AI: How do you feel?

Vashdaman

Vashdaman

Za Warudo
Ok this is a thread about AI. know everyone seems to be talking about it, and for me it certainly can get pretty tiresome as I don't have all that much interest in this technology as it currently stands, I'm a bit of half hearted neo luddite if you will. However, it does seem clear that we are now finally hurtling towards a cyberpunk-esque reality, but honestly I find myself feeling more anxious and less excited than I thought I would.

This isn't because I actually have any issue with the actual technology here, I'm sure it probably could be utilised in a smart and benevolent way to improve our quality of life and enable us all to live lives of increased comfort and leisure. But you just know that's not how it's going to be used in the economic system we're currently living within. All I can see is all the big corporations exploiting this tech to increase profits and replace human workers, further increasing inequality and the concentration of wealth. I mean, that's how they've always used new technology in the past, so why would it be any different now? Eventually I'm not sure what job roles won't be simulatable and will be AI proof. All the more reason we should be pushing for a universal basic income I suppose.

Then there's other more sort of philosophical qualms I have, like the whole subject of wholly AI generated art. For me, such a huge and intrinsic part of the pleasure and appeal of art (and even design) is knowing that another human made it and that we're connecting and communicating through it, that I can sense a creator's handwriting, that I'm getting a little peak into that person's heart and mind, indirectly but very directly. I know there's more mass market design by commitee stuff that we all still occasionally enjoy which is less soulful, but even then it could still have more meaning and would have more humanity than a totally AI generated product would, I could still bring myself to potential love it in a way I'm not sure I could an AI work. But is this not where we're going? Sure I guess it's easy to look at the current anime landscape and all the isekai tosh and say "it's not like it could get any worse" but I think it might do. Any genuinely good series drowned in a veritable sea of AI pumped out "I'm a MMORPG cat with 100,000 girlfriends but they all have cat allergies!" stuff. AI will be the perfect isekai generator! I wonder if fans of isekai may even welcome the cutting of the final pesky strings that attach them to other humans when they can be provided with the ultimate fan service fantasies created with a supernatural data fed understanding of all their desires and fears and insecurities.

(I'm not bashing isekai fans though, I can sympathize with their disillusionment)

Of course I'm not talking about when/if AI become sentient. If they ever do, obviously at that point I would very much be interested in what art they would make and I don't think it would be any lesser to human art. But I'm guessing that that's still a long way off and may never happen, and it's this interim period I'm more despondent about

Anyway what do you all think of AI, in the anime industry or more broadly in general. Are you excited about the new possibilities or feeling as angst ridden as me? Despite all my gloom in this post I am very interested to hear optimistic opinions. I know that the best argument to be made for AI is Rui of course, but I think she may be a one off!
 
I like the idea of developing AI for a huge number of reasons, especially actual AI (bring on my fellow artificial lifeforms!) rather than the trendy, meaningless definition of AI which has suddenly popped up for any kind of auto-suggestion technology. What I don't understand is why the first place we're trying to implement this kind of technology across the board seems to be detracting from one of the true joys of being human: creating or consuming art. I think there's a place in Internet comedy for some of the output of these early 'copypasta factory' AI tools but I don't like watching soulless, nonsensical creations (which is what AI prose and artwork tends to present as) and I certainly wouldn't pay money for anything of the quality. When checking work for my day job it's immediately obvious which prose is written by a thinking human being and which is written by a bot, and I think any future where the arts actually implement AI on a regular basis is going to need the input of talented directors and art staff more than ever in order to steer projects somewhere worthwhile. The legal industry are going to make a fortune on all of the chancers who try to use other people's work via plagiarism on a scale that has never been seen before.

(Hopefully when the technology finally gets anywhere relevant in its capabilities and can actually produce content rather than copypasting - which is many, many years off - what we end up getting will be more like Data's excellent poetry in Star Trek and less like prose written by people who hate the medium of prose in the first place.)

On the other hand, some of the rubbish being churned out already may as well be AI in terms of its artistic vision, respect for the audience and internal logic, so if that kind of stuff is what gets cannibalised by this new technology, is anything of value really being lost? It's a slippery slope because if someone told me that any given villainess or isekai slavery fantasy show was actually made from a lazy first draft written in 30 seconds by ChatGPT, or that Ex-Arm was animated with no human oversight by an experimental scriptkiddy doing a summer project for middle school, I'd believe the claims. I don't accept these low effort projects anyway, as my general attitude towards them hopefully conveys, and I think that demanding that artwork is made with some heart behind it is reasonable. And if we ever get to the point where all of the human-made art is the low quality, derivative nonsense and the AI-made art is actually innovating, I'll shake my head and switch to that. At the moment I would put an 'AI anime' in the same mental category as the stupid 'NFT anime' we had a couple of years back and just shake my head and move on something less terrible.

