this post was submitted on 19 Mar 2025
236 points (96.5% liked)
Technology
67050 readers
4051 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
Lest we concede the point, LLMs don't write. They generate.
What's the difference?
Parrots can mimic humans too, but they don’t understand what we’re saying the way we do.
AI can’t create something all on its own from scratch like a human. It can only mimic the data it has been trained on.
LLMs like ChatGP operate on probability. They don’t actually understand anything and aren’t intelligent. They can’t think. They just know that which next word or sentence is probably right and they string things together this way.
If you ask ChatGPT a question, it analyzes your words and responds with a series of words that it has calculated to be the highest probability of the correct words.
The reason that they seem so intelligent is because they have been trained on absolutely gargantuan amounts of text from books, websites, news articles, etc. Because of this, the calculated probabilities of related words and ideas is accurate enough to allow it to mimic human speech in a convincing way.
And when they start hallucinating, it’s because they don’t understand how they sound, and so far this is a core problem that nobody has been able to solve. The best mitigation involves checking the output of one LLM using a second LLM.
It's interesting how humanity thinks that humans are smarter than animals, but that the benchmark it uses for animals' intelligence is how well they do an imitation of an animal with a different type of brain.
As if humanity succeeds in imitating other animals and communicating in their languages or about the subjects that they find important.