Vincent Perricelli has written a summary of the joint 16th National Conf. on AI (AAAI-99) and 11th Innovative Applications of AI Conference (IAAI-99) held in Orlando this July. He says the speakers had great optimism about a new "AI spring" arising from successful applications in space exploration, genomics, neurosurgery, and elsewhere. CMU's CMUnited-98 robot soccer team was highlighted at the conference (), as was Greg A. Keim's best-paper-winning "Proverb" crossword puzzle solver. Eric Grimson at MIT has a video surveillance system that learns to recognize unusual events, such as jaywalkers in rush hour traffic. (It has processed a billion images in three years. See ). Genome studies have developed new clustering and computational methods for assembling sequence data and understanding gene functions. AI and HCI are beginning to enable tasks -- e.g., difficult brain surgeries -- that would otherwise be difficult to impossible. This is almost a resurgence of expert systems, but more focussed on specific applications than extensible tools. We know more about biological intelligence now, and have new tools in Bayesian reasoning, analogical reasoning, planning, genetic algorithms, machine learning, ontologies, and agent-based systems. Plus we have more computing power, esp. in embedded systems. Success in applications (and in funding) will also bring cross-fertilization from other fields. (Nils Nilsson did warn that niche applications fragment the field, whereas grand challenges such as human-level AI lead to fusion. Fritz Lehmann of Cycorp had similar feelings about XML semantic text labels -- an extension of HTML -- as "the road to Babel," whereas James Hendler of UMD and DARPA/ISO believed that XML ontologies could be the "greatest thing to happen to AI.") Patrick Winston says that AI will advance rapidly as we come to understand the brain's I/O channels, which contain much of human's intelligence capability. Much can also be learned from surgical rewiring of animal brains. Ruzena Bajcsy noted that the federal government is increasing support for information technology research, and that the money may start flowing more to CS than to supercomputers, bandwidth, and infrastructure. Nilsson expects to see "robotic factotums in about 2010, robot chauffeurs in 2015, and conscious devices in 2050. Borrowing from Nobelist James D. Watson, Nilsson urged AI researchers to learn from the winners, take risks, have a fallback, have fun, and stay connected."

One notable application at IAAI-99 was Eero Hyvonen's Interval Solver, an Excel spreadsheet add-in for interval constraint satisfaction problems. Given value intervals for all variables, Interval Solver can compute bounds on possible values for the function. It can also give variable ranges for a desired function value, and can solve inequalities. Delisoft, Ltd., .

Another useful tool may be Andrew Moore's AD-trees (all-dimensions trees, and ) for storing and updating probabilities in Bayesian networks. AD-trees are very compact data structures that can sometimes speed Bayesian learning a thousandfold.

A list of conference speakers and papers is available online at . Proceedings can be purchased from AAAI Press, , and audio tapes from Audio Archives International, (800) 747-8069. AAAI-2000/IAAI-2000 will be held 30Jul-03Aug00 in Austin; IJCAI-2001 will be in Seattle. [David Kinny , comp.ai, 31Aug99.]

----- "Far away there in the sunshine are my highest aspirations. I may not reach them, but I can look up and see their beauty, believe in them, and try to follow where they lead." -- Louisa May Alcott. [PAQOTD, 07Apr99.] -----