would you drink the serum with timber wolf estrogen

wait... @gamermcbaggin.bsky.social do you see it....

"Machine learning" can only make sense to people who do not value or grasp learning experiences such as meta-learning, generalization, or internalization. "Machine learning" equates learning to statistically improved quantifiable results. Better stats. Number going up. No wonder capitalism loves it.

If you wanna ban all "AI generated content", you'd have to ban a lot of movies that got made before the current AI shitboom: en.wikipedia.org/wiki/MASSIVE...

Which I guess is doable but isn't that just panicking and overcorrecting what is, in the end, empty marketing verbiage?

A lot of the same tools used by "Bad AI" are used in procedural generation, texturing, animation engines that try to simulate physics use math that could be called "AI" - because the only real criteria for calling a program "AI" is if it was a tool developed by teams trying to produce Actual AI.

Having the choice of going to a medical doctor be one filled with anxiety, with painful strategic cost/benefit thinking and a constant fear of being dismissed patronizingly doesn't seem too good for the general state of Society.

I know, I'm very insightful and original.

And while we should root out and fight against those malformed thoughts and the entirety of the "AI" marketing coup, we ain't ever gonna get a responsible contract that incorporates such an understanding into our contracts.

We need to define, precisely, what is "Bad AI" and why it is bad.

Machine learning image generation is merely a very inefficient way to do exquisite corpse style remixes of existing illustrations.

But what we know call "AI", I realize this just now, is not just the program. "AI" is also the cultural phantasms, metaphors and semantic abuses around the tools.

And frankly, I think being pissed off at the "AI" part is superfluous here. Fucking people over by claiming their work as your own is already bad and we already have (shitty, bourgeois) laws about it.

Being angry at the obfuscation tool and not the dipshit using it seems very misguided.

So a possible way to split "bad AI" from boring old stochastic programming tools from the 70s is by the data/computing costs.

But a lot of the labor problems of "bad AI" comes from the fact they are tools for creating barely distinct remixes and collages of existing images. Theft, rather than cost.

Before the current "AI" shitboom marketing coup, google and similar companies used the term "Big Data". Big Data is not the same as "AI", but most "bad AI" depends on big data to function. Big data is also the source of the absurdly high computation and resource costs of "bad AI".

But since the term "AI" is fucking vague and mostly meaningless, and with no clear referent, how exactly do you define what is and isn't allowed? We're in a "know it when I see it" phase, but that's not good enough. The field being full of malicious misinformation doesn't help either.

Concretely: We want an enforceable way to stop the flood of "AI" shit into illustration, writing and videogame communities and storefronts. We want enforceable ways to cut out bad agents from those communities and issue refunds.

I don't want Steam to be flooded with gen"AI" shitslop. I hate bloops7

If you wanna ban all "AI generated content", you'd have to ban a lot of movies that got made before the current AI shitboom: en.wikipedia.org/wiki/MASSIVE...

Which I guess is doable but isn't that just panicking and overcorrecting what is, in the end, empty marketing verbiage?

A lot of the same tools used by "Bad AI" are used in procedural generation, texturing, animation engines that try to simulate physics use math that could be called "AI" - because the only real criteria for calling a program "AI" is if it was a tool developed by teams trying to produce Actual AI.

If you're petulant enough, one can call ELIZA "AI generated content", even thought it is a innocuous program. Or how about markov chains? Which while not ML, can be used to generate text. Caves of Qud uses markov chains to great effect, but nobody sane would call it "AI generated content".

How exactly do you enforce a ban on "AI Generated Content" when the term "AI" is bullshit? What most people mean by "AI" these days is Machine Learning and its subsets. And machine learning is, once the mystification is gone, a way of creating weighted stochastic functions in a search tree.