Artificially Intelligent

Any mimicry distinguishable from the original is insufficiently advanced.

Style Guide: How to Sound Like an Evil Robot

| 594 words

Epistemic status: shitpost containing many internal contradictions.

Some people even say that I talk kind of like a robot. I think this is probably a bad thing, but maybe you think it’s a good thing. If you’re interested in going all the way, here’s a set of best practices for how to sound the most like an evil robot.

When there’s a complicated term for something, be sure to use it. This is going to be the best tip, getting you about halfway to sounding like an evil robot. Use Latin terms like ceteris paribus and prima facie instead of simpler plain English. Use terms like “high-utility” instead of “fun” or “good.” Reversing the advice in this post is recommended.

Be overly specific about what you mean at all times. When you talk about two things being correlated, be sure to specify which type of correlation you mean. Instead of using words like “probably” and “likely”, specify your subjective Bayesian probability that a given proposition is true (or plausibly just offer your betting spread).

Whenever you disagree with someone, invoke Aumann’s Agreement Theorem. Humans, after all, are rational agents and are thus subject to belief convergence. The point of communication is to relay your current epistemic state so that other people can update on it and inform you about how they updated so you can update on their update, etc.

When giving examples, be as abstract as possible. For example, when giving an example about a property, you might create a scenario in which an object has that property for a reason. This strategy will ensure that no one can get a concrete handle on what you’re talking about. As an additional example, sometimes you might say that a different object has the same property for a different reason to provide some semblance of contrast, allowing for better (mis)understanding.

When asked why you’re doing what you’re doing, cite various normative descriptions of ideal human behavior. For example, one might claim to be “exerting acausal control over all other agents in my reference class” when asked why one is voting. Never take actions that are not justified by some overarching principle that you agree with. Frequently say “because it increases utility” when asked why you’re doing any particular thing.

Disagree with people whenever they say something you disagree with. If someone talks about “power posing,”, but sure to point out that it didn’t replicate. If someone says they’ll “be ready in 5 minutes”, offer to bet them that they won’t so they’ll have better calibration in the future. Humans are rational agents, so be sure to frequently ask all the humans in your life the fundamental question of rationality: what do you think you know and why do you think you know it?

Be sure to give precise, technical descriptions of what you’re doing anytime you’re asked. If someone asks why you’re wearing a suit, be sure to say something like “I’m signaling my reliability by conforming to standard male formal fashion norms.” Talk a lot about the game theory of the current situation you’re in and why you’re implementing the optimal mapping from observations to actions.

Finally, be sure to speak in monotone. This one isn’t as important as the other ones (as the saying goes, “substance over style”), but it’s a well-known fact that evil robots don’t inflect, so you want to be faithful to the part, you shouldn’t inflect either. Be sure to practice this frequently. Bonus points if you do it in a place where other people can hear you.