514
Running AI is so expensive that Amazon will probably charge you to use Alexa in future, says outgoing exec
(www.businessinsider.com)
this post was submitted on 25 Sep 2023
514 points (96.1% liked)
Technology
59377 readers
4800 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I don't understand this. Hasn't Intel or Nvidia (or someone else) been making claims about their next CPUs having AI functionality built-in?
Having "AI functionality" doesn't mean they can just get rid of their big/expensive models they use now.
If they are anything like Open AI's LLM, it requires very beefy machines with a ton of expensive RAM.
Well that's exactly what I was thinking when these companies were making these claims... like HOW could they possibly handle this locally on a CPU or GPU when there must be a massive database that (I assume) is constantly being updated? Didn't make sense.
EDIT: this entire website can go fuck off. You ask a simple question about some reasonably new tech, and you get downvoted for having the audacity to learn some new stuff. People on here are downright pathetic.
"AI" doesn't use databases per se, they are trained models built from large amounts of training data.
Some models run fine on small devices (like the model running on phones to make better pictures) but others are huge like Open AI's LLM.
Wouldn't that data be stored in some kind of database?
The training data isn't stored in the model. You can take an existing model and fine tune it on a whole bunch of additional data and the model size won't change.
No, the data will influence the model.
Some of the data may be found in the model itself (i.e. the AI generated images outputting mangled author signatures from the original works that were used during training) but not in the traditional form of a database. You can't directly retrieve that data back in its original form even if some models can be coerced to do something similar.
It's basically a statistical model built from training data.
The training of these huge models also cost a fortune. Allegedly in the millions of $ in a data and processing-intensive process.
You're right. Run an llm locally adjacent to your application sandboxes and local user apps and your office will lower its heating bills.