August 1997, Volume 14, No. 8
|
|
August 1997, Volume 14, No. 8 AAAI '97: a report from Providence |
|
|
It's that age-old paradox at play again: AI proponents want their work to be taken seriously by the world at large, but too often they hang their hats on whatever hook has captured the public's attention. That's why, for instance, you could find most of the 1,000 AAAI '97 attendees watching the robotic competitions, while the audience at the Innovative Applications of AI track — which focuses on deployed, real-world systems achieving significant paybacks — was usually outnumbered by empty chairs. (By the way, the 1,000 attendees represents a marked 33% drop from the 1,500 who attended last year's AAAI '96 in Portland, Ore. Clearly, something is amiss if the field is as popular as it's ever been, the media is beside itself profiling the latest successes computers and robots have attained, and yet the premier showcase event for the AI field can only scare up 1,000 people. Maybe everybody decided to go to the nearby resort beaches.) Be that as it may, those who attended the many and varied
programs AAAI '97 had to offer were treated to a
retrospective of how far the field has come, as well as a
glimpse at the widely divergent paths future researchers
will be taking. Like teenagers, the AI field is something of an enigma. Where else, Allen asked, could you find a discipline where people with no technical background make claims about the viability of the systems? Equally confounding is that those who ought to know better — the researchers actively working in the field — are constantly misjudging the progress that is being made and are continually moving the goalposts closer or further away as the mood hits them. "We're now at a significant transition point in the field," Allen said, "where we are seeing 'working' systems." By "working," Allen refers to systems designed to work within a specific domain, rather than a "canned" demo. The defining goal of artificial intelligence, he noted, is not to build an artificial person, but rather, to create intelligent behavior. To that end, then, he defined AI as "the science of making machines do tasks that humans can do or try to do." To those researchers content to isolate themselves in pursuit of theories, Allen warned that "AI will not develop without the building of systems, just as civil engineering does not progress without the building of bridges. System building is the experimental foundation on which the science of AI is based." Later in this issue, we will look at the state-of-the-art
in AI systems, as presented at the Innovative Applications
of AI sessions. Richard Doyle, a member of the Autonomy Technology Program at NASA's Jet Propulsion Laboratory (Pasadena, Calif.), delivered a stirring address on the emergence of spacecraft autonomy, highlighting the role of artificial intelligence and AI developers in keeping NASA financially able to continue funding planetary exploration projects like Pathfinder. "The single greatest driver which has led to the emergence of spacecraft autonomy as a legitimate, perhaps critical application of AI technologies within NASA is the need to reduce the lifecycle costs of space missions," Doyle said. Autonomy technology will help reduce mission operations costs, through the automation of spacecraft functions, as well as closing loops onboard and decoupling spacecraft from ground. Autonomy will also contribute to a reduction in the ground workforce needed for mission support. The Remote Agent is NASA's showcase effort in spacecraft autonomy, according to Doyle. This system consists of a smart executive, a planning and scheduling module, and a mode identification and reconfiguration module. The Remote Agent "receives mission goals as input, and the smart executive provides robust, event-driven plan execution and runtime decision making," he explained. It is being designed as a core architecture for autonomous spacecraft. According to NASA's Mel Montemerlo, the funding of AI at
NASA is holding steady. For the past several years, NASA
researchers have begun requesting AI technologies for their
projects, and AI personnel are climbing the corporate ladder
at NASA. In addition to autonomy projects such as Remote
Agent, other AI areas of interest at NASA include turning
data into knowledge, and intelligent rovers. The irony of the situation, of course, is that Deep Blue
does not have any artificial intelligence, and its creators
have never made any claim that it did. On the other hand,
$100,000 is nothing to sneeze at, so the Deep Blue team
happily accepted the award. Developers of earlier chess
programs were also honored at the award ceremony, in
recognition of the pioneering work they did in the field.
Garry Kasparov himself was also invited, but he declined
"due to his vacation schedule." Uh-huh. Perhaps the blame lies with the configuration of the Rhode Island Convention Center, but whatever the reason, the handful of vendors at the AAAI show were all but buried behind both the robot competition arena as well as a Hall of Champions, devoted to man-vs.-machine matches in such games as chess, scrabble and backgammon. There actually was an exhibition at the show, though only three mainstream AI vendors showed up: Brightware, Franz and Harlequin. The rest of the 16 booths were occupied mostly by book publishers and a handful of research labs. Clearly, the AAAI show is no longer the place for vendors to showcase their wares.
|
|
|
Intelligent Systems Report Resources: Contents of Current Issue | Subscriber Access — Index of Past Issues Free Access — Index of Past Issues Search | Subscriptions | Reprints | Links E-mail to the Editorial Department of ISR: Lionheart Site Resources: Lionheart's Home Page | What's New | Site Index | Newsroom | Contact Software Surveys | Online Sponsors | Visitor Survey Other Publications: APICS — The Performance Advantage BDO Seidman, LLP Comprehensive Guide to Manufacturing Software Operations Research/Management Science Today | Intelligent Manufacturing Subject Areas: Manufacturing & Resource Planning Operations Research and Management Science | Intelligent Systems | Lionheart's Publishing Services
|
