We are John Henry fighting the steam-drill. We just don't know how fast it's coming. And personally, it doesn't do to tell me people will want to keep me on as a kind of artisan or artifact. I want to make very good things, not just be liked. I want to be good at this.
I honestly don't think we are! It's really looking more and more like the kind of AI shit that's being pushed right now is reaching dead ends left and right, and it's only a matter of time until the industry collapses in a big way.
John Henry was racing a steam-drill that actually did what it was supposed to do. We artists are racing a steam-drill that can barely drill, can't drill as deep, and mostly drills in the wrong locations. And the corporate ghouls pushing for it everywhere are too out of touch to care
that it doesn't work, but that inefficiency is catching up with them in the end regardless. AGI is nowhere in sight, an actual profitable system of monetization is nowhere in sight, Sam Altman thinks a quarter of the world's population is gonna pay $200 a month for his slop machine!
Shit sucks right now, and the extent to which this stuff is being pushed still is extremely depressing, but it's also proving less and less desirable to use from a financial and practical viewpoint, and that bubble is getting awful, *awful* close to bursting...
don't forget that they'd still be losing money on that, because they lose money on $200/month subscriptions.
A big difference is that the steam drill was actually competent at driving spikes. The jury is still out on whether *in practice* (to heck with benchmarks) gen-AI will reach that level. Right now it's mostly just good enough to lul people into falsely thinking it is reliable when it isn't.
i was told that this steam-drill was going to make a full-length studio ghibli lord of the rings movie for pennies on the dime and was super easy. i still haven't seen it.
like yes there are instances where ai was tempered and good enough to fool the average person, but snake oil and magick crystals never replaced modern medicine despite how many people constantly fall for it
Zach is, unfortunately, extremely credulous about AI as a topic, despite people repeatedly pointing him towards reasoned and extensive critique of the technology and it's limitations.
i suppose i should clarify that it's genAI that i'm referring to because there are uses of ACTUAL ai that got poisoned by the unfortunately common buzzwordy use of the term (like i inadvertently did in an earlier post). there are scientific advancements in ai, just not in the ones being shilled
I've read the critiques. I have a stack of books I've read on this about a foot high. I buy the concerns and the skepticism about e.g. the singularity or whatever, but e.g. that last article I posted has fans of a folk singer who couldn't tell her from AI.
As for video, no it can't make a full movie yet, but compare veo3 (which is now all over Facebook and YouTube) to the near-gibberish that was possible with state-of-the-art models 2 years ago.
Like yeah, there are limits, maybe hard ones, but from where I sit the artists are losing ground on a monthly basis if you do nothing else than consider what the public wants en masse.
so if ai output is outpacing artists, then what will happen when the ai inevitably feeds off of itself? is it going to remain as perfect as it's envisioned?
I don't think we know that. I agree (or anyway hope) the arts are less amenable to synthetic data than e.g. math or coding.
i've seen ai-generated garbage at markets that i guess do kind of well, but if anybody can do it, then it just spoils the selection that much more. it happened with the ghibli trend where everything wound up being yellow and then needed ai to fix it. that sort of fixing doesn't seem intuitive.
We do know, though! We know that "synthetic data" has an extreme deleterious effect on LLMs and image generators, this is KNOWN. They've already demonstrated it.
I disagree. When training an image-processing AI (traditional, pre-COVID kind) having to produce synthetic data is not the exception. This can be as simple as rotating the image slightly or as complex as making collages and curing the seams (is it a wonder then that GenAI images look like they do?)
It already is
???? From where I sit people are becoming increasingly hostile to Gen AI in the same timeframe. I havent read an article or had a conversation with a real person about Gen AI in months where it wasn't referred to as slop. If you want to devalue yourself go for it. Dont devalue the rest of us.
The public doesn't know what it wants and is driven significantly by elite signalling. You not saying that "AI is trash" is directly driving their acceptance. Also, AI is trash, so that shouldn't be hard to say.
People immediately pointed out the album had a tiny number of views - it fooled, like, two people. People have explained the economic issues behind the AI bubble and you consistently ignore them. Like I said: you're very credulous.
Only two of her fans were fooled, that we know of. The other musician interviewed said he found out about the AI copycat because fans started telling him his new album was weird and bad. I think you're overselling it.
Oh, sure - but everything I said remains the case. He buys into the marketing, sadly.
