AI Disruption? You’re soaking in it!

2022-12-07

On a recent trip to the paperclip store, I was asked whether AI disruption was a real thing, or just some doomer Skynet fantasy. Oh, it’s real, I replied — but it’s more 3PO than T800. And should that worry you? It depends!

So far we’ve seen fairly predictable disruption from generative AI — and here it’s ChatGPT and Midjourney leading the pack. Let’s look at four examples that shed some light on where the current advantages — and dangers — lie, and what we might expect to follow.

A picture equaling one kiloword, our first example is concept art. That’s the imagery, now almost always computer-generated, that 3D and 2D artists use for mocking up the visuals when designing video games and, to a lesser extent, films. We’ve heard what’s happening in the video-game industry: the concept artists are pretty much Midjourney wranglers now. This was probably inevitable, since the job itself had so much in common with image-generating AI: take a description and make a picture that brings it to life within a particular genre. Then make another one. Repeat. Of course there will still be some great original concept artists translating their brilliant original visions, but there’s no reason to think they’ll work for major studios.

Next let’s consider customer support. We’ve long complained about the robots who call us and the robots we’re forced to talk to when we seek help from our bank, our service providers or (gods help us) the cable company. For those cases when the customers couldn’t be forced to beep through a menu, there exists a vast network of “phone support” offices in lower-wage countries. These jobs are going away faster than you can say “Sam Altman Regulatory Capture.” Because the one thing we all learned really fast when ChatGPT came out was that the “chat” part was for real. Smarmy is the default setting, though it’s easy to imagine edgy customer service being a thing in the future. We can laugh at IBM for trying to sell you an AI Blockchain, but when they say they’re not hiring people who can be replaced by AI, this is the first group of people they’re talking about.

Will AI make people happy with customer service? Nope. Will it make them less mad? Maybe. Will they just have to take it? Probably — and the people feeding it to them will cite endless metrics for why people hate the AI less than they hated the beep-through menu labyrinth, and why that reduction is chiefly responsible for their 5000% quarterly profit increase. And here too, you’ll have a number of people become AI Wranglers, a few become AI Whisperers, and the rest will have to find something else to do with their lives.

These first two examples show disruption in areas where the AI is self-evidently better at doing most of the job than the humans who have traditionally done it. There will be many, many more such areas: instead of ten humans and a supervisor, you’ll have two wranglers and a prompt engineer. The next two examples are not so obvious.

For the third case, consider a thing that took people very much by surprise and has implications we’re not even close to understanding: computer programming. The AI can code. And it can code better than your average Junior Software Engineer. It hallucinates sometimes, and it introduces bugs, but no worse than many of the Juniors.

At this point, a strong programming team of, say, five senior coders and one team leader has a fascinating force multiplier: an army of virtual interns to write code and a council of virtual graybeards to give advice. Asking is the bottleneck. But this presents a risk we’ve not seen before: a sort of industrial “demographic collapse.” You still need Senior Engineers, and they only grow out of Junior Engineers, but you don’t actually need the Junior Engineers now, so where do you get the Seniors?

Maybe you get them from the same source we mined in the days before all the careerists piled into “tech” — that is, from the enthusiasts and natural-born hackers. And maybe you don’t need so many of them. Maybe hiring people to prevent your competition from hiring them is not going to be a viable strategy anymore — after all, the truly committed hackers have never been available for Rest & Vest schemes. That’s a future decidedly unfriendly to the FAANG crowd, but we are here to talk about disruption.

Finally, we see an infiltration of that most canonical area of human creativity: writing itself. And not little essays like this note, which for all you know was written by ChatGPT 5.0 Beta, but the (more or less) original writing that feeds our endless stream of motion-picture content. In the current contract dispute between the Writers Guild of America and the Alliance of Motion Picture and Television Producers, one of the key points of contention is who gets to use the AI, and for what. One side wants it to be a writer’s tool, the other side wants it to be a writer. We can feel relatively safe that the AI is not going to write another Better Call Saul, but right now there’s a new episode of a formulaic show queued up on Apple TV+ with the tag line: “She will do anything to protect her family.” GPT 3.5 Turbo can already do better than that, and it’s the cheap one.

But there’s another disruption here, soon to be streaming down the line to a device near you: the normalization of the “AI Voice.” As much as it tries to sound like us, the AI always sounds like itself as well. In the next couple of years we’ll be hearing that voice a lot, even if there is a layer of human writers between the machine and the screen. This could be fairly benign: when Pixar normalized 3D rendering to the point of displacing most traditional cell animation, we got a ton of new titles and the kids didn’t mind. For connoisseurs, there’s still Studio Ghibli and friends. But the normalization of the AI Voice could also be darker, specifically in combination with ubiquitous social media to which a generation of heads is constantly bowed, or — maybe next year — wearing their iFaces. Who’s up for some Lord of the Flies as Goal-Oriented GPT?

Each of these disruptions is worth a long-form essay of its own, and yet each is just a small glimpse of the transformations that are here now and gaining steam. The AI Future will come slowly, then all at once. Your kids won’t ask you where you were, but they might ask the robot, and it’ll tell them… something.