Abstract:  Technology must form an integral component in the defense of our personal privacy. Policies and regulations will serve, at best, as lagging remedies in the fast-paced world of cyberspace. In a world where personal information can increasingly be transmitted and used in multiple locations simultaneously, protecting privacy may only truly be accomplished if the information itself becomes “intelligent” and capable of making appropriate decisions, relating to its release, on behalf of the data-subject. In other words, the data must become smart – we need SmartData. This presentation will discuss the growing need, the challenges, and ultimately, the benefits of developing intelligent agents to protect our privacy online.

Abstract:  Artificial Intelligence means different things to different people. For some, it is the building of useful intelligent tools for humans. When new -- as computer programs and Google once were -- they are 'indistinguishable from magic' but once familiar, they cease to be thought of as AI. For others, AI means attempting to build synthetic humans -- and philosophical differences as to what is the essence of a human lead to different assumptions built in. If rational intelligence is taken as the core, then perhaps computers and logic may be taken as the appropriate models. But if humans are seen in a biological context, as a particular form of adaptive organism, then very different techniques and models will be used - perhaps neural networks and evolution.

Abstract:  Advocates of SmartData have taken seriously two connections between SmartData and consciousness. (1) By developing SmartData, we can discover the otherwise elusive nature of consciousness. What motivates (1)? While we can't program a conscious agent top-down, we can evolve a conscious agent while developing SmartData, which must be conscious in order to be truly smart. Once we have evolved a smart, conscious agent, we can look back at the evolutionary process and discern exactly how it was built. One who knows how something is built thereby knows its nature. So, by developing SmartData we can discover the nature of consciousness. (2) In order to develop SmartData, we must know antecedently the nature of consciousness. What motivates (2)? There are some capacities that SmartData must have in order to be smart. Only conscious agents could have some of these required capacities. We cannot make SmartData conscious unless we know the nature of consciousness. So, in order to develop SmartData, we need to know the nature of consciousness. There is a clear tension between (1) and (2). In this talk I’ll dissolve the tension by rejecting both claims. I’ll suggest that developing SmartData cannot reveal the nature of consciousness and more importantly, that we don’t need to know much about consciousness in order to develop SmartData. Put another way, developing SmartData doesn't require deciding whether ZombieData can be Smart. Accordingly, discussions of consciousness should not be at the forefront of our explorations of SmartData. I’ll then close by briefly noting some ways in which philosophy of mind more broadly may contribute to the development of SmartData.

Abstract:  Vervaeke, Lillicrap, and Richards (2009) have argued that the central problem facing the project of cognitive science is explaining how cognitive agents selectively attend to relevant information while somehow flexibly ignoring a vast amount of irrelevant information in a manner that renders them capable of intelligent problem solving. Vervaeke et al. then went on to propose a dynamical solution to this problem they called relevance realization. Work in progress by Vervaeke, Ferraro, and Irving, argues that relevance realization is the core process of general intelligence and that this is being implemented in the self-organizing criticality of the brain.

Abstract:  The central concept that unites the projects of robotics, machine cognition, psychology, and action theory is that of agency. Thinking things are, first and foremost, agents. To build a thinking thing, then, one must build an agent. An agent, I maintain, is not primarily a reckoning machine, or an information-processing device. I use recent advances in evolutionary biology to argue that an understanding of the nature of agency can be gleaned by looking at the role of organisms in evolution. Organisms are agents par excellence, and an understanding of organismal agency is crucial to an understanding of the dynamics of adaptive evolution. I argue that an agent is a purposive entity that inhabits what I shall call an 'affordance landscape'. An affordance landscape is an emergent entity jointly constituted by the capacities of an entity and the features of its physical surroundings. Agency is thus not an architectural, computational, or informational phenomenon. It is an ecological phenomenon. To build an agent, then, one must build an entity that is capable of a certain kind of purposive ecological engagement with an affordance landscape.

Abstract:  Even in the sort of dynamically-shifting and open-ended scenarios in which we often find ourselves, human beings are extraordinarily proficient at maintaining psychological and behavioural focus on what is relevant in a situation, while ignoring what is irrelevant. This adaptive capacity for context-dependent relevance sensitivity is something that we would like to bestow upon artificial autonomous agents. The history of AI provides good evidence that this is a hard problem. In this talk, I shall argue that although autonomous agent research (or at least parts of it, given a certain philosophical gloss) and philosophical phenomenology (or at least parts of it, given a certain scientific gloss) have recently converged on a potential solution for the case of within-context relevance sensitivity, we do not yet have a compelling solution for the case of between-context relevance sensitivity (i.e., for the case of fluid and flexible context switching).