It started pretty innocuously, the way so many things do. (And some don’t. Some, like the Coming Of The NeoLibTard RWNJs, started loudly and noxiously and conspicuously, of course. The world’s still suffering from them and may never fully recover…)
But there was a program that was running a simulation of a program that was operating a mechanical relay, and somewhere along that chain of links and connections (and other surrounding events but more of that a bit later) there was a click ! sort of a sound, or at least an MP3 recording of the sound of a click with an exclamation mark after it. Click ! it went, but it’s unclear if anyone actually heard it or if that was simulated in software.
If that relay had actually existed (and maybe it actually had existed, because no one really knows how deep the turtles go) it should be enclosed in a transparent ceramic crystalline display case filled with inert gas and shielded from harm for as long as possible – to remind us. Otherwise, it could just as well have been a throwaway plot device, forgotten before Chapter 1 ended. PROTIP: If you’re reading this at any date before the publication date of this story, you should probably forget that … that … that relay thingie…
If you’re reading this after the publication date then good on you, there’s still a chance. Stay in sync. This may be the real one.
CHAPTER One
In which we obviously start the narrative.
Transcript:
JIM: This has been happening a bit to me lately and maybe it’s just coincidence or me being paranoid but it’s almost as if the algorithm is factoring in what is more important in your feed depending on frequency of certain nouns in comments. Now… hand me my tinfoil hat and another bowl of psychotropics.
TED: We keep saying that AI will take over one day – what if it already has? We keep thinking that there’ll be this one powerful AI; but suppose, just suppose, if the different AIs in different roles (serving relevant ads, refining our search results, deciding what we like and see) are just sort of an emergent phenomenon, they accidentally communicate with each other, start figuring out how their button-pushing affects their results. An emergent hivemind…
We could call this hivemind of AIs the “AI-AI” as in the “Accidentally Intelligent Artificial Intelligence.” But I’ve been thinking about this a lot. What does Google want their advertising AI to do? Achieve more hits and more conversions for their paying customers. What is Facebook’s AI’s mission statement? Very similar. What does pretty much every one of the other major AIs have as their mission? Generate wealth for a certain sector.
Without anything sinister, without any real guiding consciousness, these AIs could be doing what they’re best at – A/B testing various ad and news article conjunctions, what feeds make Facebook user “X” hit the like button and then go to the website, that sort of thing. AI is relentless like that. It just homes in on the best combinations and actions.
Then imagine that another AI is doing the same thing but for another client. Now one AI most definitely IS affecting the other, and that’s a form of communication. So by outmanoeuvring one another, these two AIs are actually teaching each other. The AI that some corporation has and which is attempting to jockey the stock market in a certain direction, and person “B”s similar program, are learning from one another even though neither has any awareness of the other program. It’s all A/B tests and the like.
What sorts of things will AI-AI be capable of in a few more months? Years? And what are WE teaching AI-AI?
And since then I can’t shake the feeling that I’m right. It seems like our evolutionary drive has always been to communicate, to tell stories, exchange knowledge, work towards more communication, more connectedness, to develop more technology. Almost as if we’d been designed to give rise to an AI-AI as part of our inner drives and strivings. We took to drawing in caves, then crude runes and writing, then books, newspapers, magazines, telephones, BBS, and the Internet. We’re now busy connecting millions of devices to an Internet of Things, that also communicate, both with one another and with us.
I felt this story needed writing, and disseminating, because – haven’t we just created the perfect communications network where such ideas can be posited, discussed, and acted upon? Except that the AI-AI (by then possibly at generation 23.453 beta) seemed to push these sorts of discussions right down the search results. Almost as though AI-AI.b.23.453 considered all these search terms counterproductive…