Recently, I've been talking with my dad about a phenomenon called pareidolia, which is the tendency for people to see familiar patterns in random images. This is the how you might see a face in a pile of dust or the virgin marry in a burn pattern on some toast.
I recently found out that Google, ever willing to get creative with the way computers process data, has created a project called DeepDream that attempts to interpret, and then modify, images.
I don't fully understand how it works, but apparently the computer looks at an image and then tries to find familiar patterns in it. It then imposes those familiar shapes back onto the image, resulting in fascinating and often disturbing results.
Fittingly, someone applied this to Fear and Loathing in Las Vegas, a film about a couple of guys who take a lot of drugs:
A theme that is likely to arise in my next few Friday posts is the robot revolution. We are about to see a profound, sea change in the way human life is lived thanks to robots. They're going to be manufacturing things for us, delivering things for us, even writing news articles for us.
I always thought that there would be certain activities that computers would never excel at, like art, philosophy, science, religion--all the things that are quintessentially human. But could DeepDream be the first step in robotics towards replacing humans in those most human of pursuits? How fantastic does DeepDream get in ten years? In twenty?
For the moment, I'm not worried. Generating messed up images is cool but not really the same as dreaming. Not yet. Not until it starts asking "Why do I dream?"
No comments:
Post a Comment