The Head of CMU's Robotics Lab Says Self-Driving Cars Are ‘Not Even Close’
A conversation on rebuilding after Uber gutted the top lab.
Uber doesn't play nice with whoever or whatever stands in its way. But the notoriously cut-throat ridesharing company surprised even its own critics when it partnered with one of the most advanced university robotics labs in the US last year—only to immediately poach a third of its top scientists and its director.
At the time, Carnegie Mellon University's robotics lab, NREC, was one of the most important hubs for autonomous car research in the nation, and Uber CEO Travis Kalanick desperately wanted in. Kalanick flew to Pittsburgh (where CMU is based) to forge a partnership with the university to work on autonomous cars. Instead, Uber lured away 40 of the lab's 100 or so scientists, including its director, and set up shop less than a mile away from the school.
Read More: Uber Is Working to End Ridesharing
Headlines screamed that the university lab was in "crisis" mode as entire projects folded, and the lab braced for its funding to be cut in half by the end of 2015. Herman Herman, a roboticist and commercialization specialist, stepped up as the new director of the robotics lab on February 23, 2015, effective immediately.
More than a year later, that initial and so-called partnership hasn't resulted in any joint projects between the company and the school, although Uber eventually gave CMU a $5.5 million gift, which could be interpreted as both a "thank you" and "sorry." Recently, the company's robot cars have been spotted prowling the streets of Pittsburgh, the result of a $300 million deal with automaker Volvo.
To find out what it's been like to rebuild a top robotics lab after being gutted by one of the most powerful companies in the world, I called up Herman for a candid conversation about what universities can do that corporations can't, and how CMU has managed to pick up the pieces of what Uber left behind.
Motherboard: It's been over a year since you lost a chunk of your staff. How would you say the rebuilding process has gone?
Herman Herman: We've been hiring steadily since that time and right now we have about 115 people. We have a lot of new projects in addition to the ones we've been working on for a few years. It's like they say: there's no bad publicity. There are companies and people who didn't even know we exist, but now know about us. We've done projects with some of those new companies and are in active discussions with several more. The rumors about our demise are not true.
Have the projects you lost when Uber poached your staff been restored, or have you been focusing more on new initiatives?
Both. If there's a certain project where 90 percent of the team members left, then it's hard to continue that project. But we also had plenty of projects that were minimally affected—these are projects that have 10 people and two of them left. We recovered quickly and have continued them just fine, and many of our [Department of Defense] projects are in that category. But we did cancel two projects because we decided it was the right thing to do. It's not like the projects were in the middle of work and we stopped, though. We delivered what we said we were going to deliver, but we didn't continue the work past that.
What are the most exciting projects coming out of the lab right now?
Many of the newer projects we really can't talk about. In general we're pretty conservative about that and don't like to hype things that haven't been proven. You'll start hearing about them more next year. But there are some things we can talk about.
Not too long ago we finished a project on a fully autonomous six-wheeled vehicle that's carried in a sling under a fully autonomous Black Hawk helicopter. This is the first time you have a fully autonomous vehicle carried by a fully autonomous helicopter. We sent them on simulated missions 30 or 40 kilometres away and the robot was deployed autonomously from the sling to do surveys. We also continue to do a lot of work in agriculture, which is one of our main areas. We just finished two projects with John Deere on semi-autonomous robotic components to automate their equipment.
"I think part of the reason that they released the news about these autonomous taxis is marketing. Technically, I'm not sure if it's ready"
Uber just rolled out self-driving cars in Pittsburgh. Do you think things would have been different somehow if that was the result of a partnership between Uber and CMU?
Maybe, and maybe not. The objectives of companies like Uber, and our objectives as a university, are different. Uber as a company has to worry about its valuation. I think part of the reason that they released the news about these autonomous taxis is marketing. Technically, I'm not sure if it's ready. They still have to have engineers in the car. So, for technical reasons, it's not there. But for non-technical reasons, they decided that it's a good idea to start talking about it.
At CMU, our focus is technical work. Would things have gone better if there was cooperation between CMU and Uber? I'm not sure. I think that we are good at doing what we do, and Uber has to do what it thinks it has to do from a technical and non-technical perspective. Having a partnership that's forced just because someone thinks it's a good idea might actually be worse anyway.
In terms of the work that needs to be done to advance autonomous robot technology over the next few years, are there things that a university can look into that a corporation can't or won't?
Absolutely. This is a very critical point. With robotics, it's not just autonomous cars—it's A to Z. Robots to take care of the elderly, robots to make food production more efficient; there are still a lot of hard problems that require some basic solutions. That is a prime area for a university to conduct research.
With autonomous cars, you see these videos from Google and Uber showing a car driving around, but people have not taken it past 80 percent. It's one of those problems where it's easy to get to the first 80 percent, but it's incredibly difficult to solve the last 20 percent. If you have a good GPS, nicely marked roads like in California, and nice weather without snow or rain, it's actually not that hard. But guess what? To solve the real problem, for you or me to buy a car that can drive autonomously from point A to point B—it's not even close. There are fundamental problems that need to be solved.
Do you think there's room for collaboration with corporations in the future, or do you think Uber taught you a hard lesson here?
A lot of our work is funded by companies, and there are plenty of synergistic ways for companies and academics to work together. That being said, there are some growing pains here. Robotics is exploding, and the adoption curve, the tech disruption curve—these are getting steeper and steeper. I've been doing robotics for two decades, and if you look at the progress made between 1995 to 2005, in the past couple years we've seen more progress than in that whole decade.
Want more Motherboard in your life? Then sign up for our daily newsletter.
- Travis Kalanick
- motherboard show
- autonomous cars
- carnegie mellon
- herman herman