SoftThought, Inc. has developed a computer system, the Advanced Cerebral Emulator (ACE), which models image input as loosely organized shapes. All work is in the design of innovative artificial intelligence systems and models. The firm's models are neurologically inspired and reflect the information processing techniques believed to be involved in the human brain. Using the firm's shape analysis system, the Blob Descriptor Language (BDL) to convert salient features of an image into reduced elements for a self-managed database. We then use a convergent neural network for processing this information, adding functionality components which influence the decision-making processes. The firm uses neural plates which are neural network aggregates, which store information about objects in the environment, actions that are available to the model and cause and effect relationships between the neural plates. When combined withthe firm's image processing Fast Identification and Tracking (FIT) sytem, individual objects can be identified in any orientation, with small databases of salient features rapidly in real-time with a microprocessor. In 1999 Sogtthought completed designs for combat identification in target recognition under a Phase I SBIR for the U.S.A.F. through Wright Patterson AFB. The benefits of this work will extend to the intelligent interpretation of information by modeling the local environment of autonomous vehicles such as robots in space, military, medical, manufacturing and other hazardous environments. The firm's next stage of work includes development of an autonomous system which learns independently of a human operator and one which reasons about the problems it encounters.
