this post was submitted on 18 Aug 2023
64 points (94.4% liked)
Asklemmy
43893 readers
1010 users here now
A loosely moderated place to ask open-ended questions
If your post meets the following criteria, it's welcome here!
- Open-ended question
- Not offensive: at this point, we do not have the bandwidth to moderate overtly political discussions. Assume best intent and be excellent to each other.
- Not regarding using or support for Lemmy: context, see the list of support communities and tools for finding communities below
- Not ad nauseam inducing: please make sure it is a question that would be new to most members
- An actual topic of discussion
Looking for support?
Looking for a community?
- Lemmyverse: community search
- sub.rehab: maps old subreddits to fediverse options, marks official as such
- !lemmy411@lemmy.ca: a community for finding communities
~Icon~ ~by~ ~@Double_A@discuss.tchncs.de~
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Given that the AI we have is prone to making things up because it “fits” according to the models it trains on, how much faith would you have in a translation done by an AI on writings made by people who lived millennia before said language models were developed?
Don't confuse modern LLM models (like ChatGPT) with AI. As the saying goes:
LLMs are a form of AI, but there is a lot more going on in the world of AI than just LLMs.
That’s a good point, and you’re right that I’m conflating them.
What other elements of AI would you imagine would be useful here?
You'd have to ask people who work in the AI field, and, alas, I'm not one of those people.
There has been a lot of language work on attempting to reconstruct the original Indo-European Language, using combinations of pattern recognition and statistical analysis of child languages. Those sorts of tools could aid in deciphering a dead written language.
https://en.wikipedia.org/wiki/Proto-Indo-European_language
However, another written language called Linear-A (of the ancient Minoans) has yet to be deciphered, despite lots of attempts at trying.
https://www.thoughtco.com/linear-writing-system-of-the-minoans-171553
So:
¯\(ツ)/¯
to expand your point, the sole job of an LLM is to, when given a sequence of words (e.g. half a sentence), predict what the next several words should be. the model has no concept of what English words mean, so instead it makes this prediction based on statistics that were derived from basically reading through hundreds of thousands of English sentences
TL;DR LLMs don’t understand languages, they’ve just memorized statistics about them
I'll have more faith once it can reliably switch back and forth between Unicode symbols and their underlying HTML entities. It understands the concept of emojis and can use them appropriately, but I can tell there's still some underlying issues in the token/object model for non-ASCII symbols.