ATLANTA -- The hardest thing about artificial intelligence (AI) is keeping your imagination in check. A visit to some robotic displays at an AI conference here opens the mind to incredible possibilities.
Imagine, for instance, CNBC's Jim Cramer, who just about jumps up and down when he talks about the "mobile Internet tsunami," doing something similar for the "robotics tsunami" as the next big industry. It is that kind of thinking that AI can trigger.
However, for the wonder of watching a robot with expressive eye movements, there is a competing reality that progress is slow. For a sense of the timeline, the Conference on Artificial Intelligence marks its silver anniversary next year.
"Early on there was this dream that robots could be generally intelligent; that they would rival and surpass humans in their abilities to do things," Leslie Kaelbling, a professor of computer science and engineering at MIT, said at the conference. "The current commercial reality is pretty different."
A lot of AI research fragmented in directions away from robotics, creating algorithms that underpin business intelligence, finance, Web and other uses. AI got separated from robotics because the machines are a pain: physical and unreliable. However, "They are getting better," Kaelbling said.
Today, robotics researchers have computers that are faster, machinery that is more reliable, and many of the algorithms used in routine robotic tasks have already been built, said Kaelbling, who asked this research community whether it was time to give robotics another try.
As robots go, the one assembled by a researcher at Brown University isn't pretty. It is an iRobot research device (the company that makes robotic vacuum cleaners that look like pancakes on wheels) that university researchers outfitted with a PC and a camera attached to it. It can zoom around the floor recognizing crime-scene like number displays.
You can use this robot to watch your cat via the Web, which is one of the things Sarah Osentoski, a postdoctoral computer science research associate at Brown University Robotics Lab, has done.
The robot on display here is more of a prop to illustrate a larger concept, which is Robotic Lab's plan to create something akin to create crowdsourcing environment for to test algorithms over the Web on robots in the lab.
Osentoski said that robot research has been constrained by one-off experiments, meaning a researcher can demonstrate a robotic capability that is very difficult for someone else to reproduce and compare because they don't have the platform.
An open platform that's easily accessible may also draw in a larger community of skills to test their ideas and build up the pool of data, or as Google demonstrated, "if you have lots and lots of data, learning becomes a lot easier," said Osentoski.
People want progress in robotics, said Osentoski, who believes researchers will be willing to share data because they have more to gain by it. "Robotics is at a point right now where it is still very preliminary," she said. The lab is due to open in September.
Andrea Thomaz, assistant professor in the Interactive Computing Dept. at the Georgia Institute of Technology, was at the conference demonstrating Simon, a human-like robot with a gentle, rounded, likeable face and eyes that will turn to you. Simon was moving blocks of various colors, sizes and shapes to demonstrate its learning capabilities.
Simon's hardware was built by Meka Robotics LLC in San Francisco. The face was designed by Georgia Tech. Thomaz directs the university's Socially Intelligent Machines Lab, which combines researchers who are from computer science, electrical engineering, and sometimes psychology students, interested in the human interaction side of robotics.
The lab focuses is on real world problems, including developing robots that can interact with humans which can involve learning from them, said Thomaz.
She believes the robotics industry is "going to be huge," and students are motivated, in part, by the idea that they might have a lifetime career in robotics.
Patrick Thibodeau covers SaaS and enterprise applications, outsourcing, government IT policies, data centers and IT workforce issues for Computerworld. Follow Patrick on Twitter at @DCgov or subscribe to Patrick's RSS feed . His e-mail address is pthibodeau@computerworld.com.