Robin Sloan is always a yes. Whether he’s in Oakland, Menlo Park, or YouTube I’m always excited to see his work. So when I saw he’d be in San Francisco talking to one of the Anthropic cofounders, Jack Clark, I jumped on it.
I, for one, am excited for the influx of Anthropic IPO money into San Francisco1. They sponsored the Long Now talks last season and now they’re starting what’s presumably an arts series. And they put out a fantastic salmon spread.
The crowd was interesting: all of the folks I met were there from Jack Clark’s newsletter and not Robin Sloan’s. I, of course, didn’t know who Jack Clark was but it let me make a great pitch for Robin’s work (hot take: Sourdough -> Mr Penumbra -> Moonbound is my Machete order). They handed out free books after so hopefully some of my targets follow through! And it made it less awkward striking up a conversation on the Muni after the talk as it was clear who the other attendees were.
The conversation was remarkably convivial: it seems like they’ve known each other for a while. Apparently Robin Sloan had been messing around with early early language models and Jack Clark reached out and demoed one of the early GPTs for him. It’s fascinating to see Robin, who’s firmly a denizen of both the arts and the tech scene, and how he melds the two.
But apparently not in his work. He mentioned trying out some of these tools for writing but it just wasn’t really working. Jack Clark agreed: apparently he does ask Claude to fill in some robot dialogues for him which is adorable. Indeed both seem to embrace a bit more of a return to the physical: long walks and conversations for Clark and olive oil making and printing for Sloan. And there’s something special about artisanal words: someone spent time on it and that’s meaningful.
What I had hoped they’d get into was a discussion of what LLM’s role is in the arts. This is clearly pretty controversial: slop is everywhere it seems and seemingly unavoidably addictive (they compared it to “The Entertainment” in Infinite Jest). But there ought to be something in there: it feels like with each new foundational information technology we get a new narrative medium that everyone hates until it’s loved: Socrates despised writing, the church despised the printing press, [[https://en.wikipedia.org/wiki/Neil_Postman][highbrow critics] despised television, and of course everyone despises social media2. So what’s it going to be with AI? They mentioned some AI “ethnographers”: folks who spend an incredible amount of time with AIs and stream it – presumably as an art form? Perhaps in a decade it’ll be obvious but at least for now I’m still waiting on this new form – and I sure hope it’s not slop.
But the one point that they kept coming back to is this: fiction is, of course, strange but the current reality is also rather strange. Sloan had a good line: writing fiction is like sampling from a very high-dimensional space and I think this fits more broadly to life as well. The whole contingent nature of things is rather remarkable: one could not have guessed back almost five years ago that we’d be here. It’s even remarkable thinking back two or even one year ago. Jack had a great observation: writing fiction is practice for imagining that the world can be different, which is the first step for changing it. And change it they certainly have.
-
Anthropicinians please consider emulating Francisco San in addition to funding your esoteric EA causes ↩︎
-
Maybe we’re right about that one? ↩︎