AI stuff (was: How to properly do L-R method)

General discussion about learning languages
User avatar
tastyonions
Black Belt - 1st Dan
Posts: 1624
Joined: Sat Jul 18, 2015 5:39 pm
Location: Dallas, TX
Languages: EN (N), FR, ES, DE, IT, PT, NL, EL
x 4047

Re: AI stuff (was: How to properly do L-R method)

Postby tastyonions » Sat Mar 16, 2024 11:49 am

Well, they aren't my problems, as I get good results with "traditional" methods and see no reason to stay sat down in front of a computer screen even more than I already do, but if chatting with bots gets some people to persevere with a language when other things wouldn't, then more power to them.
3 x

Cainntear
Black Belt - 3rd Dan
Posts: 3538
Joined: Thu Jul 30, 2015 11:04 am
Location: Scotland
Languages: English(N)
Advanced: French,Spanish, Scottish Gaelic
Intermediate: Italian, Catalan, Corsican
Basic: Welsh
Dabbling: Polish, Russian etc
x 8813
Contact:

Re: AI stuff (was: How to properly do L-R method)

Postby Cainntear » Sat Mar 16, 2024 1:55 pm

bombobuffoon wrote:
tastyonions wrote:I’ve been assuming that most of the “AI tutors” out now are just GPT or one of its competitors with the thinnest of skins stretched over it. Eventually there will probably be bots trained specifically for the domain of language learning that don’t require as much direction from the student but that’s not reality yet.


It appears we are stuck in the never ending cycle of "when chatGPT version x comes out those problems will be solved".

Yup -- it's a failure to understand what ChatGPT does. ChatGPT does general chit-chat well because it's seen more general chit-chat than a human would encounter in multiple lifetimes. ChatGPT needs an *enormous* amount of data to do anything, and there simply isn't enough data on language lessons to train it.

ChatGPT does an easy task well based on unimaginable amounts of data.

Thinking that if it can train a computer to do an easy task with lots of data means you can train a computer to do a difficult task with less data is kind of silly.
2 x

bombobuffoon
Yellow Belt
Posts: 85
Joined: Sat Mar 02, 2024 10:33 am
Languages: English N-C1
Finnish A0-A1
x 179

Re: AI stuff (was: How to properly do L-R method)

Postby bombobuffoon » Sat Mar 16, 2024 7:16 pm

Cainntear wrote:
bombobuffoon wrote:
tastyonions wrote:I’ve been assuming that most of the “AI tutors” out now are just GPT or one of its competitors with the thinnest of skins stretched over it. Eventually there will probably be bots trained specifically for the domain of language learning that don’t require as much direction from the student but that’s not reality yet.


It appears we are stuck in the never ending cycle of "when chatGPT version x comes out those problems will be solved".

Yup -- it's a failure to understand what ChatGPT does. ChatGPT does general chit-chat well because it's seen more general chit-chat than a human would encounter in multiple lifetimes. ChatGPT needs an *enormous* amount of data to do anything, and there simply isn't enough data on language lessons to train it.

ChatGPT does an easy task well based on unimaginable amounts of data.

Thinking that if it can train a computer to do an easy task with lots of data means you can train a computer to do a difficult task with less data is kind of silly.


It seems to me that these are interpolation machines. They are terrible at extrapolation though.
Therefore they need a lot of data to do interpolations but are awful at extrapolating outside their data sets. Kind of the opposite of what one might expect from an "AI".

The reality of the situation is obvious. "AI" with all its supposed "improving every day" technical potential, has already long ago hit a brick wall. And its also a commercial failure. It does not make fiscal sense that these sorts of things can be scaled up, let alone do so well enough that they will replace any existing systems.
2 x


Return to “General Language Discussion”

Who is online

Users browsing this forum: No registered users and 2 guests