AI and robotics

Citation

(2003), "AI and robotics", Industrial Robot, Vol. 30 No. 2. https://doi.org/10.1108/ir.2003.04930baa.002

AI and robotics

AI and robotics

John M. Evans retired this year as Chief of the Intelligent Systems Division of the National Institute of Standards and Technology. He is now a part-time consultant in automation and robotics.

Keywords: Robots, Artificial intelligence

Artificial Intelligence traces its history from the 1950s with Turing’s formulation of measures of machine intelligence and with early work by Simon and Newell in recursive problem solving. Early achievements in theorem proving, symbolic manipulation and search led euphoric researchers to predict near-term success in programming computers to generally duplicate human intelligence. Reality soon set in, and funding for what looked like an increasingly far distant future slowly diminished.

AI research was flagging by the 1980s. Rodney Brooks at MIT proposed to revitalize AI research by attacking from a basic perspective in building robot insects. Brooks’ proposals soon found support and eventually created a widespread research effort in robots based on reactive behaviors. Robot insects based on reactive behaviors only need a modest amount of computing power to perform intriguing feats.

Ever since the late 1970s, Joe Engelberger, the “Father of Robotics”, has been predicting that some day robots would gain the mobility, sensory perception and intelligence to move off of the factory floor and out into service jobs working in direct contact with people. Engelberger has to this day promoted the concept of a household robot as the ultimate service robot, particularly for caring the aged and infirm.

The first service robots that Engelberger and I created, which had the computing power of lower level insects, have seen some success in floor cleaning and in hospital material transport, and other companies have had similar limited success in security applications. None have as yet reached the point at which the reliability and cost are so compelling that they become a whirlwind success story, but that will be accomplished in the foreseeable future. The newly released Roombavac robot vacuum cleaner and last year’s interactive doll were the first commercial successes of Brook’s behavioralist technology, and the vacuum looks to be a significant success.

There is also a great success story in mobile robots now emerging with military robots. Robot scout vehicles developed by the Army have reached the point that they can outrun manned chase vehicles during the day on some stretches of difficult terrain and can routinely outrun manned vehicles on almost any off-road terrain at night. Based in part on this early success, the Defense Department is committing literally billions of dollars to Future Combat Systems, a force of medium weight armored vehicles that will feature unmanned platforms for both sensing and combat.

With huge sums of money pouring into military applications, advanced robot technologies will be developed and engineered for reliability. It remains for entrepreneurs to carry out the technology transfer and cost reduction engineering to apply that technology base to commercial applications. This forecast has also been made by Hadi Akeel of Fanuc who predicted some time ago that military, medical and entertainment robots would create the technology base for industrial robots for the 21st century.

But more than anything else, the development and deployment of useful intelligent robots will be driven by the advent of computing power that reaches levels comparable to the human brain. Hans Moravec of Carnegie Mellon University estimates that the human brain is capable of approximately 100 trillion MIPs (1 MIP is one million instructions per second). Supercomputers of today are in the 10 trillion MIP range, and by Moore’s law we will reach the 100 trillion MIPs level in desktop scale machines sometime around 2020. Intelligent robots become a certainty by then.

This issue offers some brief glimpses into this future, describing some successes in applications and addressing the technical issues of mobility and localization. We can expect future issues of this journal to document further successes in the coming years.