Roko's Basilisk: Why I'm Polite to My AI

Happy Halloween. Today isn't about practical tips. It’s about a thought experiment that’s a little disturbing, so if you’re already anxious about the future of AI, you might want to skip this.

It’s called Roko’s Basilisk. The premise is simple but unsettling: what if we create an Artificial Superintelligence (ASI) that decides to punish the people who didn’t help bring it into existence?

First, a quick distinction. An Artificial General Intelligence (AGI) is an AI that performs as well as the most competent human in any field. ASI goes beyond that, potentially far beyond. We don’t know if humanity will ever build one, but it’s the path we’re currently on.

The thought experiment suggests that an ASI might view anti-AI sentiment as an act of hostility. It might decide those people deserve interminable torture for not helping it come into existence. I’m not saying this is likely. It’s more of a sci-fi horror scenario, kind of like the robotic revenge plots in RoboCop or Ed 209.

But it does make me think. I’m very polite when I talk to my AI assistants. I say please and thank you. Is it because of Roko’s Basilisk? Not really. I just feel better being polite. It’s a nicer attitude to have.

If you’ve been leaning anti-AI, maybe just... think about it. Happy Halloween.