@Gargron LLMs are more sophisticated versions of the old ELIZA chatbot. At least ELIZA's creator Joseph Weizenbaum did not attempt to overhype what it was.

https://en.m.wikipedia.org/wiki/ELIZA

@Gargron
Not the only danger, it's also turning human work and creativity into a seemingly unnecessary commodity.
The effect of which is already easy to see and guess, it will make things worse for all of us, whether you lose your job or need support in any form or like to enjoy art or....

@Gargron
Large Language Models are to intelligence what elevators are to teleportation.

@Gargron It's not the best term for sure, but there is clearly increasing complex reasoning of different kinds happening over the past 10 years or so.

@Gargron one word. #anthropomorphization

@Gargron I've been telling people this forever and they don't want to hear it. They're so enamored with the science fiction of "AI" that they don't want to believe that these companies will lie for money.

@Gargron
Are you saying the "revolutionary artificial intelligence technology" in this toothbrush is just marketing? That there isn't an Nvidia GPU inside analyzing my teeth in realtime? https://oralb.com/en-us/products/electric-toothbrushes/genius-x-limited-electronic-toothbrush-white/

@Gargron I think AI is a real discipline, but now people are calling any program they want to hype "AI".

@Gargron I try to avoid it when talking about LLMs.

@Gargron

I like "Zombie Algorithm."

@Gargron Yes. Anyone who has ever complained about autocorrect should know better than to think "AI" is anything but more of the same.

@Gargron I agree,
I've been using the words Statistical Model, but in my head it's much closer to "Sparkling Weighted Random Number Generator"

@Gargron AI has no intelligence. It’s a processing system.

@Gargron I think you’re pulling a giant “No true Scotsman” fallacy here. I agree that ChatGPT is not going to become sentient (anytime soon). But I think it’s reductive and an oversimplification to claim that there is no intelligence there. Not in the least because we don’t even know what intelligence is exactly.

What your Toot is also doing - as are the myriad of ignorant replies - is severely underestimating the usefulness of ChatGPT and related technologies.

@Gargron I've come across "virtual intelligence" and think I like that terminology more. The program is used to generate something which appears to have qualities of "intelligence" but is only a simulation of such. In in a similar vein, I think "machine learning" is also a little too anthropomorphic.

@Gargron GIGO. It's all about the training.

@Gargron Which is why I only refer to it as "large language models" or less often "machine learning," when it applies to VFX work in film and TV.

@Gargron I'm absolutely not an AI fanboi but we may have to reframe the conversation. Who says humans aren't statistics machines as well? It's not like earth shattering thoughts just pop into our minds without massive amounts of training, observation and mimicry.

@Gargron Remember that "beeyonaire" who owns the Republican judge? Harlan Crow? Well, being a lifelong Dickens fan, I realized Dickens himself couldn't come up with a better name. So I fired up my AIs and asked them to write a "Dickens" paragraph describing the character. They failed. So now I say, artificial--perhaps, but intelligence--not so much.

@Gargron Plagiarism machine works.

@Gargron

** S E A S O N ' S G R E E T I N G S **
~ Admiral Founder Rochko! ~

I think "auto-data" is better coining, or in general, like ChatGPT, a " generated data interplay (data-play)" might be the proper minting of the input exchange.

Good points. Furry Trunk Salute!

@Gargron
indeed , let's call it #SALAMI as dear Stefano Quintarelli suggested 🙂

https://blog.quintarelli.it/2019/11/lets-forget-the-term-ai-lets-call-them-systematic-approaches-to-learning-algorithms-and-machine-inferences-salami/

🖖

@Gargron Oh, the irony. Aren't you just parroting Emily Bender? A chatbot could have said this. How can we know that what you say is more than just your cognitive statistics?

1. There is intelligence in current LLM based AI. A different sort, but still intelligence. Language competence without comprehension.

2. Most of what people say is pretty much at the level of parroting.

3. What you say is half true, half misleading.

Several people on this thread have mentioned this sort of idea.

@Gargron I wish these bots would stop being personified. Statistical databases don't have personalities.

@Gargron
i say LMA (large model algorithm) and if people get confused by it they will ask, opening the door to me explaining how "ai" really is just a large model algorithm.
if they dont understand it, refuse to acknowledge that i am talking about the same thing and continue to talk about "ai" i know enough about their technical competence about the topic.

