Conversations with novelist Jennifer Egan and AI pioneer Demis Hassabis, and a new profile of software visionary Douglas Engelbart.
I mentioned a month or two ago that I was taking over as the host of The TED Interview podcast, which has been true delight so far—and unsurprisingly many of the initial conversations have drifted into some of the core themes about the creative process (and how technology can enhance it) that we’ve been discussing here at Adjacent Possible. The most recent episodes have been particularly relevant, so I thought I would share a few key excerpts from them, though naturally I encourage you to listen to the conversations in their entirety.
The first conversation was with Pulitzer-Prize-winning novelist Jennifer Egan, author of A Visit From The Goon Squad, Manhattan Beach, and several others. Her latest book, The Candy House, is a sequel of sorts to Goon Squad, and shares much of that book’s extraordinary formal inventiveness: each chapter jumps around in time and perspective; one is composed entirely in 140 character tweets. (Goon Squad famously had an entire chapter structured around PowerPoint slides.) Egan and I covered a lot of ground in our chat: we talked about Dungeons and Dragons, the role of the novelist in making sense of technological change, non-dystopian sci-fi, GPT-3, and much more. But at one point I asked her how she comes up with her stylistic innovations, and she had a very interesting response:
I often have a wishlist of things that I would like to try. So for example, PowerPoint was on that list long before I was able to use it because it's actually really hard to write fiction that works in PowerPoint. I definitely had a wishlist in my mind as I worked on Candy House. I hoped I could write something in the first person plural; I hoped I could have an epistolary chapter completely in the form of letters; I really wanted to use Twitter, at 140 characters. I actually wrote that chapter much earlier because of the kind of inadvertent poetry and the kind of the serialized nature of reading a story on Twitter. So there were certain things I knew I wanted to do, but that in and of itself doesn't lead to anything.
My entry point when I actually write fiction, oddly enough, is time and place. And I write my first drafts very improvisationally because I'm looking to get beyond what I can think of consciously. My conscious ideas are not good enough; frankly, they're not original. So I've got to get out from under those and get to something that surprises even me.
And what I have found so far, knock wood is that, in the end, I'm usually able to imagine my way into a story that ultimately requires a form from my wishlist, but it takes a lot of trial and error. I have to find a story that can only be told in an unusual format. So I'm looking to what I write to tell me how to write.
I found Egan’s description of her formal wish-list here very familiar, even though I write non-fiction, not novels. Two of the books of mine that I’m the most proud of—The Ghost Map and Enemy Of All Mankind—originally began with a very hazy sense of the kind of book that I wanted to write, with no specific idea in my head for the actual subject matter. (I wrote about the strange origin story of Enemy in an earlier post—I sat on this intriguing idea I had for the structure of the book for about five years before I finally found the content that would make sense for it.) It occurs to me that keeping a “wish-list” of intellectual/creative challenges, even if you’re not exactly sure yet what the exact subject matter will be for those challenges, would be a productive routine to have, for writers and non-writers alike.
The latest TED Interview episode is a conversation with Demis Hassabis, the founder of DeepMind, and arguably the leading thinker in the world of artificial intelligence right now. If you share my interesting in the long interaction between games and AI—and also my obsession with “sandbox games” like SimCity—you’ll want to listen to the first section of the interview, where Hassabis talks about his years as a teenager designing classic sim games like Theme Park and Black & White in the 1990s. But at the very end of the interview, I mentioned the classic essay written by Ada Lovelace in the 1830s, where she predicted that future computers would be capable of creative work, not just calculation, and asked him what where he thought AI was currently in terms of genuine creativity. Here’s what he said:
This is a fascinating question. I would put creativity into three buckets. If we define creativity as coming up with something novel or new for a purpose, then I think what AI systems are quite good at the moment is interpolation and extrapolation. Interpolation is sort of averaging from examples—so you give it lots of images of cats and it generates a new cat. So it’s a kind of sophisticated averaging. Extrapolation is more like what AlphaGo did, which is play 10 million games of Go, look at human games, come up with a new Go strategy or chess strategy that's never been seen before…
But what's missing is true invention. And you can see that because our systems like AlphaZero and AlphaGo can invent new strategies for Chess and Go, but they can't invent Go. So that would then be the highest level of creativity: can you invent a game as great as Go or as great as Chess? That they can't do… And I think if we solve that, one could then have systems that do what we would regard as true creativity. Imagine [giving the instructions]: invent me a game that I can learn in five minutes, but that cannot be mastered in many lifetimes, but only takes four hours to play, and is beautiful aesthetically. Something like that. But all of those words are super high-concept. I mean, maybe I should type that into one of our language models and see what it would do, but I'm pretty sure it wouldn't come up with Go, right? And so, that's the kind of instruction I think we'd like to give our systems.