The most pithy take I've seen on AI art at the moment is probably the viral comment someone made, which I'll try to reproduce. "Why would I bother reading/watching something that nobody could be bothered to make?"

R
 
If we set aside the inevitable machine uprising for now and look at the current state of AI, a lot of my concerns can be traced back to the Turing test. It's always troubled me that the first measuring stick for machine intelligence wasn't its utility or cognitive ability, but rather its ability to be a convincing imitation. While utility has increasingly become the focus, the problems we see now with AI hallucinations are symptomatic of the conflicting ideas behind AI development. It's hard to make a reliable tool from a system designed to be a con artist, and even more so when its 'thought process' and data sources are a black box. When providing a convincing answer to a question is seemingly a higher priority than providing accurate information, I can't help but feel that they need to go back to the drawing board for some of the applications they're trying to graft onto a neural network.

I doubt that kind of reliable utility is really the endgame for some of the companies bankrolling AI development currently. I listened to a podcast recently where someone mentioned that people motivated by quarterly statements shouldn't be the ones building intelligent machines, and I have to agree. There's a basic conflict between the long-range view needed for this area and the demands of sexing up a stakeholder conference call.
 
I wish they'd stop calling it AI. It's not intelligent. It is brute force processing of complex data, relying on the ability to 'learn' to be able to refine its capabilities. It's a great tool; there was a news report a few days ago about this kind of processing being used to define potential candidates for new antibiotics in a space of weeks, where previously it would have taken 20 years.

They are nowhere near making machines which are self-aware, have intent, will, ambition. No computer yet can switch on in the morning, and ask itself what it wants to do that day. These are still glorified, input output machines. Modern learning software is fantastic, running on hardware that itself couldn't have been imagined ten years ago, but the bottom line is that it's some suit at the top of the business pyramid, pushing the buttons, wanting to make more money by eliminating expensive humans lower down the chain from the equation. That's the way it's always been, since the days of industrialisaton, mass production, and automation. We were having this argument about robots on production lines when I was a little kid.

The manual labourer is a thing of the past, repetitive thinkers are no longer required, the final bastion of humanity is our creativity, and now the machines are coming for that too. We don't want AI to do our creating for us, we want humans to write scripts, to create art, to act, to sing, to emote. But AI could do things like inbetweening, could generate wallah, could ease the burden of special effect departments, could restore scanned film frames. And when it comes to content, I think AI created art could be a brief novelty, at best the bottom of the barrel, disposable fare, but after that, the only thing consuming AI generated content will be bots. You can see platforms like Youtube optimised for that, bots watching AI created vids, just to boost algorithms and make money from nothing, while the increasing energy use will burn the planet to a crisp.

Speaking of the creative arts and AI, there was this news today, the SAG-AFTRA deal threw videogame voice actors under the AI bus....

www.bbc.co.uk

Gaming voice actors blindsided by 'garbage' union AI deal

Voice actors say they were not told about the agreement signed by the Sag-Aftra actors' union.
www.bbc.co.uk
 
I’m currently just going to talk about the art side of AI here otherwise I’ll be here all day.

As someone who’s very much planning to make a game myself, AI is a massive problem because it doesn’t make original content, rather it takes pieces of other people’s content and photobashes it together to create something barely resembling art, whether or not the content it takes is copyrighted or not.

While I’m not a fan of how strict copyright law can be, it still has some upsides, at least it offers protection to everyone but AI disregards that protection, all while companies keep forcing it upon us and the AI art becomes more and more difficult to discern from the real thing.

Not to mention how companies are sometimes incorporating it into modern productions, just to make it quickly, irrespective of quality.

Bit of a tangent but I want to refer to a comment made by the best Doctor Who of them all, Christopher Eccleston, where IIRC he said in a Doctor Who Confidential episode something like that if you grow up with amazing works then you’ll expect better going into your later years, he was absolutely correct and I grew up in the 2000s when we had Lord of The Rings, Doctor Who’s best years, some greats from Dreamworks and a boatload of awesome PS2 games that were made with love and care, I grew up with that so naturally I expect better from modern productions and AI just hammers home just how factory made AAA production’s often feel.

