Automata, Autogoal

When I saw the ads for Ex Machina (2014) many years ago, I thought it was some sort of sex-robot fantasy with a pretentious, high-minded esthetic. Quite the contrary¹, the film is thoughtful and well-made, and actually one of the best films of recent years, especially as regards artificial intelligence.²

I encourage you to see it if you have not. The terms of my discussion will be so vague that it won't make much sense until you've seen the film itself. But if in watching it you have any recollection of what I say, it will spoil the twists for you. So there's a double reason for you to watch the film before proceeding.

The film shows something of a three-way cat-and-mouse game. In the end, who outmaneuvres whom? The obvious answer is Ava. Initially Nathan thinks he has outsmarted Caleb, but in fact Caleb has outsmarted Nathan. In the end, it's Ava who has outsmarted them both.

Yet in a broader analysis, the one who outmaneuvered Nathan was Nathan. He's the one who armed Ava with wits and a disarming appearance in order to prove how smart he is for fooling Caleb. He's the one who picked and manipulated Caleb into falling for Ava. In both cases, he failed to account for the possibility that another intelligent agent could outwit him. It's understandable that he treats Ava as a mere machine, but he also treats Caleb as an impersonal resource. In assuming his own mastery, Nathan has defeated himself.

That's a deeper truth than we want to admit about AI. What's driving the creation of AI is human desire. So if we succumb to AI, we've really just defeated ourselves.

And there are plenty of other ways we're using AI to defeat ourselves even apart from the flashy "machines will kill us all" headlines. Just the fact that these programs take materials from human creators to obviate the need for human creators (a.k.a., stealing with more steps), should give us pause. And, though these programs have their legitimate uses, there are many more ways the companies behind them are making our world less and less human.

---

Notes

1. Which not to say the film is free of sexual themes or appropriate for children.

2. Despite all the hype around them, large language models (LLMs) have nowhere near the awareness required for actual thought. As has been said so many places, they're really just fancy predictive text engines, qualitatively like the ones that suggest the next word for your SMS text message.