#or that I haven't used generative AI in the game development process (???)

[beekeeperspicnic]

Giant hug for everyone who has been onto the game's Steam discussion board and replied with classy snark to people grumbling on there.

I'm reluctant to do that as a developer but when I see you guys doing it, just know that I'm cheering for you.

[kosher-martian]

Ain't got no use for AI.

Look, I get it. I work in IT (if indirectly), but I think we need to kill this AI thing in its crib before it results in a majority of the US workforce doing back-breaking non-automatable jobs for pittance.

My job is 70% procurement and 30% e-waste disposal.

My job exists because:

Our procurement software has a UI that I would describe charitably as "openly hostile".

Far too many employees have access to the procurement software.

The finance department does not hold regular trainings on how to use the software and there's no process to notify finance of new employees who have been granted procurement access so the new employees can be trained.

Until recently, there was nothing built into the software to stop people from buying things with money they don't have or that is allocated to other projects. (A janitor could buy a high-spec Mac Studio using another department's money or simply not pay for it after it arrives.) Technically "not paying" is still very much an issue I have to solve.

Until recently, there was nothing built into the software to stop people from buying things without approval. After a few high-profile issues, we finally broke down and bought a plugin to correct this issue.

There is nothing built into the software to prevent people from buying devices that do not comply with our standards or are wildly inappropriate for their work tasks. (Secretaries getting high-spec gaming PCs just because they had the money in their budgets.)

Until I pointed it out, there was no policy in place to divert newly-delivered IT equipment to the IT department for set up and endpoint management. Until I pushed for a policy change, IT equipment could (and often would) be purchased and delivered directly to the department unmanaged or in a few cases a less-scrupulous employee's home address for personal use.

Our inventory system is arcane and there are a sum total of three people with access to it. And that's the system we primarily use for IT lifecycle management!

All of these things could be fixed by just buying better procurement software, let alone procurement software enhanced by AI. The only reason my employers haven't liquidated me is because right now the annual cost of having a human clean up all the human errors caused by the cheap lousy software they bought is still lower than the annual cost of software that would prevent the human errors from happening in the first place. But for how long?

What about all the jobs AI is already taking? All the writing and editing and media production jobs that simply won't exist because it was easier to let a machine do it? (Ignore that all the CGI people have four thumbs and 12 toes, human. Watch our slop content and enjoy it! You aren't worth real art.)

All the marketing jobs that will be liquidated because the algorithm knows us better than we know ourselves. Where are they supposed to go?

What about all the programmers and developers being laid off because ChatGPT does a decent enough job and if you lower our expectations enough, people will just accept a lousy app to purchase their burritos and yell at the human driver when the burrito is wrong?

Where are these people (and many, many more) supposed to work?

We closed the factories. We didn't find new high-quality jobs for the factory workers, did we?

All the factory workers went to work in low-paying retail and service economy jobs. Oh wait, they've been laid off from those jobs too because we had to have self-checkout and self-service kiosks and now entire Dollar Generals are staffed by one employee.

Drive a taxi? Sorry that was replaced by sharing your car with strangers (Uber). Oh no wait sorry, the self-driving cars are taking that too.

Call centers? We shipped those overseas and then shuttered them because a phone tree with a realistic human voice and intentionally vexatious wait times will take care of the pesky callers.

Gig work? If AI hasn't taken them already, they will soon.

Agricultural work? Slaves Prison labor will replace that soon enough (and frankly prison laborers will likely also take many other menial tasks like package delivery or fulfillment warehouse jobs soon)

Again, what are the vast majority of people on this earth supposed to do? Because if you think we're getting UBI in the US, you are deluded.

When the car replaced the horse as the primary means of transportation and other machines took the place of horse labor, the population of equines declined. They weren't needed anymore.

Between 1915 and 2006/7 the horse and mule population declined 63.07% in the United States. The US equine population in 1915 was 26,493,000 (horses and mules). In 2006, the United States had around 9,500,000 horses (1), and the United States Census of Agriculture for 2007 (table 31) counted 283,806 mules and burros (2).

And while the horse population has rebounded in the years since, it's because of enthusiasts wanting to breed more horses for entertainment and luxury purposes.

No one found the horses new jobs when the horse jobs disappeared. We simply stopped breeding horses. (And that's assuming, charitably, that we didn't juice the decline by culling the horse population.)

When our jobs disappear, do you think our overlords will just keep us around? I mean sure we're taking care of the population problem for them (a little too well actually, thus the anti-abortion laws), so they likely won't turn us into glue.

They'll still need people to fix the machines when they break and to continue building and creating better and better AI and machines to take care of their every whim.

And they'll need some people to do all the work the AI and robots cannot do yet.

And they'll want people for entertainment and luxury: Actors, singers, athletes, and playthings they can use as they please.

But ultimately what all these "tech-bros", "technocrats", "effective altruists", and "techno-progressivists" really really want is world that caters exclusively to their needs and all the inconvenient people who have needs of their own are swept into the recycle bin and discarded without a second thought. All problems are easily solved when you eliminate all the people:

Littering? No people, no litter. You, glorious techno-overlord, would never litter.

Climate Change? Less people, less carbon. You'll still be here, though, don't worry my liege!

Economy? The robots do all the work you used to have to pay people to do. And all the profits go to you, sire!

