vogon

the evil "Website Boy"

member of @staff, lapsed linguist and drummer, electronics hobbyist

zip's bf

no supervisor but ludd means the threads any good


twitter (inactive)
twitter.com/vogon
bluesky
if bluesky has a million haters I am one of them, if bluesky has one hater that's me, if bluesky has no haters then I am no more on the earth (more details: https://cohost.org/vogon/post/1845751-bonus-pure-speculati)
irl
seattle, WA
You must log in to comment.

in reply to @vogon's post:

more explanation for the !!!!:

  • neural networks (the trendy fundamental technology in "deep learning"/AI) can do a lot of things well, quickly, and at large scale, but they "learn" a bunch of mistaken assumptions and can't "explain" the "logic" they use to make a decision. in this particular case, an explainable neural network for tag recommendation would be (and is) able to tell you which words in a petition are responsible for the machine thinking that your request is the ambit of the Third Undersecretary For Large But Not Too Large Machinery and telling you to talk to them first.
  • very few people do research on explainable neural networks in the west because our industrial research apparatus values being able to explain the output of a bureaucratic process at close to $0.

very few people do research on explainable neural networks in the west because our industrial research apparatus values being able to explain the output of a bureaucratic process at close to $0

Plus also if the machine's process can be audited then it breaks the whole "ooOhH we don't actually kNoOoOw how it works!! therefore you are obligated to worship it as a god, and also me as its priest" thing, which is alarmingly central to the US tech industry

Gah, I was going to "liberate" the article for you but it's sadly in a pretty rare journal - I've got a massive network and still don't have access to it. $30 institution price is steep for the PDF of a single article, screw these journals.