Thoughts on Kyle Drake's stance on AI

[Seraphim]

I can appreciate the perspective that Kyle's piece takes, but I think there's some axises that we aren't considering in this conversation that are worth looking at, such as AI's impact on the environment, and also the fact that there's a big human workforce behind the training on AI models, whose work conditions span a range of quality.

Not even taking into account the ethical angles on scarping artist's work, those factors alone deter me from any consideration of generative AI. The environmental aspect is too big for me to ever really seriously considering it.

I have a good friend who uses LMMs to generate scenarios for our TTRPG game, and it constantly provides sludge and bad examples that he then needs to spend time on cleaning up, and I just can't find the logic in using those services. It feels like he's creating more work for himself instead of just thinking of a scenario from his own brain. 

[Capybara]

but he basically says we should embrace AI, including the use of AI "art" and using AI for therapy.

No he doesn't. I think most people in this thread have acknowledged that this blog post is a year old, but since it was written before AI boomed as much as it has, this is just optimistic curiosity about the future of it. People like him who work in tech fields tend to have a broader scope of what "AI and machine learning" entails, and have complete blind spots when it comes to the environmental impact or the ethics of scraping people's art/photography and writing. Most hobbyists tend to jump to AI "art" and "writing", when he clearly elaborated on more uses that aren't inherently bad. Earlier this year Neocities briefly had a chatbot of Daria as an April fool's joke, if you didn't know he has neutral-positive views on AI, you should've known earlier this year. I get wanting higher standards for owners of indie services, but I just can't see "Don't use Neocities, Kyle Drake supports AI" (oversimplified but not untrue) as any different from how people continue to use Twitter, Tumblr, services provided by Wix, etc...

That said, I still lean towards being against AI and that it needs to be regulated so artists and photographers won't be out of their jobs, the worst cases usages should be wholly illegal (ex. deepfakes, simulated illegal content), and that AI "art/writing" should be a reference tool instead of a replacement for art. I'm normally against the argument that a neutral tool shouldn't exist because people could potentially "use" it wrong, but I'm tired of the bad AI has done being overlooked because it makes funny shitposts or pretty pictures.

[JINSBEK]

First, I’m glad that Kyle Drake asks people to call it “machine learning” instead of AI. “AI” is a stupid buzzword used to sell the similar kind of technology our old flip phones had with predictive dictionaries. Calling it “intelligence” is insulting and misleading.

What everyone is afraid of is workflow automation. Right? The last start-up I was a part of was originally an AI R&D company. The founder and CEO, a man actually deeply concerned about the state of the environment for the future of his young son, was an “AI engineer” himself (that’s not the only type of software engineering he knows, by the way), and worked as a consultant for companies like Zapier. Long story short, it turns out machine learning was NOT the tool our userbases needed the most (or at all), and we ended up letting go every single AI engineer in the company. We pivoted away from AI, mostly because it added nothing to our product’s UX, and actively detracted from it. I wish we had axed ML from our product earlier—everyone wished that.

I’ve befriended photographers who report to me that they and their colleagues have lost 60% of their income to AI. I see other photographers who say they cater to a who has clientele has zero interest in AI, so it doesn’t affect them negatively. I befriended a photographer who actually now passionately creates detailed, Hiëronymus Bosch-esque murals using manual lassoing of areas (“in-painting”), continuous refining, and working within a manually set node-based environment, really not unlike any other node-based environment that has to be manipulated by anyone working in VFX. (Obviously, giant tapestries like these are separate from his photography work.) His wife is an art photographer and he has introduced her to his new ML passion-hobby; I sat in on him going through his workflow to create a composition about mortality, and she made an amusing little work, using the same techniques, on AI robots being used for money laundering. I befriended another artist who began his artistic career as a lapidarist, paints in physical media and was a pioneer and leading fractalist, a man who has been creating art for longer than I have lived, and now creates generative “digigraphs” in much the same complex, intentional workflow as my previous friend. (And he still does lapidary and works with other media, as well.) Last year, I had the privilege of coming in contact with a group of French avant-garde AV and graphic designers who use generative assets in many of their collage-based graphic design and video works. I’m also close to the games industry and hear of the coming and present bloodbaths due to ML being used to replace the artists who used to make promotional and concept art. And I have former colleagues who work with the cinema and TV actors who are rightfully concerned that they could be replaced by generative actors (or even impressions of themselves, if they give up the rights to their own likeness!), and the striking American TV and film writers who do not want to lose their jobs to machine-generated schlop.

