Ever since I first learned about neural nets, I've been fascinated by how they relate to the inner workings of our brains. For example, during Tesla's Autonomy Day, they explained how they rely on visual information from cameras around the car, rather than Lidar, just like how humans use their eyes. They showed how their model accurately predicted a curving road behind a hill without actually seeing what's behind the hill. It's the edge cases in traffic that make it difficult to achieve autonomous driving, which is why these models require massive amounts of training data covering various situations.

As a thought exercise, I sometimes like to think about the human brain as an AI model: the input to our brain is what we see around us, stories other people tell us, breaking up with our partner, etc. The output is the actions we take and what we say.

We tend to produce similar outputs if we don't read, actively listen to people, change our behaviour after a breakup, meet new people, travel, etc. With each of those things, our output can completely change.

That's what I find magical about the human brain: it only takes a single input to our training data to completely change the way we think.