By Glenn Smith
As many members of the SciArt community will be aware, the current explosion in artificial intelligence research is being increasingly driven by the life sciences. In particular, the core AI algorithm which is currently gaining so much attention and achieving such incredible results in applications as diverse as speech recognition, photo interpretation, game playing, and even weather forecasting - the Deep Neural Network, or DNN - is itself closely modeled after the architecture of the animal/human brain.
Of even greater possible interest to the SciArt community, however, is the fact that, with formalistic board games having been more or less thoroughly mastered by the computer, AI researchers are increasingly turning their attention to the graphic arts as the new arena of choice for their work - and with astounding results. Imagine, for instance, a DNN-based system which, having been "trained" on examples of paintings by Van Gogh or Picasso, can then take an arbitrary photograph and re-render it as if painted by one or the other of those artists. And of course there is also the much-publicized "DeepDreaming" research, wherein a DNN "stack" trained to be able to discriminate, for example, between photos of various dog breeds is instead presented with a photo in which there is no dog - a photo, say, of a horse and rider - and one of the characteristic inner states of that stack itself then output as an image: there are dogs everywhere, their heads sprouting in psychedelic fashion in place of the pommel of the saddle and the ears of the horse, and with even the horse's nostrils converted into the bulging eyes of a pug.
All of this is the territory to be covered by a new special issue of the MDPI/Arts journal entitled "The Machine as Artist (for the 21st Century)" - see the formal announcement as well as the introductory essay - and which special issue has the further ambition of connecting an aesthetic capability on the part of the computer to improved prospects for both a "friendlier" AI and an increased sensitivity to the health of the planet as a whole. Serving, in turn, as Guest Editor of this comprehensive effort is Professor Frederic Fol Leymarie of the University of London. The focus of his research team is the development of algorithms which will allow robots to create stroke-based drawings, a capability which represents both an increased delicacy of operation, and an enhanced understanding of the actual entity being depicted (see Baxter, above).
Learn more and submit HERE.