Bruce Sterling is a well-known SF writer, but I think he is not quite on the radar of Less Wrong, so I'll say a bit about how I locate him culturally. I'll make a distinction between SF that is primarily about marvels of science and technology, and SF that is primarily about how it feels to inhabit technological society. I consider Sterling to be a leading practitioner of that second kind of SF. He wrote one novel, the posthuman space opera Schismatrix, in which he indulged his inner Singularitarian to the maximum; but after that, his works have been more earthbound, though still about the transformations of life that could be wrought by technology. 

He's also been prolific as an essayist and public speaker, still offering commentary on technological society and world affairs, but in a supposedly factual vein, rather than an overtly fictional one. I admire his talent for digesting these complex topics and producing an intuitive summation in a few pithy turns of phrase; I consider him a role model in that regard. And so now he's trying to sum up the state of the AI world, after ChatGPT. 

He aims for an indulgent, god's-eye view of the current-year antics of everyone involved in AI. He acknowledges the prophets of doom, the social justice warriors, the captains of industry. He nominates three myths as encapsulating the times: Roko's Basilisk, the Masked Shoggoth, and the Paperclip Maximizer. Every one of them comes from within the rationalist sphere, or adjacent to it, so we're doing pretty well when it comes to mythic potency! 

But he does believe that after this carnival of ecstasy and dread, there will be a hangover rather than a singularity. I suspect he's the kind of guy who thinks that superhuman intelligence might well be possible, but only decades from now, after still-unknown paradigm shifts, and that language models are a false dawn in that regard. So when the next AI winter comes, we'll still be here, and we'll still be human beings, but we'll also be living with all the consequences of having created glib, multitalented, relentless artificial persons. 

My own bet is that this isn't a false dawn, that for better or worse, we're now directly on course towards artificial intelligence that comprehensively transcends human abilities. But I appreciate Sterling's role as skeptical yet open-minded observer. 

New Comment
1 comment, sorted by Click to highlight new comments since:

As far as Bruce Sterling's views go I would expect he still has similar thoughts as when he wrote https://www.edge.org/response-detail/26093 .

Intelligence might not be the best way to frame the behavior of ML agents. Debating whether or not an agent is intelligent obscures a lot of other aspects of ML agents.