|AAAI Fall Symposium on|
Anchoring Symbols to Sensor Data |
in Single and Multiple Robot Systems
Abstract: Classical artificial intelligence systems presuppose that all knowledge is stored in a central database of logical assertions and that reasoning consists largely of searching and sequentially updating that database. While this model has been very successful for disembodied reasoning systems, it is problematic for robots. I will discuss an alternative class of architectures "tagged behavior-based systems" that support a large subset of the capabilities of classical AI architectures, including limited quantified inference, forward- and backward-chaining, simple natural language question answering and command following, reification, and computational reflection, while allowing object representations to remain distributed across multiple sensory and representational modalities. Although limited, they also support extremely fast, parallel inference.
Abstract: I shall argue that the original symbol grounding problem is caused by the assumptions of a realist semantics and an externalist view on symbols. If a cognitive (conceptualist) view on semantics is assumed, the problem becomes more manageable. Conceptual spaces will be briefly described. I will argue that these representational tools can be used to handle the meanings of symbols and thereby anchor them.
Abstract: The Selective Tuning model of visual attention is one of the strong hypotheses for explaining experimental results from psychophysical, neurophysiological and imaging attention studies in primates and humans. I will present the basics of the model and show some examples of its functionality as well as its biological predictions. Previously I had also shown how strict subsumption architectures cannot scale up to human-like performance without the addition of attention mechanisms, intermediate representations and goals. I will overview those results and then present a suggestion for how the integration of the selective tuning model of attention and behaviors can be accomplished in order to form a new more powerful architecture for control.
Back to the program
[ Symposium Home ]