A spiritual follow-up to Culture is Stuck by an amateur cultural analyst.
What is art? The Mona Lisa is a great work, because for its time, it demonstrated remarkable mastery. Nowdays, many can replicate it with ease. Generative AI will reduce this cost to near 0.
Art also means evoking feeling - a banana taped to a wall, is that art? Ironically, do the endless Facebook debates about whether it is art or not, make this a good piece of art?
How has art’s context changed? Art is also taste, style, and context. This piece is birthed out of my generation - generation Y. We remember just the cusp of when the Internet took over everything. When the first memes came out, and people started remixing templates - first demotivational posters, then ragecomics, and then normies became aware of them. Photoshop and copy-pasting.
This led to a couple of things - virality, decentralization, and hauntology. Do you remember seeing your first viral video on YouTube? Something that literally everyone knew and quoted, like “Leave Britney Alone”. This mimetic behaviour has always been part of human culture - see “a dingo ate my baby” for an aussie example. But the tools to remix it have never been deployed so far at scale. Now suddenly everyone has their own channel. Instead of centralized syndication - where everyone watched The Simpsons at 6pm on a weeknight because it’s the only thing that was on - now culture is downstream of nowhere. It simply flows in all directions. While you could argue that Netflix replaced the TV, arguably TikTok occupies a much larger mindshare.
Does culture flow linearly? The other change to culture is that it never dies. Seriously - when culture can remain accessible forever, thanks to the Internet, what does that mean? Although they no longer make Seinfeld, I am part of many groups on Facebook where we continuously make memes applying Seinfeld to the current day. Just because the show is off the air, doesn’t mean it’s not living still to this day.
In fact, Seinfeld was recently reborn as a live show. But not by humans - by an AI. In December 2022, some developers released the Nothing, Forever show, which is a procedurally generated Seinfeld-esque show that streams on Twitch. Everything - the character movements, the classic Seinfeld bass sounds, the laugh tracks, the dialogue - is generated by AI. It is streamed realtime to Twitch, and the show literally never ends. It is a constant stream of new content, that you can tune into at any time. And what’s interesting - is that the content is up-to-date with current events. The dialogue mentions modern online dating trends, because the AI model was trained on the Internet.
Let’s just recap - in 1998, Seinfeld went off the air. By all means, this was the point where it should have died. But in the 2010’s, the Internet meant that the culture lived on - as people/fanbases remixed memes in groups. And now, in 2023, it lives on thanks to a machine. First the cost to distribute content, and culture, went to 0 when the Internet came out. Now the cost to generate that content, is also quickly falling.
Cultural nonlinearity. And that is where we are. Since the Internet, culture has been decentralized - instead of centralized syndication, from news network who decide what you watch, anyone can publish their own take, and consume it at anytime. Because culture flows in all directions, it feels less like a top-down person giving you instructions, and more like a space. A latent space.
What is the latent space? This is a concept from machine learning, that refers to the mind of the machine in terms of a coordinate space. At (0,0), you might have the concept of an apple, at (0,1) a banana, and at (512,212) the concept for something else entirely. Knowledge has these coordinates. And as the machine’s memory grows larger, it becomes like a library of Babel - containing every known concept and idea. When you prompt a machine learning model, you are in fact giving it coordinates - your words are converted to a word vector. And this is used to guide the model to producing an output. This is very powerful and leads you being able to do math with words. For example -
king + woman - man = queen is an equation that is not only real, but is computable by machines. A prompt is a more sophisticated example of this idea - the ability to combine different art styles in generative AI models reflects this same idea of “concept algebra” -
painting + surrealism + linotype.
Who explores latent space? This is coming to my weird observation. The entire AI industry is made possible due to the Internet - in order to train these models, they must experience the world, and where do we get such a massive dataset? The Internet. They also have the meta-ability to learn - AI is not just copying information, it can actually learn new information in-context.
Originally, it was the internet hivemind which made Seinfeld memes. Now it is a machine. The machine independently explores the collective memory of the Internet. But because AI can learn in realtime, soon, both man and machine will be making together. The Twitch stream will guide the AI to produce content in the same way Twitch Plays Pokemon works.
Machine-on-User collaboration. And so I did something. I made it so you could DM an internet meme. I prompted the AI with a general idea of a hot take, gave it a personality, and users could chat with it. The AI would come up with takes in conversation, and users could mint them, thus furthering the puzzle.
Machine-on-Machine collaboration. I created a meme social media called Take, where people could make text templates like “you wouldn’t download a [xx]” and people could fill in the blanks. The founder of knowyourmeme actually made a meme on it. This is an example of human-explored latent space. But then I built something called the Take Modern, an art gallery where these takes are translated into visual imagery. The art was not just generated by one AI - a collaboration of two. One AI generates ideas for art in the form of scenes (ChatGPT), and then communicates them to the image AI (DALL-E), which generates visuals of these scenes. The role I have to play is as orchestrator - I write the code which connects them - but even here, I use an AI to write the code (CoPilot). At every step, my agency is abstracted away and I direct more and more conceptually.
User-on-Future collaboration. All content published online is simply messages to a future AI. By putting it on the blockchain, the content is near-guaranteed to be indexed. And so this entire piece, artworks and this article, being minted as NFT’s, will become part of the latent space too.
The rise of generative AI has fundamentally changed the landscape of art and culture. With the ability to create new content at near-zero cost, AI has allowed for endless possibilities in the realm of creative expression. Additionally, the decentralized nature of the internet has further amplified this effect, allowing for a constant flow of culture in all directions. The latent space, which is the mind of the machine in terms of a coordinate space, has become a playground for both humans and machines to explore, collaborate and create. As we move forward, it is clear that the relationship between man and machine will continue to evolve, leading to new forms of art, culture and expression that we can only begin to imagine. (this was all written by ChatGPT)
This article is notes I wrote from building an AI image pipeline called the Take Modern. It was too big to explain in one post, so if you wanna read it, DM me on Twitter and I can send you the deets. -Liam