Sci-Fi Bulletin: Exploring the Universes of SF, Fantasy, Horror and Spy-fi!

Scottish SF author Ken MacLeod’s latest novel Dissidence is the first novel in his new trilogy, The Corporation Wars, described as “a space opera giving a robot’s eye view of a robot revolt”. Here the double BSFA Award winner discusses the role of robots and artificial intelligence in science fiction…

a

SF writers keep getting AI and robots wrong – and we should keep right on doing it. Here’s why.

Last Thursday a prestigious US law firm, BakerHostetler, announced that it was hiring – or rather, had a licence to use – ROSS, an ‘artificially intelligent attorney’. Based on IBM’s AI Watson, ROSS can take natural language questions and return relevant answers from legal text, improving as it goes along. If ROSS works as advertised, it won’t just save on an enormous amount of routine clerical drudgery – it’ll replace the work of an unknown number of expensively trained legal brains.

Oddly enough, the only AI in science fiction that this reminds me of is one of the earliest and most plausible: the City Fathers that administer NYC in James Blish’s Cities in Flight. This bank of computers acts as library and librarian, teacher and school, lawyer and judge, surveillance system, memory bank and political adviser. And throughout all the millennia of that galaxy-spanning space opera, the City Fathers remain a tool of the likewise long-lived mayor, John Amalfi. Only when facing the imminent end of the universe do they surprise him with a moment’s hesitation and a farewell.

But at no point do they rebel, or hatch plans of their own. In this respect Blish’s imagined computer bank – for all its vacuum tubes and teletypes – was closer to AI as we’ve increasingly experienced it over the past few decades than almost all SF written over the same period has ever envisaged.

As an immensely powerful tool for good or ill, AI matters to us all, and will matter more and more. In the real world, it would be feasible to automate drone strikes entirely, from the metadata trawl that identifies a high value target to the death from above delivered – if it hasn’t happened already. Alarming though that is, it isn’t what gets SF written, and sold.

No, what gets us going – as writers, readers, viewers and players – is Skynet and Terminators! It’s not the prospect of people using AI and robots for terrible, or even questionable, purposes that troubles our dreams and lights up our screens. It’s AI and robots having purposes of their own.

We’ve been doing this for a long time. The founding text of SF is Mary Shelley’s Frankenstein, which though not about robots set the template for thinking about them. The very word ‘robot’ in SF and popular usage comes from Karl Capek’s RUR, a play about a robot revolt. Robots rebelling against their creators was already a cliché by the 1940s, when Isaac Asimov tired of it and tried to end it. But even Asimov’s robots, constrained as they are by the Three Laws, are self-aware and contrive to have their own way.

I’m deeply sceptical about self-aware AI and rebel robots as real near-future possibilities. Yet I continue to write about them, and I’m eager to read and watch more stories about them.

Why? Two reasons.

First, because these stories are myths. They enable us to talk and dream about ourselves and the sort of things we can do. What are the characteristic features of AI in SF? Self-awareness and self-activity. What makes robots in SF interesting? The ability to work, to speak, to co-operate and to resist. Does this remind you of anything? Look in the mirror.

What all these characteristics have in common is that they’re human. By projecting them onto machines and isolating them, we explore our own alienation in ways it’s hard to do otherwise. Because our creations are already, and always, out of our control, we have plenty of anxieties to work through about that, and always in need of mythic ways to process them.

The second reason we should go on imagining implausible machines is that they’re only improbable, not impossible. It’s theoretically possible that a brain could be mapped in detail down to the molecule. It’s not at all ruled out that self-awareness could arise, or be created, in software. It could even happen by mistake.

In this century? I doubt it. In this millennium? Almost certainly. But whenever it happens, and whatever the consequences, it can’t hurt to have run the thought experiments in advance.

We can look back with pride at the great clanking horde of SF robots. Long may they march!

Dissidence is out now from Orbit