Taxes? LOL you don't have to pay taxes. Taxes are for the peasants.

You get my point.

The techno-overlords are tired of negotiating with us because they don't believe they should have to. We aren't their equals, if they even consider us human. They need some of us to do their bidding and some of us to keep around to torture and dominate to make them feel like gods. Beyond that, the rest of us are just excess people breathing up all the air.

AI has to go, not because the AI itself is dangerous, but because the people whom the AI ultimately serves have told us time and time again who they really are. It's time for us to believe them. We know their vision for the world, and it's a world without us. 1. This figure comes from a report by the Food and Agriculture Organization of the UN and is often cited in writings about horse population decline, but I could not locate the actual report on their website.

2. https://agcensus.library.cornell.edu/census_parts/2007-united-states/

[inoppositionflorien]

There is no useful way to categorize generative AI as stealing that doesn't also categorize essentially every creative work ever made as stealing.

Further, there is no way to develop generative AI which does practical things, like translation and protein folding, without developing generative AI which does literally anything else, because protein folding, translation, and making a creative work are functionally the same process from a programming perspective, and programmers tend to like open source things.

These things are pretty well established, but also seem to routinely be ignored. So why? Well, I have some speculation about why people might be wrong on the internet. A number of these things may be true in any individual of course.

They just don't know these things to be true, despite them being true, because they haven't heard them yet.

They have heard them but don't believe them, because it would be convenient if these things were not true (because they also have one of the later reasons too)

It's part of an ideological package deal they picked up somewhere, picking up some components of the ideology means they must pick up the others, and part of the highly online progressive ideological package is "anything made by a programmer that attracts media attention must necessarily be bad"

Everyone wants to be an artist, but they see potential competition. Thus, the competition must be defamed. It is also socially acceptable to lie, exaggerate, or generally wildly speculate about machines, so absurd easily disprovable claims get repeated.

People remember how much fun it was to bully cryptocurrency people, because they were a relatively small community that clearly signposted themselves, skewed ideologically right-libertarian (an extremely rare ideology in the real world and thus a desirable target), and defined themselves around a product that made very little financial sense and was impractical for basically everything (and was made by and for goldbugs but that's beside the point). They then assume, partially because of option 3, that anything that programmers make must necessarily make little financial sense and be impractical for basically everything, even when this is not true, and then assume therefore bullying Generative AI people might also be fun, because "they're basically the same thing"

Having enemies is fun in general, and a decently insulated mid-sized community has decided AI people are an ideologically acceptable enemy to have, so if you're going to have an enemy, why not have them as the enemy?

Valorization of certain forms of labor over others. (why is the corporate artist more important to defend than the auto worker who got automated out of a job earlier?)

Desperation for there to be something humans are better at than a sufficiently well-programmed computer, they already dominate board games and repetitive industrial tasks, and now they're coming for slightly less repetitive tasks, and if we don't have slightly less repetitive tasks, what do we have?

Fear of SEO spam and people using AI to lie on the internet (this is a particularly weird one because we already have SEO spam and it's comically easy to trick people on the internet, and it's doubtful generative AI would make either of these problems worse. I've seen people reblog the most obviously fake things like they're real! Using AI to lie on the internet would be like trying to kill an ant with a nuclear bomb, it might work, but also you could just lie about someone pretty directly or create a fake post that you claim is from an opponent and then mock them for it, and basically no one is going to check it. This particular moral panic is actually recycled from when photoshop was new)

Maybe they just genuinely do want more restrictive copyright law that allows companies to maintain very close control over their IP? I have seen this take in the wild but not so much in my tumblr circles, generally it's something about "fanworks butcher the source material and parodies are often lazy or shallow, therefore IP should be more restrictive to prevent those things from happening."

They don't like generative AI for some other reason (they're an educator and got handed a generated essay and had to grade it, they blame it for contributing to the continued decline in quality of search engines, they dislike how companies keep trying to make A Virtual Assistant Like From The Movies a thing and don't fully remember that this "we're going to make virtual assistants a thing" has been going on since at least Clippy and probably way longer) and then generalize their dislike to every aspect, because one aspect annoys them but they don't feel that's justification enough to be angry about it.

Anyway that's my speculation. Maybe there's other reasons why people keep making incorrect statements too, but I can't think of any that wouldn't be covered by these points right now.

[hummingbird-games]

Dev Diaries

March 5th, 2024

So. We're 3 months into the new year. Huh.

Updates from me? I started a new job! I've been reading/catching up on a toooooon of manga lately (I'm also sloooooowly making my way through a nonfiction read and can't help thinking how Corey would heavily sympathize with me. Sigh.)

I also got another free month of Spotify (lol, I think I'm the only person I know who only uses the free plan because I'd rather spend that money elsewhere??) so I've been on a listening spree and flagging songs for a writing playlist I will use for the majority of the HSDJY 2 drafting process moving forward.

I still have yet to play BG3. My family gifted me the physical edition, so I've been impatiently waiting for it to ship out. It's been 84 years JFJEHFJHJF!!

Hmmm...nothing else interesting has happened to me (that I can publicly share ☠️) so on with the game development updates!!!!!

