I offer Cassandra's Complete Class Theorem¹.
All "good use cases for AI" break down into one of four categories:
I. Bad use case.
II. Good use case, but not something AI can do.
III Good use case, but it's already been done without AI.
IV. Not AI in the LLMs and GANs sense, but in the sense of machine learning, statistical inference, or other similarly valid things AI boosters use to shield from critique.
https://wandering.shop/@xgranade/115766140296237983
___
¹Not actually a theorem.
I had to add Category IV in my self quoot to cover that boosters kept showing up and claiming that things that have been done for years and years are evidence of why LLMs and image-generation GANs are useful. You know, the same ol' bait and switch I pointed out the other day.
I like how you logically put it. For me, it comes down to "Does the technology do more good than bad and is it a net positive for society and/or the environment?" I honestly do not believe what is being marketed to us as "ai" does enough good to justify its existence at this time, outside of a research center. It might be useful some day, but people are literally dying because of how it is being sold. The world is on fire and they are only adding more fuel by building data centers.
This is a very good point, and something I need to keep at the front of my mind. Even if there was somehow a magic use case for LLMs and other AI bullshit machines, that would not justify:
• Making fascists rich.
• Displacing labor rights.
• Fucking over the environment.
• Enclosing culture behind corporate ownership.
• Giving fascists a giant disinfo machine.
The "Thermodynamics of Bullshit"
Environmental: Every "funny" AI-generated image of a cat in a tuxedo is a literal withdrawal from the planet’s cooling systems. It is Carbon-to-Cringe pipeline engineering.
Labor: AI isn't "automating" work; it is laundering it. It takes the collective output of human culture, strips the names of the creators, and sells it back to them in a distorted slurry.
The Fascist Subsidy: every token generated effectively transfers wealth to the massive corporate/fascist structures that own the compute.
Enclosure of Culture:GenAI is just a giant "copy-paste" machine built to enclose human creativity behind a subscription wall.
Environmental Nihilism: Every "low-effort" AI image is a glass of water poured into a server's cooling system while the planet burns.
[rude laughter]
Theorem¹: wouldn’t that make it Cassandra’s Complete Class Conjecture?
@zkat That's fair, and I definitely bang on that drum as well, but I'm also not above shitposting at the special pleading from the slopbros.
Perhaps I should be...
@codinghorror @zkat Maybe put differently, I don't see any excluded middle here worth saving, and for the same reason I don't see an excluded middle between "opposes fascism" and "is fascist."
The drive to always seek out an acceptable moderate position is very exploitable, and that is something AI boosters are very wont to do.
The problem here is the IV cases. For example, I’d be hard pressed to find harms caused by Mozilla’s language-translation models:
And, in the current marketing environment, things like that are also branded ‘AI’.
When the models can be trained ethically, machine learning typically does well in places where there is no harm in a wrong answer but significant value in a correct one. CPU branch predictors now often use neural networks. If they give a wrong answer, the CPU does a small amount of work that it throws away. If they give a correct answer, the CPU does useful work instead of idling. Getting the right answer 95% of the time gives a 10x or better speed up relative to not doing it. This is a great place to use machine learning. But a lot of the places where it’s proposed have significant negative real-world consequences from wrong answers.
@oxi There's a reason you don't train classifiers on synthetic data when you don't understand how the simulator works.
Based on having done a bunch of ML shit back in the day, I'm gonna guess that's a Category I.
@xgranade I have used one at work (we were told to). The positive use I found simply saved me some time - I could have done the work, but decided to let the AI do it, so I could be a Good Little Employee.
Another time we tried, and it didn't solve the problem at all.
I have so far been unimpressed. Anything I have found of use I should have been able to find using a decent search, but of course, these have all been shittified with AI, so no longer do their job properly.
@xgranade weather forecasting, using AI to infer areas that can't be measured using contextual and historical data. NOAA does this, and it's great. Saves lives. Lots of applications like that.
LLMs are pure trash.
@xgranade okay. now make marx sing worker's songs, I'll wait.
That said, I'm absolutely against 99% of the western AI hype.
As a marxist, I know that a tool is not to be thrown out because its main wielders are fascists.
We shot fascists with their own guns. We use fascist social media to recruit new people.
@xgranade
I am tempted to simplify this by saying that the great use cases for AI are exactly the ones which existed before the current hype of LLMs and Generative AI, when the field was in the hands of academics and researchers, not controlled by investors and the worse of the worse of the VCs.
Things ranging from contextual translation to image recognition to optimization to genetic algorithms and evolutionary computing. When running fast and fewer resources were the goals.
Cassandra is only carbon now
Bryan (he/him) 🇺🇸 🏳️🌈
F4GRX Sébastien
tuban_muzuru
Stu
divVerent
Hashbangperl
The Long Tail
Daniel Terhorst-North
Arvid Requate
Mike Anderson
Kat Marchán 🐈
Jeff Atwood
David Chisnall (*Now with 50% more sarcasm!*)
Alex Wilson
Usagi
./coldwave -v 
pascal guldener
Jeremy List
Oxı̇
Niedliche Nacktschnecke
Schroedinger
Quinn Norton
haui 
Matt Boehm
dr_a
Mark von Übelgarten
Jeremy Kahn
cthos 🐱