The Artifice of Intelligence

A fine Scots blizzard of April sleet and hale made for an apt setting to take a sick day. What I did get up to, however, was listening in to an episode of the BBC’s Horizon on artificial intelligence. Can’t say I recommend it, as the necessary thrust of the show was a tech-dizzy sense of inevitability as progress surely marches ever on. While IBM’s Watson, and a parade of variously impressive robots, were given their praise, not once was the presence in the background of this little thing we use millions of times each day. Google, you might have heard of it. (And rivals.) The internet, I’m not the first to reckon, is a closer step along the path to whatever AI we might ultimately discover. And the fact that we can index and search it, this endlessly complex and forever changing thing, to serve up half decent results in an instant is an everyday feat I find naturally more convincing than acing chess or Jeopardy.

I’m being a touch unfair, of course. AI is a an age old problem, as is its coverage. A much better documentary, as I remember from digging it up a good while ago, was The Machine that Changed the World. It may be twenty years old, but its impressions of the imminence of our artificial peer’s arrival were pitch perfect for today’s. Telling. And they even bothered to cover the internet, in charming period dial-up modem and virtual reality treadmill style. Is there anything quicker ageing than technology?

But all this is to say that I’ve the problem to address for myself, writing of a future so often supposed to contain Turing’s answer. Rummaging in my 2010 notes again, I see that I’ve considered this before.

(Put on Yarus by John Berberian if you like. Yes, the archives have a soundtrack. Why would you think they didn’t? Oh, right. Well, this one is a mood piece. See if there’s sense in it.)

(And beware distant spoilers. But they are so very distant indeed. Aegean? What in the name of epistemology has that to do with this? Don’t expect complete answers. Work. In. Progress.)

Andala Tech: Computers

The Extent of AI

Let’s just lay it down here: there is no hard AI in Andala. I wouldn’t know what to do with it. Real, old-school, artificial intelligence is the stuff whole stories are made out of. And I’m quite sure to their eventual detriment. It’s not to say that my avoidance of what is, thus far, quite beyond us is in any final sense better. You can well shake your head at the internet’s absence in so many an old future’s tale. Foretelling the future’s a mug’s game, it really is. But for those of us who insist, we can at least remain consistent to our own, doomed, rules.

Within the narrative: why not?

Philosophical questions abound wherever “AI” is spoken. As good a warning sound that bull lies ahead as we all often know. It’s only natural, I suppose. A mind with a power switch? An alternate, potentially infinite, realm of existence? All the ancient unknowns the humblest of computers used to garner, how quaint, are turned new again. We really don’t know how this story goes. And so the fiction.

But that’s not especially useful to me. Not for this particular novel and universe. Hence my determined scowl to side-step the whole ado.

But I could still use an answer for why not, from my people’s perspective.

I’ve heard word of great equations with hard AI and various quantum or statistical limits. Oh really? Not that I’ve gone back to check their reasoning, but it sounds perfect hogwash. Intelligence, for as much as we typically find it awkward to analyse and place within our systems, is here already. There’s nothing unique about the conditions within the human brain. Practically speaking, we may not have gotten anywhere close yet, but there’s no magical barrier holding back our eventual progress. I don’t doubt we will experiment in very synthetic intelligence, discovering as we do quite how common or arcane our own natural kind turns out to be once you’re running the process. Emergent phenomena have the hint of transcendence to them, effervescing somewhere above our day to day intuition and handiwork. But we will know them. And they very well may change us. Boy, that would be a second computer revolution to witness, or to miss.

Yet I’ve got to put it off.

Various different realms of excuse occur to me. Computational density is a good one. A wholly practical “no can do” on elaborate brain emulation thanks to incomplete technology. It’s my safest bet. Though obviously at odds with my eventual plans at Mintaka.

A rather more involved alternative is the prospect of a ban. Perhaps it was achieved in the meantime, and prove so wildly disruptive in whatever place that it did, that the whole thing was sealed in embargo. Has a ring of conspiracy about it, typically something I don’t like but which fits with the Aegean much better than my simplest choice. The real promise to this one (if I can find it in myself to accept any such thing as Pandora’s Box resealed) is its partiality. Most may not have AI, but some can, so long as they know to keep entirely quiet. Yes, this is the one a writer’s instinct would snap at, I know. I’m thinking synthesis.

Truth be out: I suppose I’ve already decided all of this in the past, now that I consider the first two options to come to mind. The events at the Aegean are about something essentially equal to the hardest core of all AI. Universal simulation: the superset of every conceivable mind. So here at least AI is a foregone conclusion. You know, if they needed it, in the explicit HAL sense of old.

I’ll write another section about simulation and the practical limits on computation in the Andala universe, as I consider it a separate subject, hence my dawning surprise. But as for computers which think they are human, or operate with apparent consciousness and free will, I’ve already described why I’m against it. The in-world reasoning may be identical, but I should also remember mind playback, come to think about it. Thoughts in progress, I tell you, even if they are for the most part just memories.

—September 2010

By and large, I think I was right. (What a surprise!) As much as I respect 2001: A Space Odyssey, copying it is not a goal, and that includes HAL. Besides, between now than then, I read Stephen Wolfram’s take on emergence, and adopted some of what I thought of it for the physics of my story.

Quite besides Andala, I’m more convinced now that Turing Machines the like we have today throughout our technology, are a poor match for the hard intelligence problem. What we have managed to construct, thus far, is something of an endlessly elaborate Mechanical Turk. Software is still a painstaking craft, every bit as specialised and handmade as the realm of industry before mechanisation. We’re the ones deciding at every step of the way, whether we use these machines directly as tools, or beforehand when writing the code that enables this. There’s something else to intelligence. Nothing holy or sublime, but something self realising, self improving, a self in the first place if you will. This is why I fear for algorithms alone to ever get there. Rule based we are not. What we are is about as emergent as it gets.

Just trying to figure out how to explain my view is a little taste of the quagmire I’d find throughout my book if I let it loose in there. And yet…

Eventually, just like the hosts of these breathless shows, I reckon AI is ultimately around the corner. The prospect is so disruptive that I can’t have it any sooner than the end of my story, should it ever get there. And that’s what I was on about with those vague allusions to “Mintaka” and “Aegean”. All that stuff about mind uploading has stuck with me longer than I imagined. There’s something I can’t quite comprehend lying in wait for Andala. That may very well be every bit as nutty a path as it sounds. But I wouldn’t be too sure. Writing is a curious, shrouded, inscrutable thing. Just like intelligence.