Numerous manual jobs are slowly being replaced by automation, but in the world of art, the place of artificial intelligence has merely been that of an assistant to the creativity of humans. Improved and new AI-based instruments are invigorating artists by giving them a platform to visualize each idea that serves to inspire them by permitting them to draw images, erase them, and then redraw. This effectively gets rids of the process of tossing discarded ideas into trash cans until they overflow with them.
Google’s AI experiment grants artists the ability to draw something, Neural networks which were trained on millions of doodles that were collected via a game have the sketch passed through them. The experiment comes up with a multitude of possible ways that an object can be continued on the basis of where the artist left off. The model is also capable of producing similar doodles and mimicking drawings.
At Brown University, there are researchers attempting to achieve perfection. Although GANs, or generative adversarial networks, have been stunning the world with their art pieces that range into the surrealistic, AI has a long way to go from truly replicating a human’s writing or drawing skills. Because human hand motion is complex and factors like each stroke’s speed and spacing, robot arms have a difficult time mimicking and keeping up.
To perform such functions efficiently, an undergraduate student of Brown, Atsunobu Kotani, and Stefani Tellex have created two distinct models. One local to handle each stroke and one global to handle the shifting action. The team of researchers demonstrated a robot capable of writing the word “hello” in ten different languages that employ differing character sets. The robot also could reproduce such rough sketches as a representation of the Mona Lisa.
The method by which the robot was taught enables it to learn in real time with greater ease. Upon observing the image that is the target, the robotic arm at once makes its attempt to draw or write. Researchers segregated the writing into a pair of steps: first, a drawing action that draws every stroke, and second, a shifting action that achieves a new point at which to start. And progress in AI continues to impress onlookers.