It’s funny how some of Elongated Muskrat’s testing and experiments involve the subjects dying.

Monkeys dying with the Neuralink experiments, and humans are dying with these autopilot tests!

Random question I’ve always wondered about in case anyone is more familiar with these systems than me. My understanding is that autopilot relies on optical sensors exclusively. And image recognition tends to rely on getting loads of data to recognize particular objects. But what if there’s an object not in the training data, like a boulder in a weird shape? Can autopilot tell anything is there at all?