What Has Gemini Been Up To? -> TKD (again)

March 16th!!! Y'all, this is the deadline the team and I are working towards for a finished and published game 😭 unfortunately my plans for full voice work won’t be realized by then (but they will be realized. Just in a few months. Honestly, it wouldn’t surprise me if y'all get the updated HSD 2 demo before the fully voiced TKD….)

What Has Gemini Been Up To? -> HSD:JY and Ko-fi

Ko-fi first!!

It’s only been a few months (since November??) but I’m super proud of my posting consistency! Granted, I’m a little anxious about how posts will look moving forward once I start moving major updates and general silliness to Tumblr. But. We shall cross the bridge when it appears.

March snuck up on me and I only have the free and any-paid-support ready, but the subbies are just There™️. That being said...when I'm quiet or posting non-Hummingbird content here, there's 99.9% chance I've made a free post on ko-fi.

Tumblr is looking like the other social media platforms right now with its flagrant support of AI, and it gives me a headache trying to decide how to keep everyone informed as well as share posts from my peers and new kids on the block. Anyhoo!!!

High School Daze goodies??

At the time of this post, a couple things have been happening that I haven't had the time (or the energy) to talk about. The obvious? HIGH SCHOOl DAZE: JUNIOR 2 HAS AN ITCH.IO PAGE!! I...have no idea when the full game will be out. I have an idea...but I don't want to say it and feel obligated to make that deadline just so I don't disappoint players. (I'm well aware that some people took one look at Crushed, went "Aww, that's nice, Gemini. Now where's HSD??" 🤣)

The first round of sprites have been commissioned!! If you peeped the key art (also done by my sprite artist, heehee she's lovely) you might notice some new outfits, some new hairdos. Fingers crossed I'll be able to update the page--to mirror the debut game's page--at the same time the new demo goes live (which will showcase all of or at least 75% of the common route of the full game. Stay tuned to find out if we'll focus on the friendship route or if I can defeat the Coding Monster to include all the variations for the romances too aha).

Writing wise? Five of the six total routes are outlined. (Florence's I've…barely started 🥲…this pre-production thing is kicking my ass). I’ve started drafting out the friendship route too, as seen by the random posts I’ve made about HSDJY 2. Well, a combination of friendship route + the common route with its lovely variations that aren’t a nuisance at all. Nope. *eye twitch*

I alsooooooo discovered that all the raw and edited music I created for HSD and for personal projects over the years using GarageBand were deleted. And I never backed them up. So. The tracks I made, the jingle for the splash screen, the main theme and it's 2 variations, and alllllll the little cute things I made that are as old as my own high school days are....gone.

Poof!

If I sound very calm about this, what an astute observation! But my anxiety is being used elsewhere, and I decided that I while I can't go back to the original files to tweak them, I've backed up the other files. And I don't mind starting from scratch with this.

But also y'all, please back up your shit. Please. Don't be caught slipping.

That poll I made a while back that now has results and I totally didn't forget about?? The boys won!! And I'm not surprised 🤣 I haven't decided yet if I'll do a live developer stream or a prerecorded one, but either way it'll happen closer to October, the 2nd anniversary date.

What Will Gemini be Up To?? -> Rest?!?!?

I assumed a lot of things about what would be done or not done by March, and that led me to loosely block out March as a 'rest' month. I wouldn't work on any projects, and instead would read, play console and computer games, and basically take the break I didn't take in December.

Well. The Knight Dance is still in production. And I commissioned sprites earlier than expected. And I didn't plan enough ko-fi content to be scheduled in my absence. LOL and I started a new job!? 🥹

But!!! But but but, I do need to take some type of break so I'll do my best to be scarce in this space (and lower the temptation to work because I see y'all are working LOL). Cool? Cool!

- Gemini 💛

[sappho-official]

I need to stress to you guys what actually happened here.

In 2018, the Elder Scrolls VI was announced to be "In Development"

But it was not in production.

As far as I know, it was in early preproduction; the gap between Fallout 4 and Starfield was filled not just with Starfield's development, but the development of the Creation Engine 2. Fallout 4 already looked dated compared to other Triple A titles launching the same year. And if there's one good thing that came out of the disaster that was Fallout 76's launch, it was that Bethesda became incredibly aware of the fact that the original Creation Engine (which used a lot of framework from the Gamebryo Engine, which they'd been using since Morrowind) was not going to be able to keep up with modern development trends.

So the engineering team started making Creation Engine 2. And it's good! To be honest, Starfield really does feel like a tech demo, which makes sense for a first game in a newly-developed engine. But only the engineering team needs to be making the engine; the other teams (design, art, narrative, animation, etc) would definitely be making things in the new engine and giving input, but Starfield wound up with a much, much longer preproduction process than your average game as a result. There's not a lot of point in starting preproduction for another game when you haven't even finished preproduction for your current project! Making your devs constantly switch between projects is not good for their productivity. Once Starfield entered full production and was like, definitely going to be a real game, a few designers would get pulled off to work fully on TESVI, while meeting with leads about what TESVI would look like. They'd write design documents, maybe do a little concept art, and pitch that to execs and Todd himself. And, considering the timeline, it's probably been able to go through a lot of revisions.

So then.....why announce it in 2018?

You see, the announcement that "The Elder Scrolls VI is now in development" was not a developer decision. It was a marketing decision. In fact, the majority of developers at Bethesda Game Studios were not aware about the TESVI announcement before it was made. And they were pretty mad! But what can you even do in that situation? Saying "no, most of us aren't working on that right now! It's just in design documents, there's no game at all!" breaks NDA, especially if you post about it on social media.