Nobody except money launderers wants the (increasingly extremely obvious) pump-and-dump of mass-generated, lazy cashgrabs, least of all, actual artisans. Nobody wants to lose their jobs to this. Unfortunately, the nature of publicly-traded companies is to garner as much short-term revenue as possible, and that often means cutting all conceivable corners—and then some. We have now multiple generations of experience demonstrating that any new technological advancement will be seized on by capital in its relentless pursuit of quickening the pace of profit extraction and accumulation. Who is using the technology, and how? The same people will always try to take advantage, you know, and the same people will always get trampled on. The tool doesn’t matter, in a system that entrenches inequality.

Honestly, I love what machine learning can produce. When trained exclusively on paintings from the 18th and 19th centuries, then prompted with word salads (or beautiful prose), it can produce stunning surrealist, abstract illustrations. Humans don’t think like machines, these machines can’t yet think like humans, so they produce something hazy, unreal, “insane” yet of perfect sense to their own internal workings, and I genuinely love that. There’s also the inherent fun in generating shitposts and you can tell who goes into this with a Newgrounds type of stance, versus the hacks who just want to sell poorly generated pictures of kangaroos on Adobe Stock. It was part of my last job to identify and ban these false “artists”. The greed and stupidity bound up together was insufferable and insulting.

...I actually have a phrase I use; the "value of potentiality", e.g. what is the value of all the potential outcomes you can get from an interaction, what new things can you learn? what relationships can you build? what can you offer to the person or thing you are interacting with?

In almost all cases the value of potentiality from AI generated work is almost 0, whereas its quite high for any human made work. For example; If a human writes the dialogue for a game character; I can find out who they are, and contact them, saying how much I like their work (or hate it!), and who knows where that interaction could lead. We could becomes friends, they could marry my cousin, we could end up stranded on a desert island, or we could just have a short interesting conversation ~ all of those are are potential outcomes that add some value to our lives, and they still exist, even if none of them ever happen and we never interact! (That's why existing is fun!)

However if an LLM generates the dialogue of a game character, then there is no further interaction I can have; the potential outcome is 0, nothing, no further possibilities. This potential interaction is part of what makes engaging with art so fascinating and powerful to people, and its why content generated by an LLM will never really be a threat; even if the content is of a high quality, it cannot satisfy the human need for possibilities, and therefor it cannot satisfy the purpose of art It might seem flashy for a while, but what you'll quickly find is that people will gravitate towards art that has a higher potentiality (For example; Thats why YouTube beats TV, because your potential of having an interaction with a YouTuber feels more real than having an interaction with a TV show)

Yes, exactly! Humans learn from doing, from working and interacting with one another, from bouncing things off, experimenting, failing, refining, trying out new ideas and finally executing well. Convenience at the expense of human experience is something afforded by the technology around us, everywhere. It’s not just ML, everything is being prioritised and designed based on convenience. And ironically, the quality of our life suffers. I think that is a real problem and that cannot be addressed by simply value-condemning a technology which companies are going to attempt to leverage whether you like it or not. A struggle has to be waged for rules, regulations, and the right to live off of one’s artistry.

[Yoylecake420]

I really appreciate your nuanced takes on AI.
I may not agree with Kyle Drake's stance on AI (I'm generally anti generative AI, I don't like AI stealing people's art), but as long as he's not selling the data or using it for unethical purposes, he should be fine.
I just really hate people who are against AI without specifying which one. AI was made to be a tool, not to be replaced. I think what we need is to condemn the corporations using AI for theft or selling data. It's also worth noting that people mostly focus on the negative aspects of things.