AI today is basically a mechanical school. AI students are trained to give a specific answer. This is at the heart of how all machine learning works, including generative AI. Even image generators do this.
“Here’s a million examples of what the pixelated representation of a hand looks like; now go and make a derivative copy.”
This is fine for objective facts, like physics and history. It is useless for art.
Merely drawing a hand is not art, it’s an objective truth (do typical humans have 5 or 6 fingers?). But art school is not about objective truths. Art school teaches creativity. Specifically challenging ideas and expression.
AIs can’t fundamentally challenge ideas and express themselves because they lack personal experience, personality and individuality.
Society at large has been fooled into thinking that speech (LLM) and other generative AI lead to AGI. But the reality is that these models have more in common with encyclopedias and stock image libraries than intelligence.
I’m not convinced your take is different - drawing an accurate sketch of a hand isn’t art, telling AI to generate a hand isn’t art, it requires someone creative or expressing something to be art, regardless of the medium(s), including diffusion/noise removal models being a medium.
Nobody’s going to claim illustrator or inkscape “made” your graphic design, so why claim the same for AI - doing so just shows you don’t understand the medium or what goes into finetuning models, parameters, inpainting, step control of loras, block weights, noise removal level and regional prompting and all the other things that differentiate a piece of AI-generated art from AI slop (not to say that you have to use all of these for it to be art, just that once you do it probably passes the threshold for it to be art)?
I have a different take.
AI today is basically a mechanical school. AI students are trained to give a specific answer. This is at the heart of how all machine learning works, including generative AI. Even image generators do this.
“Here’s a million examples of what the pixelated representation of a hand looks like; now go and make a derivative copy.”
This is fine for objective facts, like physics and history. It is useless for art.
Merely drawing a hand is not art, it’s an objective truth (do typical humans have 5 or 6 fingers?). But art school is not about objective truths. Art school teaches creativity. Specifically challenging ideas and expression.
AIs can’t fundamentally challenge ideas and express themselves because they lack personal experience, personality and individuality.
Society at large has been fooled into thinking that speech (LLM) and other generative AI lead to AGI. But the reality is that these models have more in common with encyclopedias and stock image libraries than intelligence.
I’m not convinced your take is different - drawing an accurate sketch of a hand isn’t art, telling AI to generate a hand isn’t art, it requires someone creative or expressing something to be art, regardless of the medium(s), including diffusion/noise removal models being a medium.
Nobody’s going to claim illustrator or inkscape “made” your graphic design, so why claim the same for AI - doing so just shows you don’t understand the medium or what goes into finetuning models, parameters, inpainting, step control of loras, block weights, noise removal level and regional prompting and all the other things that differentiate a piece of AI-generated art from AI slop (not to say that you have to use all of these for it to be art, just that once you do it probably passes the threshold for it to be art)?
Transition periods are usually hard, for some
I agree completely.