If you are looking for an answer to the question What is Artificial Intelligence? and you only have a minute, then here's the definition the Association for the Advancement of Artificial Intelligence offers on its home page: "the scientific understanding of the mechanisms underlying thought and intelligent behavior and their embodiment in machines."
However, if you are fortunate enough to have more than a minute, then please get ready to embark upon an exciting journey exploring AI (but beware, it could last a lifetime) …
Thus, configuring itself does not involve the development of new component types . The development of new component types, e.g. for addition to the component catalogue, rather is a typical task in the domain of design. Configuring of technical systems is an ubiquitous industrial engineering task and has been a domain for expert system application ever since the longtime paragon R1/XCON . Most configuration expert systems agree in using frames or objects for the representation of factual knowledge about component types [2, 3]. For the knowledge about how to use the factual knowledge and how to fred a good configuration quickly, two approaches are used.
Furthermore, when a role of an 2. No primitive of C1 is disjoint from any primitive of C2 ultimate concept does not have a positive at-least restriction, the individual may decline to fill that role by specifying an 3. restricted-roles(C1) CC_ restricted-roles(C2) at-most restriction of zero, e.g., (at-most 0 SECONq3ARY-4.
The definitional knowledge will be represented by a set D of formulae in our language. The consistency information will be a set / of integrity constraints. Now we are ready to outline the main idea of Constructive Problem Solving [Klein94]: given a configuration domain (characterized by the definitional knowledge D and the domain constraints I), and a concrete configuration problem (described by a goal G), a solution will a finite mode/C fulfilling the following conditions3: and C I D 3.G C I D I Thus Constructive Problem Solving (CPS) represents in a well-formalized way what is actually happening configuration: to a given configuration problem G a solution C is generated which is a model, i.e., which has to fulfil this goal and the general constraints in the domain. In this way CPS mode/construction provides a well-founded approach to the synthetictype of configuration problem solving.
Configuration is the task of selecting and arranging parts to provide a desired function without violating part-specific constraints. Configuration problems arise in design, manufacturing, sales, installation and maintenance. The parts need not be physical, for example, on could configure an insurance policy or an investment strategy. Configuration problems have a long history in AI going back at least to the pioneering R1/XCON expert system for configuring computer systems. Recently the area has been revitalized by: Renewed industrial interest: Edward Feigenbaum highlighted configuration in his "Tiger in a Cage" talk at the AAAI-93 configuration.
The problem of inductive learning is hard, and-- despite much work--no solution is in sight, from neural networks or other AI techniques. I suggest that inductive reasoning may be grounded in sensorimotor capacity. If an artificial system to generafize in ways that we find intelfigent it should be appropriately embodied. Tiffs is illustrated with a network-controlled animat that learns n-parity by representing intermediate states with its own motion. Unlike other general learning devices, such as disembodied networks, it learns from very few examples and generalizes correctly to previously unseen cases. Induction In artificial inductive learning, the machine is trained oll only part of the set of possible input-output pairs; once it reaches some criterion of success on this training set, the machine is tested for "correct" generalization on the remaining test set.
A new view of mental rotation in humans is presented. Rather than being a perceptual phenomenon, mental rotation of objects is supposed to be an imagined action in the sense that its only difference to real action is the absence of motor output. A series of experiments is reported which shows that the difference in speed between mental and manual rotation are negligible and that performing rotational hand movements interferes with mental rotation and vice versa. It also could be shown that the preparation of rotational hand movements is already sufficient to influence mental rotation. The general role of motor processing in dynamic visual imagery is discussed, considering the underlying neurophysiology.
Some of the key tenets and tendencies of the Embodied Artificial Intelligence (AI) approach (taken mostly from (Brooks 1991)) are: T1 AI should be embodied in robots. T2 AI must be situated; the robots must operate in the real world, not an artificial simplified one. T3 AI requires adequate perception, not perfect perception. T4 AI can be reactive, that is, based on simple behaviors that are fairly direct links between perception and action, not mediated by representation and reasoning. T5 AI can emerge from simple behaviors, thanks to interaction with the outside world.
Modular-functional decomposition is a fundamental tenet of Computer Science. Cognitive Robotics, with strong roots in Cognitive Science and Biology, replaces modularfunctional decomposition with a more opportunistic approach. Nonetheless, we can extract heuristics with both analytic and synthetic power: architectural principles for neo-modular systems. This paper describes three neo-modular principles: Imagination, Shared Grounding, and Incremental Adaptation. It includes examples of each drawn from existing systems, and concludes with an illustration of these three principles used in concert to build a system which progresses first to hand-eye coordination, then to planned complex reaching, and finally to shared attention.