The problem isn't a binary thing. AI may not replace all animators. But what happens if AI allows one animator to do the work of 2?
then we aren't there yet and going into hypotheticals. ai has already been replacing jobs but not in a way as efficiently as the people in charge have hoped. the ai being implemented needs a babysitter of equal skill to watch over it. right now this sort of wishful thinking is a scam.
Except for voice acting. AI voices had already decimated a huge swath of it.
voice acting might be disproportionately affected, but i still don't see it being the end-all, be-all substitute or that this is a losing battle when the main companies replacing them are ones that are doing legally dubious things anyway
It's in techbros' best interest to convince you that AI is both here to stay and has the utility of something like a steam drill, because it lets them conveniently ignore all the evidence that genAI is floundering simply by insisting "it'll get better" until people give up arguing with them.
As it stands, there's not really any evidence to support the idea that it's here to stay aside from a handful of boring niche use cases. The only people insistent that it will stay are the people who benefit from its use, and there's no reason to trust them anyway so who cares what they think?
They use a similar argument when they insist that AI will magically gain consciousness as an excuse; most cases of "ChatGPT begging for its life" are the result of people prompting "beg for your life" or some such, but they get cranked out so Joe McDumbass can point and go "SEE?! We got RESULTS!"
Any suggestion that these results are faulty can be easily binned just by going "well, it'll get better eventually, so don't stand in our way!" Rather conveniently for the people chanting that, it never seems to get any better.
"I want to be good at this." i don't understand what this means here. Is this implying you can't be good if you don't use AI? Or implying that AI potentially getting 'better' at art than you some how stops you from being good at art in your own right? What does this mean?
I believe he means good at art. If "anything done by a human hand is ok", then you can literally draw and write like a 5-year-old and you'll check that box. But if your goal is to be good at the craft, you have to understand perspecive, anatomy, nuance, intent (you know, the stuff AI struggles with)
I understand that he means good at art, I just don't understand what that has to do with AI. Your suggesting he's saying that literally -any- human made art will be satisfactory for people who don't want AI, and that he want to be better than just 'a human made it'?
yep, pretty much how it works with souvenirs: "Oh look at this pretty scale model of the city…can see it was done with an injection mold though" "But honey, this other one is HAND MADE! That automatically makes it BETTER!" "Didn't we come to see the specific building they left off?" "But HAND MADE!"
Okay, so. If I need art made for my twitch channel, for an event announcement or a asset for stream. I'm going to expect it to be good quality, AND I'm going to make sure it's done by a human. My standards arnt reduced because I won't use AI. I still need quality work. Art isnt by default a novelty.
That take also barely makes sense in context, because AI still has nothing to do with if ones art is Good or not. The AI's quality is irrelevant if you goal is to be good of your own merits.
All I see is report after report about how the move to AI was ineffectual and expensive. Programmers who use it make more errors, customers hate it in customer service, and it can't do simple math. Not sure that it's going to do much better in art. I'm sure it will be used, just not a replacement.
The hype curve has been particularly helpful to me in navigating this whole "Hey, how about we change what AI means?" thing https://en.wikipedia.org/wiki/Gartner_hype_cycle
The stuff I'm working on now, in poetry and in comedy, is about as hard as I can run my brain on creativity. And part of that is just thinking, you know, let's race the steam-drill as long as we can. This seems to me the proper attitude and I don't think it has to be a negative one.
I don't know if I'm allowed to discuss this subject or if I look like I have 7 heads and 10 horns, but I don't think visual/textual artists are going extinct for the same reason that luthiers still exist today, while you can build 100 instruments a day on an assembly line in China with much...
higher precision. Granted, most string instruments these days come from assembly lines and luthiers are rare and pricey, and most who retired in the past 30 years were not replaced by a younger generation. There are 2 sides to that question: the technical one ("I want a drawing of XXX for...
a specific purpose") and the poetic one ("I want to have a piece of your beautiful soul for when I need to connect to it"). Obviously, the first one is indeed going to be replaced by a more efficient machine than the meat-n-bone one. But I'd be surprised if anyone goes into such craft solely for...
the technical aspect, even though it is an important one. Good technique is often thought as a necessity, not an end-goal in itself. The second one is not under threat of anything: there will always be people whose random assembly of thoughts finds beauty in the random assembly of thoughts of...
