Talk: Inner Speech for AGI

Talk: OpenNARS for Applications: Overview and New Features

Talk: Machine Intelligence as Generalized Operant Behavior

Abstract: Contextual behavioral science is a broad research field emerging from behavioral psychology. Relational Frame Theory (RFT) is a contextual behavioral account of language and cognition that we previously have argued provides a valuable perspective on the necessary criteria for general-purpose intelligence. The fundamental thesis of RFT is that language and cognition are all examples of so-called generalized operant behavior. This is defined as particular forms of abstract goal-driven behaviors that are generally applicable across contexts. In this talk, we will show a series of experiments carried out in a subset of OpenNars for Applications that is only able to carry out sensorimotor reasoning. Increasingly complex tasks will be demonstrated up to an example of generalized operant behavior. We will explain what particular NARS rules are needed for the specific experiments. Finally, we will argue from an RFT perspective why semantic reasoning would be needed for more complex tasks such as stimulus equivalence and arbitrarily applicable relational responding.

Talk: NARS-based natural language understanding and NLP query processing

Talk: Explainable AI for First Responder Safety

Talk: The 'Explanation Hypothesis' in Autonomous General Learning

Abstract: General machine intelligence assumes that an agent can improve its own knowledge with respect to more than one goal related to more than one phenomenon, without outside help: A general learner's learning is not limited to a strict set of topics, tasks, or domains. General autonomous learning machines are still in the early stages of development, as are learning machines that can explain their own knowledge, goals, actions, and reasoning. In this talk I present the hypotheses that autonomous general learning requires (a particular kind of) explanation generation, and review some key arguments for and against it. Named the explanation hypothesis (ExH), the claim rests on three main pillars. First,that any good explanation of a phenomenon requires reference to relations between sub-parts of that phenomenon, as well as to its context (other phenomena and their parts), especially (but not only) causal relations. Second, that learning about new phenomenon in a self-supervised fashion requires (a kind of) bootstrapping, and that this bootstrapping – as well as subsequent improvement on the initial knowledge thus produced – relies on reasoning processes. Third, that general self-supervised learning relies on reification of prior knowledge and knowledge generation processes, which can only be implemented through appropriate reflection mechanisms, whereby current knowledge and prior learning progress is available for explicit inspection by the system itself. The claim thus construed has several important implications for the development of general machine intelligence, including that it will neither be achieved without reflection (meta-cognition) nor explicit representation of causal relations, and that explanation generation must be a fundamental principle of their operation.