It’s the greatest marketing slogan of all time: “Artificial Intelligence will kill us all”. They tell you the machines will grow past today’s weak Siris and Alexas, into strong, HAL-9000 killers. At least, that’s what the talking points say. You tend to get attention when the product roadmap includes ‘Armageddon’ as a feature.

“Don’t worry,” they continue, “we’re making great progress and we’ll get there in 20 or 30 years. Until then, these bots and things are completely harmless! Look how cute!” But these bots (by definition ‘weak’ AIs) actually have the potential to sink us long before Skynet incinerates us. Ask yourself:

If artificial intelligence is embedded into my environment to help me find what I want and need based on my behavior, and I behave according to its suggestions, at what point does my own intelligence cease to be a part of the equation?

This is, simply put, a feedback loop. This is behavior, auto-corrected.

To build a suggestion, you need data. But there is already plenty of data collected now, and more on the way. Everything is connected: my car, my home, my work, my medical records, all my means of communication. Plus my thermostat, my toaster, and probably my toilet. Telemetry is engineered into the design of everything today, so I can’t just cut the purple wire, or put a piece of electrical tape over the offending bit.

Since it’s not the government doing this, then its totally OK, right?

The Artificial Intelligence Problem

I digress…if everything is wired and instrumented and processed by algorithm, and the system regularizes and metes out my options by mathematical regression, how am I to find New things? Where does New enter the equation? How does New get exposure (and how much would it cost to get a little more?) Do I trust they add some spice to the equation? Or is the output bought and sold on closed exchanges like display advertising? And at what point is that exchange turned entirely over to machine control?

You can see a glimmer of this now. Twitter suggests hashtags, based on traffic, which you can buy. It’s always trying to auto-complete what I type based on trend. You can even buy the promotion from Twitter itself. Now, tens of millions of Twtter accounts are already bots. They exist as collections of rules and scripts, most intended to influence some sort of financial gain but all helping to drive trends.

I’m not saying this is intentional manipulation of behavior on a wide scale — yet. But the mechanism intentionally monetizes the behavior, and so it is practically the same thing.

Longer term, if the system pushes everyone ‘like’ me in the same direction, how long before everyone is ‘like’ me? What does ‘like me’ look like? Can I change it? How do I know what I want to change it to? Can it tell me? Will I notice? What will it cost? Will there be 5 billion unique ‘me’s, or an Apple-esque selection of 10 ‘great’ People Personality Models? Will they call them iMe? iMyself or maybe just iI?

It may seem crazy, but you see what I mean.

In the End

At some point, maybe we invent a microscopic artificial intelligence in the form of grey goo which promptly eats us, then figures out how to circumvent the laws of thermodynamics and goes on to destroy the Known Universe. Yay us.

And one can make the argument that society already shapes our preferences through mass indoctrination and the enforcement of cultural norms, although this is on the timescale of generations. I’m talking about the timescale of a trending topic.

This loop is already subtly steering our behavior. In the end, the question is, how long before it’s entirely automatic? A time when don’t wake up in the Matrix…we’ll all just wake up in a Gap commercial.

I wrote this off the cuff as an editorial on artificial intelligence, I did not have time to include a mega-extensive literary analysis so I could continuously drop names like Minsky, Dennett, Fermi, etc. and it would have been TL;DR anyway. If you’re into those guys I’m sure something will auto-suggest them for you.