Rodney Brooks at NVCA: Why the World Needs Robots, and What VCs Need to Watch Out For
(Page 2 of 2)
industrial robot. (Current machines have integration costs that are 5 to 10 times their capital cost, Brooks said—“it’s a big-company game.”)
Nevertheless there are pitfalls for VCs looking to invest in robots. “It’s tricky to be a robotics investor,” Brooks said. Not worse than other areas, he said, but the field has particular challenges. Such as:
—It’s difficult to tell what is hard and what is easy. “A number of discussions with VCs have worried me,” Brooks confessed. Basically, a lot of fancy robotics demos out there look impressive but are not suitable for real-world products. The Honda humanoid robot, for instance, is a “wind-up toy,” Brooks said. If you change the environment slightly or something unexpected happens, the robot goes out of whack. He also told a story of visiting a lab in Tokyo where researchers had their robot pour Brooks a cup of tea. When they gave it a bottle of beer instead, the robot poured beer all over its feet—because the beer bottle was an inch shorter than the tea bottle. (Perceiving and understanding its environment is hard for a robot.)
—It’s difficult to predict whether the market will be responsive to a radically new solution. This one isn’t unique to robotics, but perhaps it’s even tougher with robots. Brooks put up a slide of the “14 failed business models at iRobot” (courtesy of CEO Colin Angle). Among them: sell nuclear power plant inspection robots; sell land mine clearance robots; sell robots to the oil industry to stimulate production in wells; earn royalties on robotic toys; sell research robots to universities and hobbyists. You get the idea—timing is key.
—Conventional top-down engineering often goes astray in developing a robust robot. People often think, well, a plane can pretty much fly autonomously, so why can’t robots do more? The difference is that the conditions a robot deals with in an unstructured environment—a home, office, or cave in Afghanistan—are hard to measure, and they change every few tens of milliseconds. As things come up, the robot needs to be flexible and adjust—and that’s hard to engineer.
All in all, pretty useful stuff from a guy who says he’s “tarred as impractical” because of his professorial career. VCs, take notice.