The phrasing of "in development" is what gets me, tbh. The engine itself and a lot of the tech (like AI and ye olde Radiant Quests) was being developed, so does that technically mean that TESVI was in development? Yeah, kinda! If you squint! But was it anywhere near a playable state or really, like, existent? No, not at all.

So why did marketing announce it in 2018 instead of waiting? It probably had something to do with the sheer amount of time they knew it would be between games; they probably had some weird timeline of when they needed to increase hype around the Elder Scrolls in general (whether that was a Skyrim rerelease or an ESO thing or whatever). And they almost definitely had a very different projected timeline for Starfield and expected it to come out in like, 2021 or something. Then covid happened and Bethesda got bought by Microsoft who probably had a lot of input in what they wanted Starfield to be because they needed it to be a financial success in order to justify their purchase.

It wasn't, of course, but that's a story for another day.

[scarecrow-the-ghost]

hi!

{

name: “Scarecrow the Ghost”

pronouns: [“they/them”, ”he/him”]

flags: ["nonbinary", "asexual", "robotkin", "questioning"]

also_known_as: [“Toby M.”, "TM-6278"]

does: [“art”, “activism”]

description: “>hi! my name's Scarecrow, and I'm an artist, activist (in training), and game developer (also in training) with autism. this is just my art blog, but there’s also gonna be some reblogs and thoughts here and there.

>I mainly work with 2D, but I've been getting into 3D art/animation and multimedia as well, with an emphasis on what viewers process when viewing the artwork (alongside the atmosphere and sense of humor). I'm also interested in virtual reality and accessibility.”

important_notices: [

“>I'd love to do commissions, but I unfortunately haven't set that up yet.”,

“>please don't repost/edit my art or use it to train AI models. knowing some basic concepts of how the technology generates images, I seriously doubt it'll do great at replicating my art style, so just don’t.”,

“>not necessarily a DNI, but I'd like to keep this blog SFW or at least worthy of a marking of 13+... so no NSFW requests or submissions please!”,

“>you’ll typically find my art under the tag #scarecrow’s art, or under ‘my artwork’ on this blog’s corresponding website.”,

">just because I'm in the robotkin community doesn't mean I don't still care about other humans/robots/animals/concepts/etc, or that I no longer have agency or free will. please don't treat me as inferior because of that. (also I'm very aware of the Soul Argument, and I'm not lettin' that stop me from bein' me.😉)"

]

contacts: {

“Discord”: "https://discordapp.com/users/1288827854385315843",

“email_me_here”: "tsmckinney10@gmail.com",

"Mastodon": "https://mastodon.social/@scarecrow_the_ghost"

"DeviantArt": "https://www.deviantart.com/ghostbot19"

}

}

[bookwyrminspiration]

thank you for the well wishes! we shall see how it goes haha. my brain is going offline as i shut off all my devices since there won't be any service (i am going camping) but just wanted to say: VERY cool to hear about the ways that video games use AI!! i understood what kind you meant & such; if i was not raised hearing about computer stuff and hadn't had several long late-night conversations with my dad about it and also was not studying programming myself i might have been more confused but no, i definitely understand and was just very intrigued! like AI is such a neat tool on the programming side like on the newer developments end it can write example code for a specific thing that you can then use as reference (w/ the insight of an actual programmer who knows how things work and thus how to correct any possible errors) but i think it's also been used in search engines and algorithms for a while, which is kind of a given? it is just a cool part of the tech field i haven't really explored much that is very interesting to me because of just all the possibilities of it. but yeah, i don't hear about it from internet people other than about the more recent big thing of AI generated art and writing, so i just had to pause and go :0! for a minute haha!

Ah, well forgive my assumption to the contrary then! Something about the way you worded it gave me the impression you'd interpreted it differently than I'd meant. But that's my error; you probably know more about the subject than I do. I also haven't explored it much, just know some things in passing since my dad quite likes video games, and I play casually as well.

The programming aspect of AI? very cool! the "AI"s being fed people's creations without their permission to create algorithmic uninspired replicas? not so cool. I've met a few people who like AIs for one thing or another and the few times it's come up in conversation I've tried not to be overbearing about it (because they do acknowledge stealing work is bad) but I think if it comes up again I will simply not hold back. I'm always trying to be nice I gotta be more assertive sometimes

but anyway! the whole process behind video games is like magic to me. I know in theory how it works, but then I see it actually work and I'm just like. what the fuck how did you guys do this. have you ever seen ori??? do you SEE the fucking visuals on that thing?? BLOWS my mind. very cool :)

[christophercant]

How to compete against AI: Future-Proofing your Art Career Part 2

This article was originally published on ChristopherCant.com

This is part 2 of my series about AI art, and future-proofing your art career.

Part 1 was an overall look at what the potential threat of AI art might be to the working artist, and an overview of the different solutions an artist might use in order to continue a sustainable income in an uncertain future.  If you haven't read it, it's over here.

After writing that article, I wanted to look a bit deeper into alternative solutions, and figure out a list of ways to make myself a more competitive artist in a world that can generate cheap, high-quality imagery in seconds.  This is as much a letter to myself as it is to anybody else.