I think about what we’ve done with AlphaFold—amazing big advances in life sciences—but what would it take for a system to come up with general relativity? And then really advance our knowledge of the world and physics? Ultimately what I want to do with AI actually is understand the universe around us. That's the whole reason I've worked on AI my entire life. And that question, I think, is going to require true creativity—and we're not there yet.
The final item I wanted to mention is a new profile in the Hidden Heroes series, this one on legendary software pioneer Douglas Engelbart, who invented the computer mouse and helped lay the foundation for what would become the graphic interface—all of which he publicly shared during a conference in 1969, in what would come to be called the “mother of all demos.” You can watch the whole thing here:
I had known about the Engelbart story for decades—I wrote about the demo briefly in my very first book, Interface Culture—but in researching this profile, I realized that part of the innovation story here was not just the software, but all the creative and technical work that had to go into actually inventing the genre of a high-tech product demo itself, including some crucial contributions from a young Stewart Brand, who had just made a name for himself creating elaborate AV experiences as part of the TRIPS festival in San Francisco. In the profile, I write:
It is telling that the version of the demo that you can now watch on YouTube begins with a few introductory words almost exclusively focused on the technical innovations behind the stage show itself, and not the NLS system that Engelbart was introducing. (“Behind the scenes, Bill English coordinated the supporting crew who managed cameras, switches, mixers, special-effects controllers…”) Today, of course, we take it for granted when people gather onstage and project images from computer screens, interacting with other computers in distant cities; Steve Jobs turned that genre into a new kind of popular entertainment, most famously in his introduction of the iPhone in 2006. But the genre itself was first conjured up by Douglas Engelbart and Bill English—with a little help from Stewart Brand—back in 1968.
The Engelbart story is also a wonderful case study in collaborative innovation, and the strange tendency of certain places at certain moments in time to produce a disproportionate number of new ideas:
A few decades ago, the musician and artist Brian Eno coined a term to describe the collective IQ of creative hubs at their peak: Florence in the 1500s, Harlem in the 1920s. He called that group creativity “scenius”. By the time Engelbart and English started thinking about staging their demo, the Mid-Peninsula scenius quotient was remarkably high for a patch of land that had been orchards just a few generations before. Right up the street from SRI was the headquarters of the radical left magazine, Ramparts. The Grateful Dead played late-night shows at Kepler’s Books around the corner. The social clusters of computer hobbyists that would become the Homebrew Computer Club in the next decade were meeting up over coffee or drinks, plotting the revolution that was to come.
BTW, if you’re interested in the long-running “scenius” of the Bay Area, I recommend John Markoff’s new biography of Stewart Brand, Whole Earth, which I drew on a little to write the profile of Engelbart. It occurs to me writing this just now that someone—not me, I’ve got too many projects already—should write a book that’s a survey of a handful of different examples of scenius at work over the centuries, one “scene” per chapter. Maybe it’s already been written, but if not perhaps someone in the Adjacent Possible community wants to have a go at it? You’d be guaranteed a blurb from me!