All the signs show that artificial intelligence (AI) is set to revolutionize our lives more than any other previous digital development. In fact, more than a few people even see the technology as a danger, fearing it could break away from human control and turn against its creators. Fortunately, we aren't at that stage just yet. At the moment, we mostly encounter AI in the form of virtual assistants such as Alexa and Siri, which do not appear to pose any threat. In addition to the classic CPU and GPU, the first mobile devices now also come with an NPU (Neural Network Processing Unit) that is capable of completing tasks up to twenty-five times faster than previous models. But one thing that most people do not believe AI can deliver is artistic creativity. However, even this seemingly uniquely human trait has been cast into doubt by the latest work by Ahmed Elgammal, a computer scientist at Rutgers University in New Jersey in the United States.
As part of his AI research, Elgammal focuses on innovative technologies designed to produce machine-generated art that is as new and original as possible. The current highlight of this work is his AICAN project, an AI algorithm that can be "trained" - i.e. fed huge volumes of relevant data - to create its own images. The exciting thing about these images is that test groups made up of experts and the general public alike have often been unable to distinguish the AICAN "artwork" from pieces by human artists. What's more, they liked the AI images - or at least they did at first. "Interestingly enough, some people were annoyed when they found out that a computer had created the artwork," says Elgammal. "They then rejected pieces they had originally liked."
Ahmed Elgammal, Computer Scientist, Rutgers University (US-New Brunswick)