Architecture for Commonsensical Intelligence
April 15, 2008
"The Emotion Machine" describes a model for making machines with human-level commonsense knowledge and reasoning. Most previous AI models were based on single techniques (such as neural-network learning schemes, statistical methods, or simulations of evolution). But it turned out that each of these could only solve certain classes of problems. So, instead we're proposing to build an open software system that can assimilate different "ways to think," so that, when one method fails, it can switch to another. The model is based on multiple levels of "Critics," each of which tries to recognize obstacles that the system encounters, and then rearranges its ways to represent problems and its procedures for dealing with them. The traditional focus on finding a "single best method" has hobbled most previous AI research. Our goal is to develop and implement an architecture that is much more resourceful and extensible. Google "Minsky Home Page" to see more details.
Presenter Bio
Marvin Minsky, Massachusetts Institute of Technology
Marvin Minsky has made many contributions to AI, mathematics, cognitive psychology, computer science, linguistics, robotics, and optics. In recent years he has worked chiefly on imparting to machines the human capacity for commonsense reasoning. His conception of human intellectual structure and function is presented in two complementary books, The Society of Mind and The Emotion Machine, and he teaches a course on these subjects at MIT. Professor Minsky has degrees in Mathematics from Harvard and Princeton. His inventions include the first neural network simulator, various mechanical eyes, hands and other robotic devices, and the widely used Confocal Scanning Microscope. A member of the NAS, NAE and Argentine NAS, he has received the ACM's Turing Award, MIT's Killian Award, the Japan Prize, the IJCAI Research Excellence Award, the Benjamin Franklin Medal, and the Robert Wood Prize for Optoelectronics.
Close
