Zach Seward of the NYT has published as an article his 2024 SXSW talk on AI and journalism. It's very good! I'm in full agreement with his conclusions. Approved usage for AI: providing assistance with search, data organization, and other creative research methods. What it should not do: write content while pretending to be something it is not. There are legitimate ways to incorporate AI into newsrooms, and Seward has outlined great boundaries around how one might.
Beirut on AI
Michael Beirut with a great take on a benefit of AI:
"A majority of the world is boilerplate, and that's a good thing," Michael says. "Because all of us are motivated by two opposing kinds of needs. One is the need for predictability and comfort. And the other is the need for surprise and excitement. If you get that balance wrong and everything is just comfortable and predictable, you get bored. And that's not good. But on the other hand, the antidote to that isn't 100% unpredictability, surprise, and entertainment, because then you get too overstimulated and start to freak out. There's a certain reason why you wouldn't want every building on Fifth Avenue to look like the Guggenheim. It just would be too crazy in an unpleasant way. This architect was admiring some anonymous building somewhere in Manhattan, and they said, 'You know, what makes a city great is its ability to put up and maintain something like this, a perfectly good kind of background building that actually does its job well and isn't calling attention to itself.' But if you examine the details, you see that they're sound and built to last, and the tenants are satisfied. And then down the block and around the corner, there's something more special and showy that provides a great foil for that.”
So maybe ChatGPT, Midjourney, and all of those other new generative AI programs can be responsible for doing the boilerplate really well. Now, you've got time to do something a little more interesting. "Everyone dismisses boilerplate as crap that's not worth anyone's time," Bierut says. "You could argue boilerplate makes the world go round, whether it's in architecture, design, or writing. Boilerplate makes things that are special look and feel more appropriately special. Maybe that's an appropriate way to think about the promise of AI, but I don't know."
This tracks with how I've been thinking about AI with respect to design (or any creative profession). It's going to be a tool which helps with boring, rote elements of a process. It'll be able to recreate and fuse older iterations of a craft in order to more quickly produce what had come before. It's just not going to be reliably able to make something humanist and new. At some point, the boundaries of what AI can produce will be obvious, and human beings will bend culture back to what is more fully human.
You can't believe the tech bros about this. They're going to claim AI will be possible of anything — including self-awareness and then becoming our new god. Whatever AI creatures they do eventually create will look and act like them, because people are only able to create gods in their own image. In the case of AI, that god will also be white, male, and techno-centric.
Personally, I'm looking forward to seeing how art and society eventually push back on the upcoming AI-obsessed culture. It should be really human. Really tactile and emotional. Think about how Grunge reset the table on music as we turned into the 1990s. AI stuff will always look clean and rich. It'll have a pristine quality to it, even when prompted to not. Because it's expensive. And it's run by rich people. The pushback to it is going to be a good time. It's going to be the Guggenheims on a street full of forgettable buildings.
