My decidedly anti-AI friends won't like this post and may wish to skip it.
As you may have noticed, I am very ambivalent about the technology. It seems to demean human creativity, burns through energy recklessly, enables new forms of deception and control ... but it also unleashes remarkable new avenues for a kind of creativity that collaborates with collective memory.
The best AI image generator, Midjourney, just released a video generation system that yields very impressive results. What's the next step? The company says:
... we believe the inevitable destination of this technology are models capable of real-time open-world simulations.
What’s that? Basically; imagine an AI system that generates imagery in real-time. You can command it to move around in 3D space, the environments and characters also move, and you can interact with everything.
An old dream of creating a real dream, now becoming real.
You can make a video from an image you've uploaded or generated with a text prompt. The video can be extended (currently) up to four times in segments of a few seconds, generated either purely by the AI or in collaboration with your instructions. This gives ordinary mortals the power to assemble a little movie that illustrates their ideas.
Here's a mini-film I made. I wanted to start with a creepy atmosphere. My prompt for an image was:
A scene inspired by a story from the bizarre cult classic The King in Yellow. A small deformed man secretly controls the actions of distant victims. This is a retro vision of a bygone future. The style of the scene is low-saturation, low contrast, sickly colors, inspiring an undefinable anxiety.
Midjourney generates results in sets of four, so you can select the one you like best (or start all over again). I picked this one.
But it didn't feel just right, so I asked the AI to create variations on the image, and I selected this one.
Then I used it as the basis for assembling this video:
We've entered a new era, for better and worse.
So, to get back to the official theme of this blog: Can a typewriter do this?
No.
However, there are at least two interesting possibilities for high-low-tech collaboration here.
1. Brainstorm at the typewriter, letting your words run free, until some sequences of visions appears in verbal form on the paper. Then give those words to the AI and see how it "imagines" them. Tweak the results until they satisfy you.
2. Start by making an AI video, then sit at the typewriter and describe it, finding the best words to evoke the specificities that have emerged in the film.
Whatever the evils of AI, I am sure that creative people in our community are going to find new forms of digital-analog interaction that will stimulate new thoughts and emotions.
Hi Richard !
ReplyDeleteI have a lot of respect and even, dare I say, admiration for you and your work. That's why I don't understand how an intellectual, a philosopher like you can write a sentence which begin by “whatever the evils of AI”! Yes, AIs are fun to play with, but one can't let fun disconnect one's moral sense. It would be like voting for Trump because, “whatever his evils”, he says fun things. The evils of AI, whatever its impressing results for creating texts and images, are too important to just overlook.
It's certainly not going away. I would argue that thoughtful people have some responsibility to try it, to get to know its powers, and to see whether it can enable some good as well as bad. Yes, my little experiments have contributed a bit to the energy drain that AI is causing, just as when I drive my hybrid to the grocery store, I contribute a bit to climate change. Or when I post on this blog, I contribute a bit of information to a Silicon Valley surveillance capitalism company. Keeping one's hands clean is very difficult.
DeleteOne more thought: the general theme of this blog, and of my book, is the role of a particular mechanical technology in an increasingly digital world. I need to keep up with major developments in that world, and experience them firsthand, in order to comment intelligently on them.
DeleteI will miss my friend Mr. Polt, now that he has obviously been replaced by a robot. :P
ReplyDeleteCreepy.
ReplyDeleteIndeed - incredibly impressive.
ReplyDeleteConsidering then that it is the impressive and seamless stitching of pre-ingested motion-snippets, it has me wondering: How can the quite enormous energy-hunger and cost of such Laputan-engines be lower overall cost than a relevantly skilled human? For text certainly, or also for video-animation. (Even if stochastic and video; AI is imho still Laputan in essence. It is an impressive formatting of "web-search results".)
How is energy priced so low; of have the companies a loophole to free-energy and/or free-money?
It is all a bit bewildering...
The energy use is skyrocketing. Tech companies are starting up old nuclear reactors and investing in nuclear fusion. It's all kind of crazy.
DeleteOn the purely economic question you raise, I don't have exact figures, but for me to buy the traditional equipment it would take to make this tiny film, or to hire a professional to make it, would be prohibitively expensive. Or else I can pay Midjourney $10 to use it for a month.
DeleteTrue, very true - it lowers the barrier of access to creation.
DeleteThe 10 per month is still puzzling, given the billions that are allegedly being sunk into this; nuclear reactors aren't to be had for pennies either I'll wager.
Similar to e.g. Uber - wondering if AI pricing is (still) dissociated from cost. Whichever way - AInteresting times ahead : )
I should add that there is a limit to the processing power that $10 buys. I used mine up pretty fast! (And will not be buying more.)
DeleteThe end is near..
ReplyDelete