37
all 9 comments
sorted by: hot top controversial new old
[-] Wheaties@hexbear.net 24 points 1 year ago

I wish the term "AI" would die. It's so unhelpful to understanding these programmes, and that's exactly why these bazinga-brained tech nerds love it so much. Can't have a conversation about the real-world effects of these things if half your audience thinks they're talking about baby HAL 9000.

Pattern Engines. That's what I've been thinking of these as. You feed enough sample pattern into it, and it spits out more of that pattern. "Large language model" is good, but that only covers the text. Pattern Engines work off of any pattern you set them up for; text, picture, audio, video.

[-] WoofWoof91@hexbear.net 14 points 1 year ago* (last edited 1 year ago)

i like "very fancy predictive text" myself

[-] StellarTabi@hexbear.net 11 points 1 year ago

Over-applied Statistics.

[-] AlbigensianGhoul@lemmygrad.ml 5 points 1 year ago* (last edited 1 year ago)

Since most of it is based on deep learning, you can call it some variant of "matrix multiplication optimisers." Giving it descriptors based on the actual mechanisms already helps remove the glamour a bit, which is why researchers often name their things by the effects rather than the inner workings (StyleGan, Node2Vec, AlphaGo).

[-] StellarTabi@hexbear.net 10 points 1 year ago

I wonder what's going to show up first on hugging face, an AI trained on something like conservapedia, or an AI trained on something like prolewiki?

this post was submitted on 19 Oct 2023
37 points (100.0% liked)

technology

23313 readers
151 users here now

On the road to fully automated luxury gay space communism.

Spreading Linux propaganda since 2020

Rules:

founded 4 years ago
MODERATORS