another person. This is what we value in an artist, the person themselves and how they poured themselves in the outcome, not the product per se. For that matter, however it is done is irrelevant. The masters of the renaissances had many assistants doing the actual work, and Dalí used to sell...
signed blanks to his students. Could a machine -without any human input- give the satisfaction that we used to only get from the soul juice of another human being? Maybe. But I suspect that there will always be a market for soul juice, albeit a smaller one maybe. I suspect some humans will...
only get confort from the idea that it is another human who felt the pain or the joy they are feeling and was able to express it so beautifully. The machine doesn't feel anything, ergo its emotions on canvas are fake, whereas I trust the human has the same emotions as I do. That'll always be...
a win for the flesh machines, no matter how good technique the silicon machines have. Finally, I think the discussion (as always with AI) tells more about us than about the technology itself. We want to have value for what we are and not what we do, and yet we worry that a machine would be...
economically better than us at what we do. Isn't that bizarre? The machine could never replace the thing-in-itself, only the phenomenons with observe from it. Maybe what we want is never achievable in the first place -except maybe for some form of love- and we are just sad that the machine shows...
I’m reminded of the line from Firefly. “May have been the losing side. Not convinced it was the wrong one”
It’s just that the closer you look at how generative AI works, it’s painfully obvious how this stuff has no chance of truly competing with human intelligence. If there’s a breakthrough in AI that makes human intelligence and creativity obsolete, it hasn’t happened yet.
What can be done with LLMs has plateaued with GPT4 and other models of the same cohort. They ain’t got more training data or tricks up their sleeves.
It's funny how people with a long view can sometimes fall to defeatism. This fight is this fight, and two fights are never the same. John Henry was doing the same exact thing that the steam machine was. Artists making art are not doing the same thing that the pixel selection algorithm is doing.
At most, the pixel selection algorithm is a cheap replacement for human labor, which will be used by corporations to generate what they need. That's the John Henry analogue. If the common artist starts using it, they're not an artist anymore. They're a consumer of the "AI" company.
The former artist can't expect to make a living generating images, because we all know generated images are worthless. So they'll either do something else (like work in the factories as the currently dominant U.S. political party wants) or become a prompt engineer for the corporation's images.
If you don't fight it, you lose. That's the one aspect that is true of every fight.
I don’t think I’ll love any AI art like I love human art. I care about the emotional connection, or the sense of…solidarity, maybe? with other humans. Are you saying there just aren’t enough weirdos to make human art sustainable, Or will the machines get better than humans at art?
So at what point in the "race" do you start carrying bottles to piss in so that you aren't slowed down?
I am getting close to feeling that way!
Argh, fine, at the risk of becoming a character in one of your cartoon panels: Then what are you fkn waiting for, Tovarisch?
Generative AI, at least the kind we have now, is not creative. It would take a serious revamp of its most basic workings for it to be (note that this means it's probably possible!) We are more like the people watching the Salamanca than the people watching the transcontinental railroad be built.
Mind, the Salamanca was an impressive feat! It not only demonstrated that steam engines could PULL things, but that they could do so without losing money! But it was limited by requiring toothed rails, vertical cylinders, and using a single tube for the boiler (and no one questioned it!)
I don't think that's a bad attitude to have, but I think sometimes you come across as being too in awe of this stuff. I don't remember what it was, but you did a thread about an AI thing a while back where it almost seemed like you were advertising for it, and it felt a little weird.
we're all just trying to understand what's going on (even the people programming the thing!)
For sure. I've done a lot of reading about it and come down on the opposite side. It doesn't do what's advertised, hasn't shown improvement in 2-3 iterations, and over 90% of projects using it have stalled. I think assuming it'll eventually replace us just helps the companies con more investors.
When thinking about AI I'm often reminded of this from H. G. Wells' The War of the Worlds, similar to your John Henry analogy.
I also think you might be over-estimating it a little, not taking into account that there might be a hard limit to the tech, and use case. Where it's been impressive, it's not necessarily been good, and where it's been good, it doesn't seem sustainable.
That is the feeling I’ve had as an artist too. I don’t have any illusion that I can “stop what’s coming.” I don’t oppose AI art absolutely, I just have no interest in taking actions to proliferate it myself. I can take a stand for my values and my art may last a while longer after I’m gone.
You are vastly overestimating the technology.
When it comes to AI, it's not necessarily a bad thing to err on the side of caution