this post was submitted on 21 May 2024
120 points (86.6% liked)
Technology
73094 readers
2386 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- 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, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
You're referring to something that is changing and getting better constantly. In the long term LLMs are going to be even better than they are now. It's ridiculous to think that it won't be able to replace any of the workers that were fired. LLMs are going to allow 1 person to do the job of multiple people. Will it replace all people? No. But even if it allows 1 person to do the job of 2 people, that's 50% of the workforce unemployed. This isn't even mentioning how good robotics have gotten over the past 10 years.
You must have one person constantly checking for hallucinations in everything that is generated: how is that going to be faster?
It's hard to improve when the data in is human and the data out cannot be error checked back against its own data in. It's like trying to solve a math problem with two calculators that both think 2+2 = 6 because the data they were given said that it's true.
How is it hyperbole? All artificial neural networks have "hallucinations", no matter their size. What's your magic way of knowing when that happens?
LLMs now are trained on data generated by other LLMs. If you look at the "writing prompt" stuff 90% is machine generated (or so bad that I assume it's machine generated) and that's the data that is being bought right now.
There is a plateau to be hit at some point. How close it is, depends who you ask. Some say we are close, others say we are not but it definitely exists. LLMs suffer, just like other forms of machine learning, from data overload. You simply can't be infinitely feeding it data and keep getting better and better results. ChatGPT's models got famous because value function for learning had humans involved who helped curate quality of responses.