Artificial Artificial Intelligence

Machine intelligences should be with us by now, shouldn’t they?

Long promised by the golden age of sci-fi, aren’t we now overdue our robot brains? Those intelligences we can interact with on a human level, bound into subservience by Asimov’s three laws, there to handle all the menial tasks and mathematical heavy lifting.

They were due to usher in a more utopian existence. Robot gardeners, robot policemen, robot friends, stylists, celebrities and sportspeople – where are you all? Why, two decades into the 21st Century, am I still making my own breakfast, choosing my own clothing and lighting my own cigarettes? Where is my robot manservant who might pleasure me in all the ways only he knows how? Why am I still stuck conversing only with dumb humans?

Ok, I’m conversing mainly with humans. I’ve taken one of those automated telemarketing calls. I’ve been accidentally lured into an argument with a Twitter bot. I’ve been momentarily convinced by a spam email that addressed me by name. And I’ve been mightily freaked out by a targeted ad that seemed to know exactly what I was thinking of buying. Yes, I do talk with robots. More often than I’d like perhaps.

We might not have AIs of any significance yet, but we do have a form of Artificial Artificial Intelligence (AAI) that is passing as AI in some quarters. Bots, algo-curators, recommendation engines, humanised info feeds, and that horrible breed of algorithmic journalism you see, where it’s clear no human eyes or fingers have been involved between press release and published article. Artificial forms of artificial intelligence, one and all. Machines not thinking in any human sense of the word, but attempting to give the impression that they can. Kinda.

AAIs are not cognitive systems, there is nothing that might be mistaken as real thought behind any of them. They are, at best, an algorithmic intelligence. More accurately they are big data extrapolators. They take data we have generated, and make statistical guesses on what we’re likely to do/say/buy next. And we, with our darling pareidolic persuasion, give them the benefit of the doubt. We see them as reasonably, mildly, intelligent.

AAIs deliver all the promise of AIs – machine intelligences we can converse with, good at menial mathematical tasks – but we stay safe in the knowledge that they’re pretty limited without a human hand to guide them. There’s little danger of a SkyNet-style consciousness emerging from these data sets. We have no fear of a product recommendation engine that dreams of being human.

AI, true AI that is, is immensely difficult. In the middle of the last century it was a problem that felt solvable, and solvable in the near future. Five or six decades later it seems further away. Much of the mechanics of human intelligence remain beyond our understanding, so simulating it digitally may now seem further beyond our capabilities that it ever did.

The optimistic AI of 50s and 60s sci-fi evolved into the dystopian scenarios of pre-millennial sci-fi, when the invention of AI could only be imagined as the product of an out-of-control technological process; a singularity. These AIs would come about via exponential machine learning, not human design. On this path AI would eclipse us very quickly, leaving only the question of how we might survive our redundancy.

These kind of horror stories put AI out of fashion for a long while, which may have subtly limited progress in the field. In the same way as media stories exaggerating crime statistics serve to keep old folks in their homes and in front of their TVs. We are taught not to venture into unknown places. Because it’s not safe for us there.

Maybe AAI is the only AI we can ever really accept. Maybe this is as far as we’re ever going to go in this direction. We’ve achieved the cuddly face of AI. Why toy with the alien, uncanny valley, intelligence of a machine, when we already have a pseudo-intelligence we immediately recognise.

We like this form of AI because it is familiar. Because it is us. We have become a culture of algo-Narcissuses (Narcissi?), entranced by a mirage of intelligence that is not really intelligent. It is merely our own reflection.

-- 30th August 2016 --

« »
Generative Art book cover Generative Art: A Practical Guide
(Manning 2011)

The definitive guide to the philosophy and practice of coding as an artistic medium.

Foreword by Marius Watz.
Novelty Waves book cover Novelty Waves: A Short Book About Digital Art
(Leanpub 2014)

Five long articles and two short stories exploring digital art.

Foreword by Jeff Noon.

Name your own price.