160 likes | 179 Views
Design/engineering-based theories of consciousness. Ron Chrisley Director, COGS University of Sussex. Introduction/overview. Role of theory in designing consciousness Sources of data for theories of consciousness Operationalising "consciousness"
E N D
Design/engineering-based theories of consciousness Ron Chrisley Director, COGS University of Sussex ESF PESC Exploratory Workshop: Models of Consciousness
Introduction/overview • Role of theory in designing consciousness • Sources of data for theories of consciousness • Operationalising "consciousness" • Design/engineering approach to constructing theories of consciousness • Misc.: other minds, prosthetic consciousness ESF PESC Exploratory Workshop: Models of Consciousness
The role of theory • Assume our primary goal is the construction of machine consciousness • Then why do we need a theory of consciousness, anyway? • To guide design (which in turn guides actual construction) • So should consider a theory's ability to guide design when constructing and evaluating it ESF PESC Exploratory Workshop: Models of Consciousness
Theory Design Artefact Flow of effect is not one-way: ESF PESC Exploratory Workshop: Models of Consciousness
Design 1 Artefact 1 Design 2 Theory 2 Artefact 2 Design 3 Rather, an interactive process Theory 1 ESF PESC Exploratory Workshop: Models of Consciousness
Constraints on a theory • Design • At least: Implementable • Better: Suggests designs • Data • At least: Consistent with • Better: Explains • Other • Simplicity • Unifiability with other theories ESF PESC Exploratory Workshop: Models of Consciousness
Sources of data • Phenomenology (e.g., entire visual field is coloured) • Phenomenological reports (e.g., subjects report that entire visual field is coloured) • Behaviour (other than phenomenological reports; e.g., grip of blindsight subjects) • Architectural constraints (conceptual; e.g., fear that bridge might collapse requires ability to entertain counterfactuals) ESF PESC Exploratory Workshop: Models of Consciousness
Sources of data • The forementioned data occur in bothb normal and pathological (cf Taylor's presentation today) forms • What about neuroscience? • For our purposes, mainly inspiration • But also a "clear case" (discussed below) ESF PESC Exploratory Workshop: Models of Consciousness
Operationalising "consciousness" • If we just think of our goal as building a conscious machine, it can be difficult to see how to proceed, or to tell if we have succeeded • Instead, we can "operationalise" the notion of consciousness: • Why are we interested in making a conscious machine, anyway? ESF PESC Exploratory Workshop: Models of Consciousness
Operationalising "consciousness" • Interested in making a machine that has: • Autonomy • Adaptivity/advanced learning capacities • Emotion/affect • Responsibility (or to which we are responsible) • Intelligence • Authenticity (own world view and goals) • Ability to integrate information from different sources/modalities ESF PESC Exploratory Workshop: Models of Consciousness
Operationalising "consciousness" • Interested in making a machine that has: • Vivid/meaningful sensation/perception • Ability to act in the world • Ability to simulate/imagine/plan • Ability to represent its own states • Attentional capacities • A belief that it is conscious, gives phenomonological reports ESF PESC Exploratory Workshop: Models of Consciousness
Design/engineering approach to machine consciousness Non-implementable theories, that claim that consciousness is • Indefinable • unknowable • epiphenomenal (no causal powers) • just a mythinvented by philosophers • not possessable by (non-biological!) machines • such that machines indistinguishable from us could lack it arenot useful for a design/engineering based approach, and should be off-limits during this workshop. ESF PESC Exploratory Workshop: Models of Consciousness
Design/engineering approach to machine consciousness • Thus, on the design/engineering approach it will not count as an objection to a theory that one can imagine (in Chalmers' strong sense) something which satisfies that theory but is not conscious ESF PESC Exploratory Workshop: Models of Consciousness
Design/engineering approach to machine consciousness • What will count is a theory's stance on clear cases (positive and negative) • It will count against a theory if it: • Implies that something isn't conscious which, pre-theoretically, is clearly conscious (e.g., humans) • Implies that something is conscious which, pre-theoretically, is clearly not conscious (e.g., a stone) • Of course, what we take to be the "clear" cases can change ESF PESC Exploratory Workshop: Models of Consciousness
Other minds • If you want to know under what conditions you could ever know a machine to be conscious, it will probably help if you figure out what it is by virtue of which you know other humans to be conscious. ESF PESC Exploratory Workshop: Models of Consciousness
Prosthetic consciousness • An exciting alternative to creating "stand-alone" conscious machines is to use theories of consciousness to create technologies which augment our own consciousness • Specifically, non-pharmocological technologies • Allows one to avoid many of the epistemological problems that arise with "stand-alone" conscious machines ESF PESC Exploratory Workshop: Models of Consciousness