Thoreau’s adage “beware of all enterprises that require new clothes” should perhaps be updated to “beware of all enterprises that require venture capital.”
Morozov argues that AI itself has much to offer, but it has not lived up to its potential to serve the public good, and the context of AI’s development explains why. I agree. My own misgivings about AI have less to do with the technology itself than with the problematic nature of who owns it, and what they want to do with it. Venture capitalist Marc Andreessen’s wildly hubristic visions of the future are par for the course in West Coast technology in that they downplay even the possibility of any downsides, brusquely dismissing these as “safety-ism.” I for one wish there had been a few more “safety-ists” around when the algorithms for social media were being crafted.
If a company is run primarily for profit, you’ll get entirely different outcomes than if it’s run for the public good—despite what the true believers in the “invisible hand” of the market preach. Social media provides the best example, and the experience of what happened with social media is a bad omen for what might happen (and is happening!) with AI. Two words—“maximize engagement,” code for “maximize profits”—were all that was needed to send social media into the abyss of spleen-venting hostility where it now wallows.
The drive for more profits (or increasing “market share,” which is the same thing) produces many distortions. It means, for example, that a product must be brought to market as fast as possible, even if that means cutting corners in terms of understanding social impacts; it means social value and security are secondary by a long margin. The result is a Hollywood shootout fantasy, except it’s a fantasy we have to live in.
AI today inverts the value of the creative process. The magic of play is seeing the commonplace transforming into the meaningful. For that transformation to take place we need to be aware of the provenance of the commonplace. We need to sense the humble beginnings before we can be awed by what they turn into—the greatest achievement of creative imagination is the self-discovery that begins in the ordinary and can connect us to the other, and to others.
Yet AI is part of the wave of technologies that are making it easier for people to live their lives in complete independence from each other, and even from their own inner lives and self-interest. The issue of provenance is critically important in the creative process, but not for AI today. Where something came from, and how and why it came into existence, are major parts of our feelings about it. We feel differently about a piece of music played by an orchestra in a concert hall than we do about exactly the same piece of music made by a kid in a bedroom with a good sample bank. The backstory matters! The event matters! The intentions matter! We have no idea of the actual origin of the text AI delivers to us. Does it matter that what we’ve scraped off the ether to feed our AIs is not by any means the whole of the world’s knowledge, but just the part that happened to have been published in printed books by the small sliver of the English-speaking world that happened to publish them—and made them available to AI bots? What kind of sausage is that? Surely Weisswurst, made of available scraps on the butcher’s floor.
AI is always stunning at first encounter: one is amazed that something nonhuman can make something that seems so similar to what humans make. But it’s a little like Samuel Johnson’s comment about a dog walking on its hind legs: we are impressed not by the quality of the walking but by the fact it can walk that way at all. After a short time it rapidly goes from awesome to funny to slightly ridiculous—and then to grotesque. Does it not also matter that the walking dog has no intentionality—doesn’t “know” what it’s doing?
In my own experience as an artist, experimenting with AI has mixed results. I’ve used several “songwriting” AIs and similar “picture-making” AIs. I’m intrigued and bored at the same time: I find it quickly becomes quite tedious. I have a sort of inner dissatisfaction when I play with it, a little like the feeling I get from eating a lot of confectionery when I’m hungry. I suspect this is because the joy of art isn’t only the pleasure of an end result but also the experience of going through the process of having made it. When you go out for a walk it isn’t just (or even primarily) for the pleasure of reaching a destination, but for the process of doing the walking. For me, using AI all too often feels like I’m engaging in a socially useless process, in which I learn almost nothing and then pass on my non-learning to others. It’s like getting the postcard instead of the holiday. Of course, it is possible that people find beauty and value in the Weisswurst, but that says more about the power of the human imagination than the cleverness of AI.
All that said, I do believe that AI tools can be very useful to an artist in making it possible to devise systems that see patterns in what you are making and drawing them to your attention, being able to nudge you into territory that is unfamiliar and yet interestingly connected. I say this having had some good experiences in my own (pre-AI) experiments with Markov chain generators and various crude randomizing procedures. Any reservations about AI get you dismissed as a Luddite—though it’s worth remembering that it was the Luddites, not the mill owners, who understood more holistically what the impact of the new mill machinery would be.
To make anything surprising and beautiful using AI you need to prepare your prompts extremely carefully, studiously closing off all the yawning, magnetic chasms of Hallmark mediocrity. If you don’t want to get moon rhyming with June, you have to give explicit instructions like, “Don’t rhyme moon with June!” And then, at the other end of the process, you need to rigorously filter the results. Now and again, something unexpected emerges. But even with that effort, why would a system whose primary programming is telling it to take the next most probable step produce surprising results? The surprise is primarily the speed and the volume, not the content.
In an era when “cultivated” people purport to care so much about the origins of the stuff they put into their mouths, will they be as cautious with the stuff they put into their minds? Will they be able to resist the information sausage meat that AI is about to serve them?
