Chomsky and AI

General discussion about learning languages
User avatar
tastyonions
Black Belt - 1st Dan
Posts: 1551
Joined: Sat Jul 18, 2015 5:39 pm
Location: Dallas, TX
Languages: EN (N), FR, ES, DE, IT, PT, NL, EL
x 3795

Re: Chomsky and AI

Postby tastyonions » Mon Mar 20, 2023 9:42 pm

Cainntear wrote:That's machine learning, not AI.

But we call it AI in common parlance because that's sci-fi.
And we call it AI in marketing because it sells well.

We call it AI because it does tasks we normally think of as requiring intelligence, like playing chess, writing poetry, summarizing articles, solving math problems expressed in natural language, making pictures, writing music, explaining concepts from physics, and even telling jokes.

I suspect that as it gets better at all of these things and adds even more impressive feats to the list, the goal posts for which things are viewed as sine qua non of "intelligence" will continually get moved so that chatbots will never be seen as "intelligent" so as not to wound our pride.
2 x

Cainntear
Black Belt - 3rd Dan
Posts: 3449
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 8601
Contact:

Re: Chomsky and AI

Postby Cainntear » Mon Mar 20, 2023 10:14 pm

tastyonions wrote:We call it AI because it does tasks we normally think of as requiring intelligence, like playing chess,

Yes, and the first time a computer beat a grand master was when I was in uni, and my lectures went to great lengths to stress that the news was incorrect in attributing the success to AI. Computer chess was a brute force instrument: it wasn't a matter of being smart, but just of having enough memory and processing power to map out enough possibilities of "if the player does this, then I do that, then it does that ... then I will win".

AI isn't the path to being good at chess.
writing poetry, summarizing articles, solving math problems expressed in natural language, making pictures, writing music, explaining concepts from physics, and even telling jokes.

That's natural language processing. The fact that we've moved on to machine learning to do it doesn't justify the use of the term artificial intelligence. It's a sci-fi term that doesn't accurately describe real computing.
I suspect that as it gets better at all of these things and adds even more impressive feats to the list, the goal posts for which things are viewed as sine qua non of "intelligence" will continually get moved so that chatbots will never be seen as "intelligent" so as not to wound our pride.

Or just... you know... stop calling it A.I....?
0 x

User avatar
tastyonions
Black Belt - 1st Dan
Posts: 1551
Joined: Sat Jul 18, 2015 5:39 pm
Location: Dallas, TX
Languages: EN (N), FR, ES, DE, IT, PT, NL, EL
x 3795

Re: Chomsky and AI

Postby tastyonions » Mon Mar 20, 2023 10:18 pm

Well, that's precisely my point. Even if we had a robot with full human (or super-human) perceptual abilities, capable of walking around, painting, improvising music, speaking, creating, working a fulltime job requiring human interaction, and doing everything else a person does, you and others would still be saying, "Bubububut it's nothing but a bunch of silicon doing natural language processing / statistics / following algorithms, that's not *real* intelligence!"
1 x

User avatar
sfuqua
Black Belt - 1st Dan
Posts: 1642
Joined: Sun Jul 19, 2015 5:05 am
Location: san jose, california
Languages: Bad English: native
Samoan: speak, but rusty
Tagalog: imperfect, but use all the time
Spanish: read
French: read some
Japanese: beginner, obsessively studying
Language Log: https://forum.language-learners.org/vie ... =15&t=9248
x 6298

Re: Chomsky and AI

Postby sfuqua » Mon Mar 20, 2023 10:28 pm

ChatGPT corrected my question this morning when I called it a large language model. It said, "you must mean neural network"
1 x
荒海や佐渡によこたふ天の川

the rough sea / stretching out towards Sado / the Milky Way
Basho[1689]

Sometimes Japanese is just too much...

Cainntear
Black Belt - 3rd Dan
Posts: 3449
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 8601
Contact:

Re: Chomsky and AI

Postby Cainntear » Mon Mar 20, 2023 10:32 pm

tastyonions wrote:Well, that's precisely my point. Even if we had a robot with full human (or super-human) perceptual abilities, capable of walking around, painting, improvising music, speaking, creating, working a fulltime job requiring human interaction, and doing everything else a person does, you and others would still be saying, "Bubububut it's nothing but a bunch of silicon doing natural language processing / statistics / following algorithms, that's not *real* intelligence!"

