Top Takeaways from Robotics & AI TechCrunch Sessions

Two robots on the display at the NVIDIA table at the TechCrunch Robotics & AI Sessions. Source: John Waters

May 01, 2019      
John K. Waters

Attendees of the third annual Robotics + AI TechCrunch Sessions event, held at UC Berkeley’s Zellerbach Hall last month, saw warm temperatures and hot technologies during a conference packed with keynotes, panels, sessions, and demonstrations led by industry movers and shakers.

This year’s event, the second held at Berkeley, featured onstage interviews and presentations from a lineup of luminaries that included Claire Delaunay, vice president of engineering at NVIDIA; Marc Raibert, founder and CEO of Boston Dynamics; Colin Angle, founder and CEO of iRobot; Melonee Wise, CEO of Fetch Robotics, and Anthony Levandowski, co-founder and CEO of Pronto.ai, to name just a few.

Held directly across from the Berkeley student union, the event swarmed with students taking advantage of deeply discounted tickets to network, investigate internships and job opportunities, and rub elbows with their heroes.

There was a lot going on at this event, but here are a few of the top takeaways:

Platforms build ecosystems

“A platform is something that people can stand on,” Claire Delaunay told her audience, “For us, this is a platform play. So we have Isaac, which is software that you put on top of the hardware. We have a platform that can move around.”

NVIDIA promotes its Isaac SDK as “a developer toolbox for accelerating the development and deployment of AI-powered robots.” The company announced that it was open-sourcing the SDK at the conference, and Delaunay showcased two small robots that were built on the Isaac platform.

Delaunay, who heads up NVIDIA’s robotics effort, argued that platforms like Isaac, which create a universally accessible robotics platform by integrating AI software with robotics hardware, have become industry necessities. Manufacturers, researchers, and startups need these kinds of tools to cost-effectively add AI for perception and navigation into next-generation robots, she said. “At NVIDIA, what we really want to do is bring the two together.”

More robot-oriented businesspeople needed

NVIDIA Isaac SDK robots TechCrunch Sessions

CloudMinds showed a robot connected to cloud AI to make it more intelligent. Image: John Waters

The robotics industry is flush with entrepreneurs, Colin Angle said during an onstage interview, but what it needs right now and going forward are more “robot-oriented business people.”

“The industry needs people who can figure out how to take this amazing tool kit that is robotics into real solutions that deliver more value than they cost,” Angle said. “If you can do that, great things happen.”

Great things, such as what has happened at iRobot. The 29-year-old company that pioneered the in-home robot has sold more than 25 million Roombas, he pointed out. “About 25% of all the money spent on vacuum cleaners is now spent on robot vacuum cleaners,” he said. “It was all about getting the value proposition right, and that’s really hard.”

Self-driving vehicle makers: Focus on the boring stuff

It’s been 10 years since that Prius delivered a pizza from San Francisco to Treasure Island, completely unmanned,” Anthony Levandowski told his audience. “Today, there are no unmanned vehicles driving in California. It looked like it was really close back then. It feels like it’s really close now. It might be another 10 years before it actually happens on a big commercial scale. So that’s why, with Pronto, we’re trying something very different. Which is how do we build technology that can ship now, and add safety and real benefits today.”

Levandowski’s company, Pronto.ai, provides semi-autonomous capabilities for commercial trucking fleets. Its flagship Copilot system assists, rather than replaces, the driver. He’s betting on this approach, he said, because it’s a promise his company can deliver on right now.

“The whole approach that people are taking of going straight to Level 4 I think is flawed,” he said. “I think an evolutionary [approach] of getting technology to be better and better over time is actually very practical and helpful. That’s what we’re pursuing.”

Levandowski said his company is happily focused on Level 2. “It is super boring,” he said. “That’s why it’s going to work.”

TechCrunch Sessions Robotics & AI Franka Ermika

Hannes May, a solutions consultant at Franka Ermika, showed his company’s industrial robot arm, Panda, at the TechCrunch Robotics & AI Sessions event. The arm can touch things softly, such as pressing buttons. Image: John Waters

The best hardware isn’t necessarily hard

Underwater robots are among the most sophisticated and highly engineered devices in the industry. But their arms are heavy and clunky, and they tend to break off in rough water. The solution, Breeze Automation co-founder Gui Cavalcanti told his audience, lies in something called “environmental hardening.”

By combining injection-molded plastic joints with air- and water-tight fabric tubes, the San Francisco-based startup created a robot arm that can survive the pressures of the deep sea and perform delicate operations in space.

The underwater versions, in which the U.S. Navy has shown interest, take in ambient fluid — seawater — to provide an arm that can be as rigid or soft as the job requires. The version that NASA has shown interest uses air to provide the same capabilities.

“Our goal is to make the kind of lightest-weight-possible robots, the cheapest possible robots, out of the simplest materials,” Cavalcanti said, “so we can take on real-world challenges with robots that cost as much as the appliances in your house, instead of as much as a luxury car.”