What do the new AI models mean for society?
A friend sent me this image. I thought — well, I wonder if I could do better?
“Yeah, the artists are borked”
I’ve been playing with these generative AIs (large language models, or LLM’s) for a while. Could one of the latest models, ‘Stable Diffusion’ could do any better? I refine the prompt: “Taylor swift, renaissance painting, extremely detailed, accurate proportions, realistic face, beautiful eyes, smiling, detailed lips, accurate lips, accurate chin”
Well yeah, I guess it could. Could we get a bit more specific, like playing the mandolin? Kinda.
How about riding a tiger, wearing a traffic cone? I tried several of these but it didn’t do well. This was my favourite weird combo. A tiger print traffic cone?
After refining the prompt a bit I finally got these.
It’s a surreal experience. Yes, the results are mundane. I could have created something similar using Photoshop. But the experience is totally different. There’s a feeling of jamming, going back and forth with something that almost feels intelligent. But more than this I get a feeling of weightlessness, of floating in the unknown. It’s diving into black water. It’s rolling the dice in the dark.
Playing with these models, I’m reminded of William Gibson’s:
“The future is here, it’s just not evenly distributed”.
I feel like I’m floating, weightless, while the rest of the world remains firmly attached to gravity. It’s a feeling of limitless, strange power and potential. The closest analogy I have is when I first saw the internet working, as a student at Sussex University nearly 30 years ago. I had my mind completely blown when I saw that I could find almost anything, for free. Text and images would download from a server — anywhere in the world, anyone could put up a web page and I could see it, here in this small computer lab in South England. The first thing I started looking for were aliens of course, being a Fortean Times haunted teenager. In the Mosaic browser on a huge Unix machine, an image of a ‘grey alien’ scanned slowly into view, line by line of data emerging from who knew where to form a picture.
My first thought was: The internet changes everything. This changes all new media, everything about the future, and how we communicate and share. The horrible old big-money controlled media will disappear, and this will take over everything. I was right about everything changing, but totally wrong about the political implications. Peer-to-peer media turned out to be a hellzone when mediated by private companies looking to turn a profit. The Daily Mail is now the biggest news site in the world. But the change coming was real.
I might be completely mis-led, but this is how I feel about the new LLM’s. I don’t know how, but they will change everything. Images now, but soon: Music, animation, video, 3D, VR.
We’re creating models of intelligent processes that will enhance, empower, transform, and perhaps replace vast swathes of human cognition and work. I have no idea how this will pan out. I hope that it will empower and enrich humanity, and enable us to solve our collective crises. My fear is that it will mostly fund a few super yachts for rich white men, while impoverishing the rest of us. The political and collective choices we make over the next decade will decide.
If you want to have a play yourself, check out these links:
Craiyon — for fun and scrappy creative results
Dreamstudio — ‘Stable Diffusion’s key site and the source of all the images in this article
Midjourney — remarkable (if pricey) image generation through a Discord bot. You’ll need a Discord account to try it.
AI generative tools — an amazing resource if you want to go deeper.
