Something that comes naturally to humans is still a challenge for computers: recognizing images. We’re able to learn what something or someone looks like. We can look at an image or sculpture of a subject and recognize what it is. Even if the image isn’t perfect, we can figure it out. Computers aren’t as good at that. That’s part of the reason classic CAPTCHA tests work — they’re easy for us to recognize but computers find it tricky.

But computers are getting smarter. Even the inventors of CAPTCHA tests don’t see this as a bad thing — when a computer beats a CAPTCHA test, it means we’re one step closer to artificial intelligence. A failure of one system is a triumph for another.

Several teams around the world are working on improving computers’ abilities to identify images. With reliable image recognition software, we’d have an improved search capability (most image searches rely on meta data or text captions, not the actual images). You could catalog billions of images automatically with a strong enough system.