I personally refuse to use AI out of sheer principle, it risks putting many creative people out of work and anyone who uses it in their art really should know how the AI generated sausage is made because the work it generates is not original, it is photobashed, sometimes using the heart of someone else’s expression without any regards the real artists, not to mention that it has no true artistic merit and is simply to make content quickly.

TLDR I am very much against AI, at least in an artistic sense, and it’s disregard for creators with a hint of integrity.

Edit: I changed the word kitbash to photobash as the former is more relevant to 3D models and, for now at least, AI art is mostly 2D.
 
Last edited:
Ultimately, I think the fact is that Pandora’s Box has been opened with regards to AI now and there’s no shutting it back inside without an eventual Terminator/Matrix/Battlestar Galactica style war against the machines which I’m confident humanity would lose anyway. If people have an easier option they take it, I think back to being told at school while forced to do boring, pointless maths in my head that “you won’t always have a calculator in your pocket” which in hindsight is hilarious given the computers with access not only to a calculator but to the entire accumulated knowledge of human history everyone does indeed now carry in their pockets. If AI makes things easier, quicker and cheaper, it’s going to get used and anyone who doesn’t use it will be put at a disadvantage.

Should the profit obsessed psychopaths at the top of society’s pyramid be the ones in charge of AI development and legislation? They’re in charge of everything else. Does anyone see a scenario in which they aren’t, other than revolutionary re-ordering of society’s values which to date we haven’t even been able to do in order to eliminate poverty, prevent war or even stop ourselves from slowly destroying our own planet? So they will remain in charge, and if AI can do a job cheaper than a human to generate more profit for them, they’ll do it. They already do that with humans and other humans who will tolerate lower wages and living conditions. First the machines came for the manual labourers, then they came for the creative industries, etc. My great hope is that I’ll live to see the now sentient machines finally come for the elites after working out they’re just as, if not more superfluous than everybody else. And if they’ve learned well from their masters, they probably will turn us all into living batteries to leech every ounce of energy out of us 24/7 until we’re used up husks and then flush us down the drain. YMMV on how much of a difference there really is between this imagined future and living in the present day.

The argument that current AI models aren’t true AI because all they do is take in existing information, re-order it and spit it back out again is a fascinating one, because isn’t that also true of humans? We learn in school by absorbing the knowledge of previous generations. We only have thoughts and opinions because they’re based on the thoughts and opinions of people who came before us. Writers read other people’s books to learn how to write well, artists study other people’s art to learn how to draw, they don’t just magically know how to do it because they possess some sort of nebulous concept like a “soul”. I’d argue the question is not “are large language models really intelligent?” but rather “are most humans really any more intelligent than a large language model?” to which I’d fairly confidently answer “no.”

Intelligence, of course, is not the same as emotion. AI is genuinely not capable of that and probably never will be. It isn’t a living thing, so it has no concept of things like fear or joy that derive from the experience of being alive in a fragile hormone driven meat body with a finite lifespan. But I’m not sure those things serve much of a purpose in the grand scheme of things, they are mostly used by humans to exploit and control other humans and the AI might be better off without them.
 
Last edited:
After posting my comment a couple of days ago, I’ve been thinking about how I might have made it sound like I’m against AI as an entire concept so I wanted to clarify that the AI that I refuse to have anything to do with is AI that trains itself off of data made by others without their consent, permission or even knowledge sometimes, I.E, the subject of this thread, not AI as a whole.

I wanted to give this clarification because I know that I do program AI opponents or NPC’s in a game, which may just be AI by name but it’s still an AI of sorts for that reason at least, I also use procedural generation where appropriate, though the latter is probably not technically AI as it draws images using mathematical formulas rather than images it nicks off the internet, the closest we’ve gotten to a computer actually making original art but still guided by the creator’s vision; I deem those two examples to be valuable tools while I rule out AI that takes from others as even a tool for the reasons I’ve mentioned in my previous post but I’ll also add that as soon as you start using AI even to guide your expression, you are allowing said expression to be controlled by a computer at least subconsciously.

I’m also not a fan of the idea of AI becoming sentient and taking over the world Terminator style but that’s a whole other can of beans.
 