@Gargron https://inventaire.io/entity/isbn:9783960541905 a book, that goes to further detail about it.

@Gargron probably the single keenest and clearest articulation of this point that i’ve seen yet. no wonder you’re the mastodon guy

@Gargron sure, that’s true of 2023-era LLMs. But what about their descendants? We’re very early in this journey

@Gargron I'd argue that it is artificial intelligence, because intelligence doesn't necessarily have to include sentience or sapience. Sure they may go hand in hand, but intelligence without sentience or sapience can only reproduce or even restructure information, but cannot create or understand it.

@Gargron I had a similar opinion for a while, and still prefer ML, LLM, etc. But as long as Artificial General Intelligence (AGI) and Artificial Sentience are considered separate things … I’m becoming somewhat ok with the term AI as a catch-all for some of this new tech.

@Gargron
I always read "AI" in the news as "Artificial Idiocy." Although the terms "Augmented Idiocy" or "Amplified Idiocy" just came to mind.
@etherdiver

@Gargron aptly put

@Gargron

@Gargron

I've taken to calling them 'text/code completers' and what they do as 'text/code completion'.

@Gargron There's an old saying that if you don't know how something works, it's AI. Once you know how it works, it's an algorithm. For what it's worth the algorithms taught in an Intro to AI class are usually much less complicated than an LLM

@Gargron thank you dear god thank you.

It's like we need to be spending just as much time debunking the "ai" label as we do explaining all the problems with what they're using it for. 😫

@Gargron

Plagiarism engines.

Sparkling autocorrects.

"Machine learning"

statistical models. Fuck em.

@Gargron

On these issues I miss the wise words of the Ada Lovelace Foundation. Did they get lost in the Great Migrations?

@Gargron Oh my, what a deafening echo chamber! AI is not AGI. LLMs are not AGI, and nobody paying attention claims or believes they are. AI is very real, it is pretty much everywhere, it is incredibly useful, and at the same time dangerous, today.

@Gargron

Maybe coopt the acronym. Instead of calling it "artificial intelligence," insist on calling it something "applied inference," or something.

@Gargron But to be fair, it is doing a better job at pretending to be intelligent than tons of humans voting nowadays… so I’m not sure I really really care ablut the broad interpretation of “I” being used.

@Gargron Completely agreed.

@Gargron always thought a better term would be "extreme fitting models"

@Gargron I recommend we start calling AI "Behavioral Statistics" as they model statistically likely behavior. Plus the abbreviation is easily understood, even by someone who doesn't know what it stands for.

@Gargron I just went to a major US marketing conference in Sept, every single vendor was touting the 'AI-fication' of their products. It's ridiculous because we are customers, we know that they've just rebranded the ML stuff they were already doing!

Exactly, @Gargron! That's why I came up with Weizenbaum's Law:

“Whenever something is called Artificial Intelligence, someone is trying to fool you.”

https://digitalcourage.social/@chpietsch/109836875762133634

#AI #ArtificialIntelligence #WeizenbaumsLaw #dontCallItAI

@Gargron essentially it is often used to bullshit. So AB for artificial bullshiting would be much more fitting or automatic bullshiting. I think the last is the best option.

@Gargron nice video from Sabine: https://m.youtube.com/watch?v=cP5zGh2fui0&pp=ygUPaG9zc2VuZmVsZGVyIGFp

@Gargron Disagree. Calling it "artificial intelligence" is absolutely justified. It's only when the qualifier "artificial" is dropped, unknowingly or uncareingly, that interpretive problems arise.

@Gargron You are also "just statistics" - according to all the science we have on the subject.

@Gargron well put!
What we have today is limited to Machine Learning

@Gargron Okay. What do you recommend we call it?

@Gargron how our brain works it is also statistics in a way, but yes biological beings are still at the order of magnitude more complex than some 2023 super computer. FYI we wrote a little blog that reviews the key conceptions and the key misconceptions: https://medium.com/mlearning-ai/ai-technology-what-it-is-and-what-its-not-and-how-it-can-potentially-help-us-solve-the-climate-9b2f428eff0

@Gargron word. It's not AI. It's software written to steal your *everything*