It is well established that AI can be weird. If you haven't dipped into the sublime world of bizarre and serendipitous NPC behaviour, this Wired article is an excellent place to start. I just re-watched a couple of them, and am still self-consciously suppressing the kind of raucous laughter that makes your fellow office workers start looking at you strangely. Or at least intensifies those looks, anyway.
Forget about NPCs, though, because right now I want to point you at the work of Janelle Shane, a scientist who likes telling neural networks to rejig concepts that humans have generally tackled much more capably. The results are often funnier than you or I will ever be.
Read the rest of this article by joining the Rock Paper Shotgun supporter program
Sign up today and get access to more articles like these, an ad-free reading experience, free gifts, and help us create more great writing about PC games.See more information