Bubububut I studied artificial intelligence for a couple of years.
2 x

User avatar
Iversen
Black Belt - 4th Dan
Posts: 4759
Joined: Sun Jul 19, 2015 7:36 pm
Location: Denmark
Languages: Monolingual travels in Danish, English, German, Dutch, Swedish, French, Portuguese, Spanish, Catalan, Italian, Romanian and (part time) Esperanto
Ahem, not yet: Norwegian, Afrikaans, Platt, Scots, Russian, Serbian, Bulgarian, Albanian, Greek, Latin, Irish, Indonesian and a few more...
Language Log: viewtopic.php?f=15&t=1027
x 14924

Re: Chomsky and AI

Postby Iversen » Mon Mar 20, 2023 11:13 pm

Iversen wrote: ... so maybe it would be simpler just to ask why "to happen" only is allowed in the third person?

Cainntear wrote:Have you never happened upon structures like this...? :twisted:

Not without "upon" (I also mentioned that "happen" occurs in the construction"make happen", but then you need "to make")

Iversen wrote:As for AI: when I first heard about it some years ago the explanation was that you let a computer run through thousands upon thousands of experiments which mostly fail, but then the computer learns which ones are the acceptable ones and build its next steps on that - and the process is so complicated that no human can follow it in detail.

Cainntear wrote:That's machine learning, not AI. But we call it AI in common parlance because that's sci-fi. And we call it AI in marketing because it sells well.

OK, correction - it was the second time. The first time I saw the expression used about attempts (or rather aspirations) to create artifical intelligence there were some people who thought it could and should be programmed in detail by a horde of engineers in white overalls. Among other things that was how some thought search machines and translations programs should be ccreated - but this would of course have been an unsurmountable task.

Then some clever people invented neural networks that could do countless failed experience by themselves, get them evaluated somehow and learn from the the few that succeeded. The price was that no human being had an inkling about the internal rules the machine had developed before it could walk up steps or play chess. I have seen a very simple version of this with a virtual robot that should learn to walk - and the criterion was that it should move along and not tumble over. And it went through more weird walks than even John Cleese could have imagined, but ended up with a version that worked. I have also seen reports on programs that can propose diagnoses based on answers to questions and medical test results. Such programs have to be trained on loads of input plus some kind of success criterion and evaluation, but ultimately they learn to propose diagnoses that aren't worse than those you get from a short telephone consultation - at least that's the claim. And all those things were categorized as AI, but of course they build on machine learning (that's the new thing about neural networks: they can learn).

And now it seems that people try to connect a neural network with virtual tentacles that extend into the world wide internet in the hope that it will look as if they tapped the real world out there. And the result is chatbots that apparently already can fool a gullible Turing machine (otherwise nobody in the school systems would have reason to fear the chatboxes). The next step where the machines begin to 'think' unexpected thoughts by themselves and not just because they eat humanmade garbage, and maybe they'll also develop some kind of self reflection - and THEN I think we have to recognize those processes as a non-human kind of intelligent thinking.

By the way human thinking is just the result of a lot of neurons firing more or less haphazardly .. but we judge it on the output.
5 x

User avatar
Le Baron
Black Belt - 3rd Dan
Posts: 3480
Joined: Mon Jan 18, 2021 5:14 pm
Location: Koude kikkerland
Languages: English (N), fr, nl, de, eo, Sranantongo,
Maintaining: es, swahili.
Language Log: https://forum.language-learners.org/vie ... 15&t=18796
x 9315

Re: Chomsky and AI

Postby Le Baron » Tue Mar 21, 2023 12:16 am

The Chat boffins already said to news organisations that they fiddle with the chatbot's output for certain reasons. Several times they've altered certain responses the chat makes in response to ways users have probed it with specific questions and lines of argument.

If any good could come out of it I'd hope the chatbots would inform many of its users to sell their internet devices and reconnect with the world they live in. Instead they seem hell-bent on devising 'tools' to end as much as possible human input into social organisation. Like a train station employee begging the management to install ticket machines and self-service advice (which fails to answer nuanced questions) so they can finally enjoy unemployment.

Once it has beat everyone at chess and and can chat with everyone (because they'll have not much else to do), what will we have achieved? Maybe someone will remember that thinking was fun and challenging Like those people who rediscover cooking. I want people to understand me better, not machines.
8 x

lowsocks
Orange Belt
Posts: 134
Joined: Mon Oct 23, 2017 4:00 am
Location: Southern Ontario, Canada
Languages: English (N), French (beginner), German (beginner)
x 353

Re: Chomsky and AI

Postby lowsocks » Wed Mar 22, 2023 9:09 pm

Cainntear wrote:
tastyonions wrote:Well, that's precisely my point. Even if we had a robot with full human (or super-human) perceptual abilities, capable of walking around, painting, improvising music, speaking, creating, working a fulltime job requiring human interaction, and doing everything else a person does, you and others would still be saying, "Bubububut it's nothing but a bunch of silicon doing natural language processing / statistics / following algorithms, that's not *real* intelligence!"