We've all heard the comments that AI is a tool, and that artists should just adapt and use our artistic skills to paint over AI outputs - but I don’t think many of us would be happy with that.

We became artists because we have a personal vision, and enjoy creating - I expect few of us would find joy in handing over our creativity and vision to something else, and relegate ourselves to cleaning up what it makes.  In that situation, it seems more to me like we are the tool, and the AI is the artist.

So let's dive into the many ways to remain competitive, without resorting to giving up the most enjoyable part of an art career to a robot!

Compete against AI by being yourself

AI is only your competition if you try to match it and beat it at its own game.  Instead, concentrate on doing your own thing, and become something that AI will never be able to imitate.

Become yourself

Stop chasing the big freelance gigs in the hopes of getting hired.  Stop trying to make your art look like the Blizzard art style, a Magic the Gathering card, or League of Legends splash art.  These styles are sought solely by companies looking for art to help sell their products, and they are the things that the AI art is learning to make.  In all honesty, I think these companies will bite at the chance to replace 90% of their artists with AI and reduce their expenses drastically; I just think they are going to try to transition to that as quietly as they can, so they don’t get a mob after them.

Now is the time to chase YOU, to dive into your own vision. When AI is churning out masses of the same stuff, you should be creating the things you love.  Of course, still learn your fundamentals - good art is still good art, and a good idea can still be undermined by bad execution.

Developing a distinct, recognizable style is key to standing out as an artist, and will become even more important.

My work - more over on my Artstation

Developing a style isn’t so much about finding some obscure trick that clicks with your tastes, or about learning from the right sources, or purposefully doing something a little bit unorthodox in your work to make it stand out.  It’s more about removing all of the expectations about what your work ‘should’ be, and unburying what it really, truly is that you like to make.  Underneath all of the expectations that your work should have dynamic perspectives, that it shouldn’t be just black and white, or should have painterly yet controlled brushstrokes, or that you should be a master at male, female, and all other forms of anatomy.

Beneath all of that is what you actually like to make.  Don’t try to do things with your art to make yourself purposefully stand out - just figure out if there are things like colour palettes, compositions, or subject matters that appeal to you more than others.  Embrace the things that appeal to you most and reject that which doesn’t appeal to you; you’ll make the art that really matters to you, and standing out will just be a byproduct of this process.

Experiment and explore

AI is able to generate a large number of variations of a piece, but it can be challenging for AI to come up with truly original, creative ideas. An artist can set themselves apart by focusing on creating new, unique pieces of art that push the boundaries and introduce new ideas.

Try out different mediums and techniques, explore completely unrelated disciplines and incorporate what you learn into your art, integrate your own personal history into your work, try out telling personal stories with your pieces.  These are things that AI will struggle to imitate. 

Use randomness, chance and improvisation to your advantage - so called ‘happy accidents’. These don’t exist in the same fashion in AI art; though you can certainly find mistakes and random noise, there is a distinct and recognisable pattern to it.  It’s not truly chance, like the happy accidents you or I might find in our works.

Taking that further, art that is heavily abstract or non-objective in nature can be challenging for AI to recreate because it lacks the ability to understand the emotional or conceptual elements of the work.  If you enjoy abstraction to any degree, try out leaning into it and see how it affects your art and its reception.

A landscape by Victoria Rose Park, a successful abstract artist

In a similar vein, humor, satire, and irony is also a struggle for AI; it’ll never be on the cutting edge of comedy, because it lacks the ability to understand the nuances of human humor and irony.  If it suits you to include comedic aspects into your work, that's one way AI will really struggle to compete.

Build a strong personal brand

Building your personal brand is important for any artist in this day and age, as it helps to establish an artist's reputation and reach potential buyers and collectors.  In the AI age, it’ll be more important than ever.

Use social media platforms, a personal website, and other online platforms to showcase your work, connect with your audience and build a community around your art.

Include a bio on your website that lets people get to know you, and highlights your personal background, influences, and the themes and concepts in your work.

Create a strong and publicly broadcasted narrative around your art, such as a backstory, or the process behind the work.

Create blog posts and youtube videos that share your thoughts about your work - about anything at all.  Just by saying what you think about things, you’ll be building rapport with people.  Hell, you’re getting to know me better just by reading this very blog post!

Importantly, be as authentic and true to yourself as you can be.  People can tell when someone is being honest and sharing with them, and most will appreciate it, and want to support your openness.

Compete against AI by showing your humanity

I mentioned this in part 1 of this series, that one of the key advantages of human-made art is the unique, personal touch that it has, by virtue of being made by an unique individual with their own experiences and perspective.  AI will struggle to imitate this authentically, and people will often be able to feel the difference.

So let’s play into that.

Incorporate evidence that your art was made by hand.  

This can be done in many ways, such as fingerprints, brushstrokes, or other marks that indicate ‘humanity’.  This is obviously already present in most traditional mediums like oil painting and watercolor, but can also be purposefully included in digital art as well.  If your digital painting style is very smooth and clean, you might want to roughen it up slightly, to let more evidence of your hand show through in the finished result.

To go even further, things like mixed media, collage and paper cutouts are currently impossible to replicate with artificial intelligence.

Personalise all of your products.

