Snowflakes and conifers, elves and presents... see a pattern emerging? Us humans are pretty good at finding the message in the noise.
Recognizing, sorting, and evaluating inputs, comparing them with mental models of the world, responding, learning and refining knowledge... we do it constantly and unconsciously.
Our pattern recognition skills have built-in biases from millions of years of evolution. A favorite internet of mine is faces in places. Our brains are primed to spot eyes and evaluate a mood. We can't help it.
Or take false positives—sounds bad, right? But, I'd rather have my ancestors run away from a sabertooth tiger that wasn't there, rather than being 100% correct on their tiger-spotting skills. Our pattern-recognition skills are optimized for—biased towards—survival.
Oh, and the pattern I was thinking of with the example at the beginning was "plural nouns." Did you get it?
Now, whether or not you are on the machine learning / artificial intelligence hype-train, looks like a new sort of pattern-recognition system is emerging. One where we’re the ones in charge of setting the end goals.
So, no biases, right? Guess again! According to Google's InceptionV3 image classifier, this little 3D-printed fella is almost certainly a rifle.
Researchers were also able to fool the algo into recognizing a cute kitty as guacamole.
Adversarial pattern recognition attacks against artificial intelligences. Now don't tell me we're not living in the future.