I wish honorable gentlemen would have the fairness to give the entire context of what I did say, and not pick out detached words. (Cobden, Speeches 46, 1849, quoted intheOED) The importance of contextual reasoning is emphasized by various researchers in AI.

A formalism is elaboration tolerant to the extent that it is convenient to modify a set of facts expressed in the formalism to take into account new phenomena or changed circumstances. Representations of information in natural language have good elaboration tolerance when used with human background knowledge. Human-level AI will require representations with much more elaboration tolerance than those used by present AI programs, because human-level AI needs to be able to take new phenomena into account. The simplest kind of elaboration is the addition of new formulas. Next comes changing the values of parameters. Adding new arguments to functions and predicates represents more of a change. However, elaborations not expressible as additions to the object language representation may be treatable as additions at a meta-level expression of the facts. Elaboration tolerance requires nonmonotonic reasoning. The elaborations that are tolerated depend on what aspects of the phenom...

Abstract. An architecture is described for designing systems that acquire and manipulate large amounts of unsystematized, or so-called commonsense, knowledge. Its aim is to exploit to the full those aspects of computational learning that are known to offer powerful solutions in the acquisition and maintenance of robust knowledge bases. The architecture makes explicit the requirements on the basic computational tasks that are to be performed and is designed to make these computationally tractable even for very large databases. The main claims are that (i) the basic learning and deduction tasks are provably tractable and (ii) tractable learning offers viable approaches to a range of issues that have been previously identified as problematic for artificial intelligence systems that are programmed. Among the issues that learning offers to resolve are robustness to inconsistencies, robustness to incomplete information and resolving among alternatives. Attribute-efficient learning algorithms, which allow learning from few examples in large dimensional systems, are fundamental to the approach. Underpinning the overall architecture is a new principled approach to manipulating relations in learning systems. This approach, of independently quantified arguments, allows propositional learning algorithms to be applied systematically to learning relational concepts in polynomial time and in a modular fashion.

This article presents a formal theory of robot perception as a form of abduction. The theory pins down the process whereby low-level sensor data is transformed into a symbolic representation of the external world, drawing together aspects such as incompleteness, top-down information flow, active perception, attention, and sensor fusion in a unifying framework. In addition, a number of themes are identified that are common to both the engineer concerned with developing a rigorous theory of perception, such as the one on offer here, and the philosopher of mind who is exercised by questions relating to mental representation and intentionality.

Bootstrapping is a widely employed technique in the process of building highly complex systems such as microprocessors, language compilers, and computer operating systems. It could play an even more prominent role in the creation of computation systems capable of supporting intelligent agent behaviors because of the even higher level of complexity. The prospect of a self-bootstrapping, self-improving intelligent system has motivated various fields of research in machine learning. However, a robust, generalizable methodology of machine learning is yet to be found; there are still a lot of learning behaviors that no existing learning technique can adequately account for. We believe a uniform, logic-based system such as active logic [1, 2], will be more successful in the realization of this ideal. The overall architecture that we envision is as follows: a central commonsense reasoner module attends to novel situations where the system does not already have expertise, and to its own failures; it then reasons its way to solutions or repairs, and puts these into action while at the same time causing "expert" modules to be either created or retrained so as to more quickly enact those solutions on future occasions. Thus what we propose is a kind of meld between declarative and procedural techniques where the former has great expressive power and flexibility (but is slow) and the latter is very fast but hard to adapt to new situations.

... this paper we study two special cases of this framework: full systems and hypercubes. Both model static situations in which no agent has any information about another agent's state. Full systems and hypercubes are an appropriate model for the initial congurations of many systems of interest. We establish a correspondence between full systems and hypercube systems and certain classes of Kripke frames. We show that these classes of systems correspond to the same logic. Moreover, this logic is also the same as that generated by the larger class of weakly directed frames. We provide a sound and complete axiomatization, S5WDn , of this logic, and study its computational complexity. Finally, we show that under certain natural assumptions, in a model where knowledge evolves over time, S5WDn characterises the properties of knowledge not just at the initial conguration, but also at all later congurations. In particular, this holds for homogeneous broadcast systems, which capture settings in which agents are initially ignorant of each others local states, operate synchronously, have perfect recall, and can communicate only by broadcasting.

Recognition of complex activities from surveillance video requires detection and temporal ordering of its constituent "atomic" events. It also requires the capacity to robustly track individuals and maintain their identities across single as well as multiple camera views. Identity maintenance is a primary source of uncertainty for activity recognition and has been traditionally addressed via different appearance matching approaches. However these approaches, by themselves, are inadequate. In this paper, we propose a prioritized, multivalued, default logic based framework that allows reasoning about the identities of individuals.

Abstract: The paper presents the results of mapping of PSI Ontologies family to the foundational ontologies: WordNet, SUMO and DOLCE. The two main outcomes of the presented research are: reported manual technique may be used as initial evaluation of ontology claiming to be gold standard for a new domain; and: usage of mentioned foundational ontologies for alignment of ontologies family of a given domain has shown differences between SUMO and DOLCE, two formal upper-level ontologies of common sense knowledge. The research reported was performed in the frame of our PSI project 1. 1

In the discussions on the limitation of Artificial Intelligence (AI), there are three major misconceptions, which identify an AI system with an axiomatic system, a Turing machine, and a system with a model-theoretic semantics, respectively. Though these three notions can be used to describe a computer system for certain purposes, they are not always the proper theoretical notions when an AI system is under consideration. These misconceptions are not only the basis of many criticisms of AI from the outside, but also responsible for many problems within AI research. This paper analyzes these misconceptions, and points out the common root of them, that is, to treat empirical reasoning as mathematical reasoning. Finally, an intelligent system, NARS, is introduced as an example, which is neither an axiomatic system, nor a Turing machine in its problem-solving process, and does not use a model-theoretic semantics, though is still implementable in an ordinary computer.

Bootstrapping is a widely employed technique in the process of building highly complex systems such as microprocessors, language compilers, and computer operating systems. It could play an even more prominent role in the creation of computationsystems capableof supporting intelligent agent behaviors because of the even higher level of complexity. The prospect of a self-bootstrapping, self-improving intelligent system has motivated various fields of research in machine learning. However, a robust, generalizable methodology of machine learning is yet to be found; there are still a lot of learning behaviors that no existing learning technique can adequately account for. We believe a uniform, logic-based system such as active logic [1, 2], will be more successful in the realization of this ideal. The overall architecture that we envision is as follows: a central commonsense reasoner module attends to novel situations where the system does not already have expertise, and to its own failures; it then reasons its way to solutions or repairs, and puts these into action while at the same time causing "expert" modules to be either created or retrained so as to more quickly enact those solutions on future occasions. Thus what we propose is a kind of meld between declarative and procedural techniques where the former has great expressive power and flexibility (but is slow) and the latter is very fast but hard to adapt to new situations.