If you sell merchandise and products, find a way to make them more personable.  Sign everything, hand embellish your prints with some gold leaf or a personal stamp, include a thank you note with each purchase, that sort of thing.

AI Art is focused on creating quantity, and lacks soul; focus instead on creating a high quality human-experience. Create the very best work you can, present it the most human way you can, and sell it in the most personal way you can.

Record and share your creative process.

When you sit down to create, record it on video.  It doesn't have to be particularly high-quality or professional, just good enough to demonstrate that you handcraft the things you make.  If you are a traditional artist, set up your smartphone or a camera to record; if you are a digital artist, use a screen recorder while you paint.  Clip Studio Paint, Procreate and Krita all have recording built-in.

Turn those recordings into videos, put them on social media, display them in your gallery alongside your artwork, think of the finished piece and the process footage as one and the same product.  If you can, narrate the footage and talk about why you made the decisions you did, where your inspirations come from, if you are happy or disappointed with the end result - anything you can talk about at all will help the audience connect to your work, when most of the art they see has no voice behind it at all.

Not only is this one way to help audience connect with your artwork, it’s also proof of your process, which might matter to some art consumers.  This might become especially important to digital artists, as there are very few ways to differentiate a finished digital painting from one generated by AI, except for a proof of process.

Compete against AI through diversifying

AI art is fairly one-dimensional, and still requires people to actually facilitate it and turn it into a product or make it part of something bigger.  It certainly can’t compete with a multi-talented person.

Diversify within your Art 

AI can’t yet integrate different art disciplines together, such as music, performing arts, theater etc into a single experience.  This requires a human, so it’s the perfect way to set your work apart. 

Incorporate multiple mediums such as painting, sculpture, printmaking, photography, and digital media into your work, creating a unique and layered approach to your art that is difficult for AI to replicate.

Take it further and create multisensory, interactive and performance-based art that incorporates movement, sound, and audience participation. 

This would create a unique and unreproducible experience that AI wouldn't be able to match.

Diversify outside of your art

Combine your art skills with other complementary skills, you may be able to create projects that make you stand out from the pack, without even having to utilise AI at all, and without having to compete with those that do.

Digital tools are becoming much easier to use year-on-year, and allow a single person to achieve much more by themselves than ever before. 

Web design, programming, social media, video editing, music production, writing; the list goes on.

Find a way to combine your art with other skills to produce something new, something that even other people aren’t doing, let alone AI.  

I’m diversifying by learning to write, learning web design and SEO and then combining those skills into educational and entertaining websites that as far as I can tell, no one else has built yet.  See if you can combine skills in interesting ways and create something new.

My current website-in-progress, FallofCamelot.com, a site about the Arthurian legends

Diversify your income sources

This isn’t exactly competing with AI art so much as it is sidestepping it, but still deserves to be mentioned.

There are many ways to earn an income as an artist, outside of freelancing and commissions, and many will be completely unaffected by AI.

Teaching is one such method.  Even if AI art somehow took over 100% of art sales (don’t worry, it won’t) people would still want to learn how to paint for fun, and these people will want someone to teach them.

In fact, as more people get their hands on things like iPads and more people from third-world countries connect to the global art community through the internet, the number of people looking to learn art seems to go up and up each year.

Teaching is something that many working artists should be doing anyway; we have some responsibility to pass on what we know to the next generation of upcoming artists, so that they can create even more amazing things than our generation did.  Where would we be without the previous generation of artists inspiring us to live as creatives?

You should also consider selling your own merchandise.  As automation comes along and threatens our very jobs, it also presents opportunities, as it makes it easier for each of us to create and sell prints, books, games and other merchandise.  These are things I want to go further into, but I’ll do it in the next part of my AI series.

Compete against AI through engaging directly with people

People don’t want to buy art, they want to buy a small piece of the artist, so to speak. Most people definitely don’t want to buy soulless art that has no artist behind it at all.  

So let people get to know you.  Selling your art is as much about selling yourself as it is your creations.  

Your motivations, what it is you are trying to accomplish and why it’s important, how your work represents the things you believe in; communicate these things to people and it will endear them to you.  When a person hears something they resonate with, they want to support it, and to help it grow.

Engage with your audience through livestreaming and video

Whatever social media platform you prefer using to share your work and talk to the people that like it, you should be able to find a way to livestream.  This is the best way to build a relationship with people online; to let them see you, to hear you, live.  They can ask you questions, and see you respond in real-time.  It’s human, and it will help you build a relationship with people.

Livestreaming on Twitch was the activity that brought me my core audience. My art was probably what got their attention initially, but it was the stream that kept them around.  I try to let my personality out as much as I can on stream, and it’s rewarded me with a small band of loyal supporters and close friends.

My Twitch channel

Whether it’s Instagram, YouTube, Twitch, or any other social media platform you use to communicate with your followers, you’ll be able to find a way to livestream.  

Recorded video can also be used to achieve the same thing, as long as it is not too scripted and impersonal; you want to let yourself shine through, and show people who you really are.

Sell your art face-to-face

Conventions and galleries attract people that like art, but who want much more than that.  They could just buy art online, but instead they want to go to it in-person, to be there themselves. They want to see the art with their own eyes, to meet the artist, to talk with them about the art.  AI art simply doesn’t satisfy any of these criteria at all - but you can.