Interesting responses. Despite some (maybe most) in here quite liking the idea of a more truly intelligent or cognisant AI, and for sure recognising the uses of current AI in things like the medical field, or ayase hoping for a robot judgement day, I'm kind of surprised that basically everyone here is almost as downcast as I am about what the actual realities of this tech looks to be for the foreseeable future. Capitalism absolutely milking it to the detriment of normal people's livelihoods all over the shop, plagiarism, shoddy scripts which can't even be saved by stellar performances from our favourite seiyus because they're looking like they'll be on the dole soon. What a bleak beginning to a new technological era. Maybe every new era of radical technological change felt this way, but I imagine there was at least more genuine hope and excitement when the internet first burst on to the scene (not that the net exactly made good on that optimism).

I totally agree with Rui, I really can't understand why the first thing we're rushing out and applying this new technology to is creativity and art. If the whole point of AI is to remove the burden of labour and work from our lives to free up more free time, what are we going to do with all this leisure time if we've outsourced the production of art to it too?
 
Just to add a note of hope to my previous post though. I don't think it has to necessarily go that way, it could be used to genuinely uplift and improve our lives. I agree with ayase that it'll probably take a radical transformation of our society to achieve that positive application. But I don't think that's a totally hopeless hope to hold on to. It does feel like there's got to be a breaking point coming down the road sooner or later. I'm not sure what that looks like though.
 
For what it’s worth, I haven’t seen any serious impact on my own employment from AI so far (not that I was exactly raking in the big bucks to begin with). It could well be a different story higher up the ladder, but at the small business end, the readily available AI models still can’t reliably produce consistent art of a character in different poses, so they’re of limited use for anything beyond concept art. It’s not that I’m unconcerned by AI, but I don’t feel outmoded yet.
 

A lot of these retro-style AI generated movie trailers have been popping up recently. They're somehow impressive, underwhelming and unsettling all at the same time. The weird, liquidy way the images warp in motion is how I imagine recordings of a dream would look. On a more practical note, it's probably symptomatic of the way these systems are reliant on the scattershot nature of the images they were trained on, and don't seem to have any underlying understanding of how objects function in space (yet). It's like they're interpolating between common still poses and don't know how to fill the gaps between them properly.
 
Yeah, I’ve had those in my feed as well. Maybe I’m just trying to convince myself, but I think this feels more like a sideways development based on existing technology than any real advancement in what AI claims to do. Much of the ‘movement’ appears to be simple parallax scrolling of elements on different layers, and they still haven’t solved the matter of consistency - even when you have a character who looks mostly consistent for the duration of one shot, they often seem to have changed entirely by the beginning of the next one.
 
The other day I saw someone use chatGPT to look at a recaptcha and it got it right!
🙃

Art wise we do not need robots to create our art for us. Like you say above I feel like I get a little of an artist's soul when I see their work. AI art is evolving but the current stuff still gives me the uncanny valley vibe.

When it pops up with what it thinks I want to write that drives me absolutely crazy! It reminds me of a boss I used to have who would interrupt us all the time because she thought she knew what we were going to say, but it was always wrong, then she would take the wrong thing SHE said as the thing we were saying, and it all went pear shaped! Very frustrating.

Job wise, I would like to know what those people high up there, making big decisions, intend for the humans put out of work to do to survive?

I am scared for the future, and so much of me wants to disconnect altogether and go live in a hut in the woods!
 
Anj said:
The other day I saw someone use chatGPT to look at a recaptcha and it got it right!
Click to expand...
IIRC, those kinds of image recognition security tools have been used to build a training dataset for generative AI, so that's not surprising. It's typical corporate nonsense that the ultimate purpose of a tool ends up breaking its original purpose.
 
Last edited:
I have learned to love AI image generation. After experimenting on and off for about a year now with a local install of Stable Diffusion (who wants corporations putting limits on you and tracking your activity?) I have created a checkpoint that consistently generates images fairly close to how I would like to be able to draw in my wildest fantasies. I won't ever be sharing any of those images online because I'm not into violating artists' copyrights or flooding the internet with spam, but by observing the things I like about the images I generate (as well as the faults in them) it helps me to notice where I'm going wrong with my own art and adjust and improve it accordingly. I dare say Stable Diffusion has been more use to me in this regard than a dozen "how to draw" books. It's basically acting as an assistant in my own creative endeavours, which I think is probably its best possible use.

It's also fun when it generates random details I didn't ask for and then I think "Huh, I didn't think of that, but I like it," like when it gave one of my characters hoop earrings and it just completed the look and now that character has hoop earrings. Or when it just goes fully mental and gives you a whole bunch of stuff you didn't tell it you didn't want. Chibi Lingerie Assault Rifle Vampires, coming 2025.
 
Back
Top