Leopards Ate My Face

7038 readers

821 users here now

Rules:

The mods are fallible; if you've been banned or had a post/comment removed, please appeal.
Off-topic posts will be removed. If you don't know what "Leopards ate my Face" is, try reading this post.
If the reason your post meets Rule 1 isn't in the source, you must add a source in the post body (not the comments) to explain this.
Posts should use high-quality sources, and posts about an article should have the same headline as that article. You may edit your post if the source changes the headline. For a rough idea, check out this list.
For accessibility reasons, an image of text must either have alt text or a transcription in the post body.
Reposts within 1 year or the Top 100 of all time are subject to removal.
This is not exclusively a US politics community. You're encouraged to post stories about anyone from any place in the world at any point in history as long as you meet the other rules.
All Lemmy.World Terms of Service apply.

Also feel free to check out [email protected] (also active).

Icon credit C. Brück on Wikimedia Commons.

founded 2 years ago

MODERATORS

[email protected]

205

Chinese firms ‘distilling’ US AI models to create rival products, warns OpenAI (www.theguardian.com)

submitted 4 months ago by [email protected] to c/[email protected]

37 comments fedilink hide all child comments

Honestly an AI firm being salty that someone has potentially taken their work, "distilled" it and selling that on feels hilariously hypocritical.

Not like they've taken the writings, pictures, edits and videos of others, "distilled" them and created something new from it.

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 2 points 4 months ago (1 children)

How does this get used to create a better AI? Is it just that combining distillations together gets you a better AI? Is there a selection process?

[–] [email protected] 7 points 4 months ago

Chains of distillation is mostly uncharted territory! There aren't a lot of distillations because each one is still very expensive (as in at least tens of thousands of dollars, maybe millions of dollars for big models).

Usually a distillation is used to make a smaller model out of a bigger one.

But the idea of distillations from multiple models is to "add" the knowledge and strengths of each model together. There's no formal selection process, it's just whatever the researchers happen to try. You can read about another example here: https://huggingface.co/arcee-ai/SuperNova-Medius