So make prints and other merchandise of your art, and be prepared to dive into the world of face-to-face sales.  There are artists today who make six-figures annually from these sorts of events, and this won't be impacted by AI art much, if at all.

Create art for specific people and locations

Your local businesses, bars and cafes, social venues and public spaces; AI art can’t connect authentically with the spirit of any of these places.  Even if AI could somehow paint art straight onto walls, it still lacks the ability to understand the context of the location.  It isn’t able to understand why certain ideas, phrases, imagery or memes are important to a certain place, and the people who go there.

Murals and graffiti immediately fall into this category, but it also goes much further: each local business has its own culture, each district, each town, city and country.  My birthplace of the Falkland Islands has immediate connection with penguins, sheep, the words ‘kelper’ and ‘benny’, Margaret Thatcher, Argentina, looking beautiful in photos yet windy beyond belief in person, etc; a massive list that only people who are really familiar with the place will fully appreciate, and only artists who understand the Falklands will fully be able to cater to them as an audience.

The Falkland Islands - if I ever need it, I have a market here

Wherever you are, there is bound to be some distinct culture nearby that you’ll be able to appeal to as an artist, and people that will support you for it.

And that’s everything I can offer you today!

If you’re here at the end of the article, you probably realise that there are many things you’re going to have to try in order to figure out what will suit you, what will actually work for you, and ultimately what you are willing to do to create a stable art career in an increasingly uncertain era.

I wish you luck figuring out your journey.  If you want to talk about it, drop me a comment. I’ll see you next time (which, by the way, will be about why AI doesn’t have to be the enemy.)

This article was originally published on ChristopherCant.com

[canmom]

in fact lemme fish this discussion out of the comments bc it's interesting!

@baeddel wrote:

i feel like writing is a place that big studios cut corners but isn't a place you actually need to cut corners at all. we have visual novels three times as long as the Bible, we have light novelists churning out millions of words per year, etc. just put a novelist on a game and you can give it ten, or a hundred times as much writing as most triple a games at very little extra cost or time. there was a game that did that, it was called Disco Elysium… so i feel like the lack of intensive amounts of text hasn't really been a technical problem for triple a, just an aesthetic/business/whatever one. therefore we haven't even advanced to the stage with writing where we have a technical problem that AI would be needed to solve!

I wrote:

mmm, I don't think the LLM would be useful from a perspective of just, generating loads of text to fill out a game with, that would entirely suck shit because it would all be bland nothing. ideally, and I am thinking very speculatively here, you would want to have a development process where you try and get the best out of the combination of an LLM and a human. so the writer might be providing interesting, nuanced, characterful text that is specific to that project, and the game would treat that writing-data as something like assets the way a 3D artist would author texture files and models and so on for the game's algorithms to process further. so you'd use the LLM to modify/interpolate it to permutations of game state, questions they never thought the player could ask, etc. etc.; the writer would be involved in an iterative process where they try out how their writing makes the LLM respond, and modify it to get the desired artistic effect, but know it could be extended into unexpected domains. that writing could be used in the form of prompts, finetuning weights, generating control vectors etc.

Like, it's widely observed that LLMs are broadly derivative, right? If an LLM pulls out an interesting expression when instructed to write creatively, especially in a niche genre where the data in the training set is sparse, there's a high probability it's essentially plagiarised from somewhere you don't know about. If it doesn't do this, since it's built on a statistical model of all human writing available, it will instead probably produce something very generic. Although you can fiddle the temperature and such, I think using an LLM like this instead of a writer is a deeply wrong-headed way to use the technology (which doesn't mean people won't try, because capital gonna capital).

But if you set it up to 'plagiarise' your own writing, you're using it in a different way. We don't create a separate 3D model for every combination of bone angles and positions and camera views; we create some data (vertices, skinning weights, bone keyframes) which the skinned mesh renderer can combine to generate an animated character. The resulting animated character is no less human-authored for being designed to pass through an algorithm before it's seen. (Of course, many AAA games use huge libraries of mocap data and techniques like 'motion matching' instead of hand authored animation, but the principle is similar.)

I think this is probably the way LLMs could actually be useful for games. Going with the 'simulator' metaphor I've been reading about recently, where the LLM weights are the 'laws of motion' for extending out (roleplaying) from the given input, you could use your writing as a tool to probe the space of LLM embeddings and movements through its state space, and in this way, prime it to generate text that has a similar relationship to your own writing as an animated, skinned mesh has to the T-posing model in Blender.

Simply writing prompts to 'off the shelf' language models that have been set up to work as chatbots is unlikely to be effective I think; you would probably need a pretty custom model that has been finetuned on the sample of writing you provide it, and to further condition the model using techniques like 'control vectors'. The text seen at runtime by any given player would be specific to their session, and be able to respond much more organically than NPCs mostly reciting canned responses, but the overall tone and style of the text would derive primarily from your writing.