Bubububut I studied artificial intelligence for a couple of years.
All right, I will ask: What qualifies as Artificial Intelligence? Apparently, machine learning does not. What about, say, neural networks and "deep learning"? Or, going back decades, the use of Prolog? Or resolution methods in logic, or A* searches, or alpha-beta pruning? Are any of these considered particularly relevant to artificial intelligence? Or just peripheral?

Note that although I have some slight idea of these topics, it is pretty limited. Please try to avoid getting too technical ;)
0 x
One need not hope in order to undertake, nor succeed in order to persevere.

Cainntear
Black Belt - 3rd Dan
Posts: 3449
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 8601
Contact:

Re: Chomsky and AI

Postby Cainntear » Wed Mar 22, 2023 9:43 pm

lowsocks wrote:All right, I will ask: What qualifies as Artificial Intelligence? Apparently, machine learning does not. What about, say, neural networks and "deep learning"? Or, going back decades, the use of Prolog? Or resolution methods in logic, or A* searches, or alpha-beta pruning? Are any of these considered particularly relevant to artificial intelligence? Or just peripheral?

Note that although I have some slight idea of these topics, it is pretty limited. Please try to avoid getting too technical ;)

Ok, so my recollection is that AI was originally applied to neural networks because they were genuinely trying to mimic brains. There was a lot of talk in my university about the fact that people were overapplying the term and that the AI department were kind of stuck named that way for historical reasons but most of what they did wasn't actually "A.I." because neural networks were taught and researched, but were something of a dead end because they were extremely inefficient and couldn't really achieve much.

Machine learning was a catch-all for systems that didn't attempt to recreate biological intelligence but could extract rules from patterns (that were often missed by humans).

The resurrection of neural networks does superficially recreate the original idea that led to the adoption of A.I. as a term, but I personally don't think that neural nets really mimic biological thought closely enough to justify the term. They've come back in because modern computers are so fast and powerful that a neural net can actually achieve useful stuff. But the way I see it, they've managed to brute-force the tech by making very shallow networks that don't really mimic the complexity of biological neurology at all, and they have effectively become just another form of machine learning, because they're just doing hidden maths.
2 x

User avatar
ryanheise
Green Belt
Posts: 459
Joined: Tue Jun 04, 2019 3:13 pm
Location: Australia
Languages: English (N), Japanese (beginner)
x 1681
Contact:

Re: Chomsky and AI

Postby ryanheise » Thu Mar 23, 2023 3:11 am

Timeline:

1943: Warren McCulloch and Walter Pitts created the first computational model of neural networks. This work was to later inspire two branches of study, one attempting to understand how the brain works, and the other to find practical applications for computers.

1956: John McCarthy coined the term Artificial Intelligence. Although he didn't have a solid definition of what "intelligence" itself actually is (and we still don't today), he describes the ultimate effort as making computer programs that can solve problems and achieve goals as well as humans, but expressly using methods that are not limited by those that are biologically observable. His use of the term A.I. was broad enough to include all kinds of chess and checkers solvers that existed then (i.e. not limited to neural nets), and to include all branches of A.I. that exist today.

1959: Arthur Samuel coined the term Machine Learning to mean programming computers to behave in a way which, if done by humans or animals, would be described as involving the process of learning. He named neural nets as an example.


But these terms were coined almost a century ago at the beginning of a long and slow exponential curve that has only very recently started to take off, with most developments having occurred within the last decade. Leading up to this point, A.I. had progressed in cycles beginning with dreams and hopes, preceding disappointment, preceding actual scientific progress. This also means that depending on the decade in which you attended university, you might have experienced a different flavour of what was considered popular and unpopular at the time. If you walked into a computer science department today, vs the 2010s, vs the 2000s, you would likely get a different sense of how people felt about the terms Machine Learning and Artificial Intelligence.

But as someone who is more interested in the science itself than the terminology, I tend to use the whatever terminology is actually in use. To do otherwise would make it very difficult to read and understand the scientific literature.
5 x


Return to “General Language Discussion”

Who is online

Users browsing this forum: No registered users and 2 guests