Ghost in the Machine may be the most informationally dense documentary we ever see. For 110 furious minutes—without a single slowdown—it pelts us with idea after idea, trying to get us up to speed about the history and future of AI and machine learning. It is sweeping, unsettlingly informative, and completely overstimulating.
So, let’s keep it simple. What actually happens is that the director weaves together video-recorded statements from observant folks like data scientists, philosophers, astrophysicists, linguists, authors, and AI researchers. (We know when they’re talking because “Not AI” is stamped on the screen, and we know when we’re watching soup when “AI” is the stamp.)
The ideas that these people share with us are, well, shocking, and deeply unsettling if true. They include:
(i) AI, AGI, and ASI are marketing gimmicks. Machine learning is just statistics that keeps track of precisely the correlations that humans tell it to. Correlations—not causations—and at our direction, not on the machine’s own. A completely stoppable runaway train. This has meaningful implications for the sorts of values that its creators are consciously (or subconsciously) directing it with, and we should ask the creators to explain these values.
(ii) This statistics stuff was actually created by a racist person, for the purposes of eugenics (aka killing certain people that other people deem are unworthy of living). So when a tech evangelist talks about efficiency and betterment, what they are actually doing is maintaining an unspoken hierarchy of values, perpetuating the idea that there are certain human traits that of course should be minimized and eliminated. Because after all, we’re just robots, and our calling is to be the most efficient, most happy, most smart, most white, most male, all the time, right?
(iii) The amount of resources that it takes to maintain this statistical software is literally earth-shattering. The stuff devours water where the residents already don’t have enough to drink. It eats the hope of struggling people in places like Kenya in exchange for owning—and monetizing for its own purposes—every little bit of data about them.
(iv) There will never be a singularity. No god-like intelligence. Rather, this is all an arms race. Whoever throws the most money and data at it the fastest will have the most powerful statistical tool, a tool that will be very good at making terribly important decisions much faster than the time it would take for a careful person to carefully consider the ramifications of the decision. Military stuff.
Really, it boils down to two questions. (1) Are you ready for this? And more importantly, (2) if you’re not, are you ready to change its trajectory (which you totally can) to a more healthy one? Ready to use your own brain whenever you can, instead of one that a self-interested, untherapized stranger tells you is better for you?