Gigaom brings you our unique analysis and commentary on the present and future of AI.
In Japan, a man named Makoto Koike started helping out at his parents' cucumber farm. He saw that his mom spent eight hours a day sorting cucumbers by size, shape, color, and number of thorns. It turns out that long, straight, thorny cucumbers bring the largest prices, but, of course, longness, straightness and thorniness are on a continuum, and are not binary characteristics. Faced with this challenge, Makoto did what anybody would do, he built an artificial intelligence based sorting robot.
He used TensorFlow and deep learning to train the AI with photographs of cucumbers. Then a Raspberry Pi is used to ascertain if something even is a cucumber. If so, the image is sent to a beefier Linux box for classification. Then he used an Arduino device to do the actual sorting.
The moral of the story is that if AI can be used on such a specialized and narrow task as sorting cucumbers, then the number of use cases really is enormous. Looking at the world with that idea in mind, you might just find yourself surrounded by AI use cases.