Patrick Winston (co-chair)
Nick Cassimatis (co-chair)
Achieving Human-Level Intelligence through
Integrated Systems and Research
Although there has been substantial progress in some of the subfields of artificial intelligence during the past three decades, the field overall is moving toward increasing subfield isolation and increasing attention to near-term applications, retarding progress toward comprehensive theories and deep scientific understanding, and ultimately, retarding progress toward developing the science needed for higher-impact applications. Recent work in artificial intelligence in addition to cognitive psychology, neuroscience and linguistics presents an opportunity to reverse this specialization and reinvigorate the field’s focus on understanding and developing human-level intelligence.
Because there are so few venues for research on integration and because the opportunity is so great, we propose to gather researchers working across the boundaries of their subfields to explore new computational techniques and research methodologies for integrating research results to produce more intelligent systems.
We plan to address three broad topics of interest. First, what can models of vision, language, learning, and reasoning in fields such as cognitive psychology, linguistics and neuroscience contribute to artificial intelligence. Is there a way to characterize these results so that they can be more easily shared and combined across subfields. Second, how can we integrate multiple perception, action, representation, learning, planning, and reasoning systems to build cognitive models and intelligent systems that significantly advance the level of intelligence we can model or achieve? Is there a way to characterize the strengths and weaknesses of each approach and determine when to use each? Finally, what kind of theoretical, methodological, or technological innovations are needed to accelerate this research? Will it require advances in cognitive modeling, cross-domain and inter-subfield ontologies, or some kind of institutional transformation?
The topics of interest lead us to encourage a wide range of presentations, including presentations focused on the integration and interconnection of multiple systems, on the contributions of fields such as cognitive psychology, neuroscience, and linguistics to integration questions, and on methodological issues having to do with integration. Specific potential topics include:
· Approaches to integrating reasoning with perception and action.
· Exploring common representations and ontologies that can be used by several different subfields so that researchers can more efficiently find, communicate and use each other’s results.
· Cross-disciplinary metrics for measuring and evaluating intelligent systems that enable the range and efficacy of techniques form different subfields to be compared.
· Determining forms of collaboration to encourage subfields to take account of and contribute to other subfields.
· Cognitive robotic architectures for combing high-level cognition with techniques for autonomous robot perception and mobility.
· The integration of language processing with reasoning in other domains.
· Integrating common sense reasoning systems with realistic environments and input streams.
· Computational cognitive modeling.
· Computer vision techniques informed by human visual processing.
· Using tasks from the cognitive development literature as set of important but tractable problems for developing intelligent systems.
· Analogical and metaphorical reasoning techniques for leveraging knowledge and reasoning in one domain for cognition in another domain.
There will also be a joint session between this session and the session on “The Intersection of Cognitive Science and Robotics: From Interfaces to Intelligence”.
Potential participants should send a one-page abstract or position paper describing their research. Those interested in presenting research should submit a paper no longer than eight pages (AAAI format, preferably in PDF) to nc at alum dot mit dot edu. Papers will be published in the symposium proceedings and there will be many opportunities for panel discussions and participant interaction. Participants from all parts of the AAAI community as well as from other fields are encouraged.
Email all questions to nc at alum dot mit dot edu.