provide a way to cut distros out. I understand packaging certain things is also a pain in these days of thousands of js dependencies, docker containers, and cloud deployment first.
@akkartik It's a tough problem. I think cooperative/community OS distributions actually go a long way towards providing third party code you can trust, assuming you run code packaged by the community. (Ignoring malicious input for now.) You know, the Linux and BSDs etc.
But now we're busy trying to throw out the advantages of that model by creating flatpak, snaps, AppImages and the like.
The cloud providers have their own set of problems since they nearly run *only* untrusted code. 😆
@akkartik I'll expand the half-formed thought a little. I was reading about various security mitigations in modern kernels, and the thought occurred to me that most of them are there to protect the system when running untrusted code.
Of course, even a trusted program can become an untrusted one if somebody can figure out how to make it do something it was never intended to do. So I guess I need to extend my statement.
Running untrusted code was a mistake. Reading untrusted input too. 😆
Interests: programming languages, open platforms, art, craft, diy, and justice. #nobot
The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!