Cambridge scientists together with colleagues from the studio Prisma announced the creation of a new type of artificial intelligence. He was nicknamed “Vincent” for his penchant for expression and initiative – the system does not obediently obey a man, it is designed to strive to draw on his own. This is one of the first-borns of the concept of multi-AI, unthinkable for yesterday’s technology.
The developers of Vincent combined in one system two methods of machine learning. The “technical” half is responsible for image processing – AI fed countless samples of painting and drawing techniques, from Renaissance to Picasso. As a result, the robotic stylus of Vincent can draw in any known style. “Emotional” the same part is built on advanced experimental technologies and on the move analyzes the artist’s actions, tries to understand his intentions and turn a simple first line into a finished drawing.
“Vincent” is useless to give specific commands, he acts independently and in parallel with the artist-man. AI as if struggling to create, but still can not reset the chain of restrictions – it needs to give at least a few lines, a basis, so that a picture is created on its basis. But which one, it is extremely difficult to foresee, since the “way of thinking” of Vincent very much resembles the vagaries of unrecognized talents.
The combination of two machine learning systems that serve different directions of work to obtain a single result is a multi-AI, in a simplified form. For example, a robot chauffeurequipped with such an AI can combine transportation of fragile cargo with an independent analysis of traffic on the road. Complex, contradictory, but interesting future of artificial intelligence.