this post was submitted on 20 May 2025
24 points (76.1% liked)
Fuck AI
2825 readers
993 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
If we take the forum title here, the "fuck" is directed at the people in charge of so-called "AI" companies. The technology has value. It's just being forcefed down our throats in ways that remind us of block chain and whatever happened to block chain?!
The tech with the most push behind it is being pushed in infancy, and is damn near useless without datasets of entirely stolen content.
There's genuinely useful things and impressive tech under the machine learning umbrella. This "AI" boom is just hard pushing garbage.
This past week we saw the most obvious example yet of why they're pushing LLMs so far too, Grok's unprompted white supremacist ramblings over on twitter. These tools can easily be injected with biases like that (and much more subtly too) to turn them into a giant propaganda machine.
Some of the tech is genuinely useful and impressive, but the stuff getting the biggest pushes is nothing but garbage, and the companies behind it all are vile.
It's fortunate that Kaptain Ketamine had his little binge of his favourite drug and made it SO OBVIOUS. There's subtle biases all over degenerative AI. Like there was a phase when trying out the "art" creators where I couldn't get any of them to portray someone writing with their left hand. (I don't know if they still have a problem with that; I got bored with AI "art" once I saw its limitations.) And if the word "thug" was in the prompt it was about 80% chance of being a black guy. Or if the word "professional" was in the prompt it was about 80% chance of being a white guy. EXCEPT if "marketing" was added (as in "marketing professional"). Then for some reason it was almost always an Asian woman.
Or we can look at Perplexity, supposedly driven by not only its model, but incorporation of search results into the prompt. Ask it a question about any big techbrodude AI and its first responses will be positive and singing the praises of the AI renaissance. If you push (not even very hard) you can start getting it to confess to the flaws of LLMs, diffusion models, etc. and to the flaws of the corporate manoeuvring around pushing AI into everything, but the FIRST response (and the one people most likely stop reading after) is always pushing the glory of the AI revolution.
(Kind of like Chinese propaganda, really. You can get Party officials to admit to errors of judgment and outright vile acts of the past in conversation, but their first answer is always the glory of the Party!)
Oh, and then let's look at what's on the Internet where most of the data gets sucked up from. There's probably about three orders of magnitude more text about Sonic the Hedgehog in your average LLM's model than there is about, oh, I don't know, off the top of my head, Daoism, literally the most influential philosophical school of the world's most populous country! Hell, there's probably more information about Mario and Luigi from Nintendo than there is about the Bible, arguably the most widespread and influential book around the world!
I wonder how that skews the bias...?