Would this actually work in practice? It would probably take quite a lot of work and experimentation by someone with both a good amount of experience with writing, and a pretty technical understanding of how to control LLM output at runtime, to get a good result. Probably the vast majority of writers would strongly prefer their writing be used directly, rather than potentially mangled by a highly unpredictable robot. I think it's an interesting avenue to think about, though. Certainly seems like a more interesting use for the technology than just churning it for slop.

using LLMs to control a game character's dialogue seems an obvious use for the technology. and indeed people have tried, for example nVidia made a demo where the player interacts with AI-voiced NPCs:

this looks bad, right? like idk about you but I am not raring to play a game with LLM bots instead of human-scripted characters. they don't seem to have anything interesting to say that a normal NPC wouldn't, and the acting is super wooden.

so, the attempts to do this so far that I've seen have some pretty obvious faults:

relying on external API calls to process the data (expensive!)

presumably relying on generic 'you are xyz' prompt engineering to try to get a model to respond 'in character', resulting in bland, flavourless output

limited connection between game state and model state (you would need to translate the relevant game state into a text prompt)

responding to freeform input, models may not be very good at staying 'in character', with the default 'chatbot' persona emerging unexpectedly. or they might just make uncreative choices in general.

AI voice generation, while it's moved very fast in the last couple years, is still very poor at 'acting', producing very flat, emotionless performances, or uncanny mismatches of tone, inflection, etc.

although the model may generate contextually appropriate dialogue, it is difficult to link that back to the behaviour of characters in game

so how could we do better?

the first one could be solved by running LLMs locally on the user's hardware. that has some obvious drawbacks: running on the user's GPU means the LLM is competing with the game's graphics, meaning both must be more limited. ideally you would spread the LLM processing over multiple frames, but you still are limited by available VRAM, which is contested by the game's texture data and so on, and LLMs are very thirsty for VRAM. still, imo this is way more promising than having to talk to the internet and pay for compute time to get your NPC's dialogue lmao

second one might be improved by using a tool like control vectors to more granularly and consistently shape the tone of the output. I heard about this technique today (thanks @cherrvak)

third one is an interesting challenge - but perhaps a control-vector approach could also be relevant here? if you could figure out how a description of some relevant piece of game state affects the processing of the model, you could then apply that as a control vector when generating output. so the bridge between the game state and the LLM would be a set of weights for control vectors that are applied during generation.

this one is probably something where finetuning the model, and using control vectors to maintain a consistent 'pressure' to act a certain way even as the context window gets longer, could help a lot.

probably the vocal performance problem will improve in the next generation of voice generators, I'm certainly not solving it. a purely text-based game would avoid the problem entirely of course.

this one is tricky. perhaps the model could be taught to generate a description of a plan or intention, but linking that back to commands to perform by traditional agentic game 'AI' is not trivial. ideally, if there are various high-level commands that a game character might want to perform (like 'navigate to a specific location' or 'target an enemy') that are usually selected using some other kind of algorithm like weighted utilities, you could train the model to generate tokens that correspond to those actions and then feed them back in to the 'bot' side? I'm sure people have tried this kind of thing in robotics. you could just have the LLM stuff go 'one way', and rely on traditional game AI for everything besides dialogue, but it would be interesting to complete that feedback loop.

I doubt I'll be using this anytime soon (models are just too demanding to run on anything but a high-end PC, which is too niche, and I'll need to spend time playing with these models to determine if these ideas are even feasible), but maybe something to come back to in the future. first step is to figure out how to drive the control-vector thing locally.

[datamined]

I took some time to really think about this for awhile, and as a working graphic designer, artist, art student, and aspiring art director I have a personal conclusion about AI art that I haven't really seen anywhere else that I think puts it into better perspective.

This isn't for debate here, so feel free to disagree but I dont want to hear it, but my opinion of AI and AI-generated artwork is that the AI is the work of art, not necessarily the images it creates. Artificial intelligence machines is the artful creation of the people that built and programmed it. I support people who want to use ai-generated images as inspiration or as a tool to jump off of, but as an artist and a designer, my interpretation of AI "artwork" is that someone who pays for DALL-E or uses any other similar machines are merely participating in a work of art, not necessarily creating it. Just as people who participate in performance artwork aren't themselves the creator of said artwork. I do not view the images that an AI generates to be "real art", it is an amalgamation of pre-existing images mushed up together by a participant. But the AI is not making the decision to input a certain color in a certain place, the AI is not making the decision to exaggerate or tone down this or that. It simply does what it does because it does. It's a thoughtless process for an AI. The AI is creating an output, much how the Human uses the hammer to push a nail into a a fence. Technically and literally speaking, yes. A hammer pushed in the nail through the wood. But a person made the decision to do that. And it was a human's decision to use these processes to create a fence. AI artwork generation, to me, is the same process with just added steps. 1. The creator writes a program that, upon anyone's input, the user can randomly generate an image within the confines of the rules the creator (or in this, case, the Artist) puts in place. 2. The participant specifies a specific frame in which the program (that the artist made) can produce something.

3. The program, like any other tool, takes that input and performs an output. 4. A result generated by that program is made. It's similar in process to open world games or games with multiple endings. Results are made depending on the way the player plays the game. But the player did not literally make those things happen due to their own choices. The creators of the game gave the player options to choose from, and the player chose within that set of options. The act, and the result, is all an output created by the game developers, the game directors, the producers... The player has a personal interpretation and emotions to the artwork that the creators made for them.

It's something i've been thinking of because appropriated art and collage are both valid forms of artwork documented and credited in art history, and to me, AI is sort of like an automated version upon the same concept except a participant can "customize" the images that collage creates. So I was trying to wrap my head around the topic based on my studies and understanding of art throughout history and how it can be seen in the future.