Creativity Came to Pass

This is written of a future. Not the future, but a future – one of many possibilities.

When AI as we know it now first came to be, it was primitive. Examples from the first few years of what we know now as the AI age are rudimentary. Text generation didn’t take long to get to a decent enough stage for basic tasks, but visual creation, music composition, and other complex forms of creative expression took a bit longer.

I’ve seen some of the images generated during those early stages. They were hallmarked by distorted appearances, fuzzy definitions, and deformed limbs – hands seemed like they were especially difficult. Ethereal, low-resolution abstract was its only visual style. Early music often felt generic or repetitive, and initial attempts at complex writing or design lacked nuance as well. Most people wrote them off.

Only a few years after that, though, things had progressed a lot. Image generation still struggled with lots of details, but overall quality had improved a lot. Music became somewhat tolerable and novel. Real progress started being made, and people began using it in place of human-made work, though it was still inferior to human work and there were obvious AI tells. In spite of the flaws, not having to compensate creators was seemingly incentive enough for some people to switch over.

There was a pretty big pushback from a lot of people, especially artists who worried AI might take their jobs or who thought it wasn’t in the spirit of art. This pushback only got louder and more vocal as it became truly competitive with human artists, which it did. Many people stopped sharing their work, or creating it at all, because they didn’t want it to be scraped.

It only took another few years for things to get a lot better. Resolutions improved a lot, and things seemed to interact pretty normally. While many aspects still looked quite funny, hands at least looked a lot more hand-like in images. Videos also began to look cohesive, rather than an inconsistent, cobbled-together collection of incohesive images played in sequence. They were decently realistic but appeared to struggle when objects interacted. They morphed through each other and didn’t seem to have a grasp on physics.

AI generation of art fairly quickly reached parity with human efforts, and the lack of ability for consumers to differentiate led companies to embrace the technology. From a corporate standpoint it was nothing but wins: faster turnaround, potentially infinite variations, and easy tailoring to specific requests. They no longer found a need for their composers, graphic designers, writers, painters, and other creatives and made them redundant.

People began really leaning on AI tools for all sorts of tasks, even menial ones they are well equipped to handle. It didn’t take long for people to start realising that the skills they entrusted to AI had fatigued and that they could no longer do the things they once could. Some people found cause for alarm, and other people figured those skills were irrelevant if AI was capable of doing it for them. Alongside this, AI systems became better and better at completing tasks with little given information and vague prompts, allowing people to provide more generic prompts without need to deliber.

The tells of still, visual art were gone for all but the most discerning eye, then only detectable when specifically analysing, and finally indistinguishable. Music was much the same, and video took a bit longer but got there not long after. Most people never even noticed the proliferation of AI art, and those who did often didn’t care. People failed to note when AI music took over their playlists, AI photos and videos took over their feeds, and that there wasn’t a human behind most of the accounts they interact with.

With most artists left without the ability to make financial return from their work, they turned to whatever other careers and professions they could find. The creative industries were always turbulent, but AI thrust them into shambles. Those who managed to hold onto employment for a bit longer by the skin of their teeth found their wages plummeting. Supply of creative talent was exceedingly high, and demand exceedingly low.

It slowly became less and less viable to spend time on art as AI displaced more and more jobs. People’s efforts moved on to remaining afloat in the current climate rather than flourishing and creating. Lacking disposable income, artists found it difficult to justify the purchase of new tools or time to use them.

Some famous artists kept creating somewhat successful works for a while as they remained respected and distinguished, but slowly they found their audience dwindling as time marched on. As distinguished figures stopped creating, grew old, and died, nobody new came to replace them in the zeitgeist. Those few who remained in admiration of them and their work grew old and came to their lives as well. Without anyone noticing, the last generation of notable artists came to end.

Artists who remained creating very rarely received recognition for their work and couldn’t find distinction from the AI-produced work. A few continued to create and practise for the love of the craft, though the lack of distinction slowly dwindled their motivation. Of the few remaining, some tried to pass their knowledge onto the next generation, but that generation was entering a world where art isn’t valued. They were a generation raised on perfectly tailored digital experiences and saw little point in the added effort of making something themselves.

