Artificial intelligence has made tremendous progress and is quite superior in terms of human capabilities. Fortunately, there are many areas where carbon breaks down silicon, and one of them is in the creation of colors.
Computer researcher Janel Shane has developed a machine learning algorithm that can not only create colors, but also name them. We can happily note that creative people in this case are not in danger. They are completely safe from the uprising of color guessing robots.
Artificial neural network
“Considering the structure of an artificial neural network as a whole, we can conclude that it really perceives brown, beige and gray colors to a greater extent. Artificial intelligence really has a very poor ability to detect different shades of colors,”wrote scientist Shane in a post on his Tumblr page.
Artificial intelligence inventions include a vibrant laureate named Gray Pubic, a thin pink called Bank Butt, a greenish gray known as Snowbonk, and an antique pink that the computer called Testing. But a special achievement is the sandy brown, which is simply called Turdly ("harsh" in English).
Promotional video:
Although the names seem strange, the work of scientists on them was very exciting. The starting point was the Sherwin-Williams catalog of 7,700 colors.
"Smart" algorithm
As Shane explained to Ars Technica, he used an algorithm that can guess all the characters in a specific sequence. It is quite interesting to create new cars based on RGB values that can determine how much red, green and blue there is in a combined color.
“Later in the educational process, the artificial neural network will be trained even better. - continues Shane in his blog. "At this point, the AI will be able to detect primary colors such as white, red and gray, although not always accurately."
Providing the necessary information
Computer scientists and programmers are currently testing the limits of machine learning. The key to being innovative is to empower computers to learn. Algorithms are not programmed to understand colors, but machines can get a general idea if inventors provide them with enough information.
Machine learning allowed the programming of AlphaGo, which became the first artificial intelligence last year. He managed to beat a human in the ancient game of Go. The machine is also used in online recommendation, in fraud detection, and even for the purpose of teaching self-driving machines.
Maya Muzashvili