Intelligent graphics
Henry Lieberman · Communications of the ACM · 1996
A NEW paradigm is not just something that's a good idea.There are plenty of mere- ly good ideas, but a new paradigm must go beyond simple innovation.A new paradigm is often introduced to solve a particular problem, but it must do more than that.It must fundamentally change the way we look at problems we have seen in the past.It must give us a new framework for thinking about problems in the future.It changes our priorities and values, changes our ideas about what to pay attention to and what to consider important.Thomas Kuhn wrote most eloquently about the impact of new paradigms in science in The Structure of Scientific Revolutions [7].He traced the history of new paradigms such as quantum mechanics in physics.Such new paradigms are introduced in response to perceived problems with current practice, are advocated by enthusiastic iconoclasts, usually meet entrenched opposition, are accepted only slowly, and eventually become standard practice themselves.How does a new paradigm come about?Of course, the appearance of new paradigms is inherently unpredictable, but one thing that can be said is that the synergy between seemingly different disciplines is often a fertile ground for the growth of new paradigms.Here, I don't mean the kind of shallow interdisciplinary activity in which an expert in one field blithely assumes he or she can make pronouncements in another.Rather, when experts in different fields look with admiration and curiosity at each other's domains, and search for commonalities and fresh perspectives on their own, truly new paradigms can result. Intelligent Graphics • H e n r y L i e b e r m a nWhat's a new paradigm? A New Paradigm of AI and GraphicsIn my career, I've always felt caught between the fields of artificial intelligence (AI), beginning my career at the MIT AI Lab, and computer graphics, or more generally, human-computer interaction (HCI).I've always strongly felt that the goal of computing should be to enable collaborative problem-solving between people and machines, and that to place all the emphasis on the intelligence of the machine (as AI does) or all the emphasis on interfaces that couple the computer to the human's intelligence (as HCI does) would only be part of the picture.A new paradigm, what could perhaps be called intelligent graphics, is needed.Certainly, many events in the history of computer science point toward this synthesis.As far back as 1962, Ivan Sutherland's Sketchpad [16], introduced both the precursors of today's interactive graphical editors and CAD systems, and the AI technique of constraint networks.Early animation systems, such as those of Ron Baecker [1], Craig Reynolds, and others, tried to move animation beyond successions of static pictures to modeling underlying processes and characters, and offering animators assistance with the semantics of their imagined worlds.David Canfield Smith's Pygmalion brought learning-by-example techniques to an interactive visual programming language.As graphical computing became more widespread, some forward-looking graphic designers such as Aaron Marcus [13] helped educate the computer industry that visual design could play the role of a dynamic knowledge representation language for human-computer communication.Artists like Harold Cohen [14] and Myron Krueger [6] welcomed the participation of semi-autonomous programs in the creation of their art.For me personally, the development of a new paradigm of intelligent graphics was strongly influenced by my collaboration with graphic designer, Muriel Cooper, and her research group, the Visible Language Workshop at the MIT Media Lab.Sadly, Muriel died in 1994, but the work of the group continues.Though Muriel was not a technical person, nor I a professional graphic designer, we shared a common vision for a new paradigm of intelligent graphics.In the remainder of this article, I will explain a bit about this collaboration and some of the research of the group that contributed to a synthesis between artificial intelligence and graphics.