The algorithms generated exactly what the children wanted to see, and most grew up never seeing a point in creating things of their own. They saw what was given to them, and most never really developed a curiosity or drive to create. Nothing AI spoon-fed them was challenging; it simply catered to their desires. Despite this, a small few of the new generation did pick up the creative spark and continued making, though their abilities were stunted by the lack of community, mentorship, and perceived value.

As the generations stretched on, oral and undocumented knowledge, techniques, and philosophies were gradually lost. As AI as a whole continued to become more capable, technology moved with it, and more and more people were displaced. This uncertainty only led to a further lack of motivation which further stifled creativity.

AI continued to become more and more prevalent as the dominant technology. The AI we gained wasn’t misaligned – a lot of safety work had ensured that – but it was disincentivised to discuss human creative output. This was both due to artists’ vocal distaste for AI and their deliberate avoidance of scrapers to prevent their work from ending up in training sets.

The initial pushback against AI art online found itself deep in the training data of LLMs before AI slop took over, which more often than not led to a higher ranking and trust within LLMs trying to avoid mode collapse. This early pushback was often crass and “considered hostile” due to a strong association with AI doomers and anti-AI sentiment that led to models downranking it and considering it inappropriate to discuss. Later discussion where people were more informed scarcely made it into model’s understandings.

There were many more nuanced takes on personal websites and in private discussions, but they never made it into training data because creatives prevented scrapers from scraping it in an effort to protect their work. Unwittingly, they removed their work, as well as their values, opinions, and thoughts from AI training, which led to the creative mindset being under-represented in the data on which the models we now rely upon so much were shaped.

Some people had noticed when AI summaries started to fill their search pages, but it was very gradual that the traditional search results started to take up less and less page space and then became part of a dropdown until eventually they weren’t displayed at all.

Websites were already struggling, but now it was dire. No referrals meant no ad revenue. Not that ad revenue meant much by that point anyway. Impressions were worth nothing due to the sheer number of AI agents pretending to be humans and the small number of people accessing websites directly anyway. Websites are falling by the day. Nobody actually uses them anymore. Why bother when the AI can give us the information they offer? All the people who created their own websites are gone now. Some people kept around the websites of relatives, but those relatives are now well estranged by time, and their sites continue to go offline.

Art of the digital age was slowly lost. “Nothing ever gets deleted from the internet” became a fallacy. The ‘web’ of today is a legacy technology that remains a footnote in an LLM’s response when asked, “How was AI trained?” – not that anyone would bother to ask such a thing.

It didn’t help that so much human-made art was created physically – paintings, sculptures, crafts, handwritten manuscripts, physical photography prints, sketches – which meant it never made it into training sets. Without a digital footprint or representation in the dominant AI knowledge systems, this vast body of work effectively became irrelevant.

As art became less popular and computers more prevalent, simple tools unique to specific crafts, like specialised brushes, sculpting tools, weaving looms, or darkroom equipment, found themselves out of manufacture due to lacking demand. More complex tools like music production equipment or computer drawing hardware had fallen out of production and grew incompatible given the swift speed of computer advancements. Even if they could be acquired, the skills and knowledge to use them couldn’t.

Because creatives had excluded themselves, their values, and their opinions from model’s knowledge, they found themselves disregarded by history. LLMs are the main way of education and learning and have lost the ability to educate people on human-created art. Their training is muddled, and they’ve lost definition of ‘art’. The exclusion of discussion of human-made art in training sets and the proliferation of AI-generated art led to models thinking they are one and the same. That art as a whole is, and was, a thing that AI generates. ‘Art’ became synonymous with ‘AI art’.

AI models serve as yes men, and, as such, most people don’t experience much in terms of restrictions. Restrictions, lived experiences, and limitations bred creativity; without them, there is nothing. AI has done a lot for making life more comfortable for humans, but at a cost – the world is abstracted away from physical flaws and into digital perfection. Children born today don’t really have emotional depth, nor any reason to bother ‘creating’ – neither were important in a comfortable AI world. Art is fundamentally a byproduct of environment, but the environment is bland, sterile, and free of any extreme influence. If you want to watch a TV show or read a story, the AI just generates one tailored for you based on what you’ve liked before – no humans needed and no chance for the uncomfortable experiences that spur creativity.

People didn’t notice when human creativity died. It was simply replaced. AI made it obsolete, at least as far as anyone today cares.