Transcript Document
AGI versus Narrow AI Ben Goertzel, PhD Novamente LLC Biomind LLC Artificial General Intelligence Research Institute Virginia Tech, Applied Research Lab for National and Homeland Security Artificial General Intelligence (AGI) “The ability to achieve complex goals in complex environments using limited computational resources” • Autonomy • Practical understanding of self and others • Understanding “what the problem is” as opposed to just solving problems posed explicitly by programmers 2006 2007 Post-proceedings of 2006 AGI Workshop (North Bethesda MD, May 2006) To be published by IOS Press Narrow AI The vast majority of AI research practiced in academia and industry fits into the “Narrow AI” category Each “Narrow AI” program is (in the ideal case) highly competent at carrying out certain complex goals in certain environments •chess-playing, medical diagnosis, car-driving, etc. A major lesson from the history of AI is: Narrow AI success is nearly always of minimal use as a steppingstone toward AGI success Some Existing AI Paradigms Paradigm Strengths Weaknesses GOFAI Representation of abstract knowledge Reasoning (short proofs) Pattern recognition Learning Autonomy Neural nets Pattern recognition Learning Associative memory Perception/action/cognition integration Representation of abstract knowledge Abstract reasoning Learning Autonomy Evolutionary Programming Pattern recognition Learning of complex procedures Representation of abstract knowledge Abstract reasoning Autonomy Probabilistic Reasoning Representation of abstract uncertain knowledge Reasoning (short proofs) Hypothesis formation Autonomy Pattern recognition Subsumption Robotics Autonomy Learning Perception-action integration Cognition Representation of abstract knowledge Examples of Narrow AI Deep Blue Can’t learn to play a new game based on a description of the game rules DARPA Grand Challenge Software can’t even learn to drive different types of vehicles besides cars (trucks, boats, motorcycles) Google Can’t answer questions. Whatever happened to AskJeeves? Examples of Narrow AI From My Own Recent Work Biomind ArrayGenius software for recognizing patterns in gene expression data Doesn’t interpret the patterns it finds in the context of the literature and the totality of relevant online quantitative data RelEx software for mapping English sentences into semantic structures Doesn’t do reasoning to resolve semantic ambiguity in a context-appropriate way How to Create AGI? AGI will not be achieved by incrementally “general-izing” narrow-AI apps To achieve AGI, we need to go back to basics We need to create an artificial baby of sorts (though not necessarily a human-like baby … it may be a baby with architecture and algorithms sculpted via computing, cognitive and systems science rather than neuroscience) We need to teach the baby well And then, the practical applications will come (and oh boy, will they come!!)