This is what Artificial Intelligence looks like, according to AI from IBM Research, the folks behind Watson. The above image was created by AI after researchers tasked it with coming up with an image of itself to accompany a New York Times series about the technology.
"For this experiment," a spokesman told me, "researchers gave AI a mission to complete in three weeks: identify a core visual concept in AI, create an original image that captures the AI concept, and present it in a way that fits with NYT's visual style."
IBM's algorithms approximated the human creative process to develop the image.
The first step was conceptual. The AI read thousands of NYT articles to derive representative concepts. The winning concept was the notion of a robot shaking hands with a human.
Using a generative adversarial network (GAN), which enlists two neural networks competing in a zero sum game to generate images, the AI created its own images based on a large teaching set of 1000 photos and images.
The AI then perused past issues of the NYT to glean an appropriate style for the image.
The intersection of AI and creativity is of growing interest as our human hegemony on creative expression seems to dwindle in the age of automation. But the process outlined above, in which machine creativity essentially derives from human creativity, suggests more of a symbiotic relationship.
As everything from advertising and filmmaking to writing become more automated, any reminder that humans are still needed at some stage to fulfill the creative process is welcome.