How I learned to stop worrying and love DARPA

Pando goes to the DARPA Robotics Challenge

By Dan Raile , written on June 24, 2015

From The Robotics Desk

As I arrived at the Pomona Fairplex in Southern California to attend the finals of the DARPA Robotics Challenge I was surprised to be greeted by the smiling face of director Arati Prabhakar on a flatscreen TV.

Prabhakar wanted to welcome me -- and other visitors -- and to explain her organization’s mission. She did so on a loop. DARPA, she said, makes early investments in technologies that change the world, solving national security problems, with successful commercial applications a frequent result.

I say I was surprised because I hadn’t expected DARPA to be quite so self-promotional. Growing up in the American ‘burb bloat in the ‘80s and ‘90s, I managed to remain ignorant about the work of Defense Advanced Research Projects Agency.

My adolescence coincided with the emergence of the commercial internet, and I learned the names of the brilliant businessmen and companies who had put all that information and recreation at our fingertips. But I knew nothing of the LSD-fueled architects of those technologies, their relationships with academia and the Department of Defense, the controversy stemming from those relationships amid the Vietnam anti-war movement. Nobody had quite gotten the story together to the point of teaching it in school. I was reading other stuff.

I certainly didn’t know that the products I was using were the result of a two-pronged interest in the power of computers to a) augment the human animal and b) surpass it.

I’ve since caught up a bit on my reading. And over the interceding years I’ve been fleetingly aware of certain DARPA-related projects and technologies that have chilled me to the bone, most of them involving the manipulation of mouse brains. DARPA itself had remained occult and mysterious, to me at least. And now here was its director acting like some kind of Disney greeter.

Was this a trap?

The rendered head of Prabhakar cheerfully invited me to learn more about DARPA’s history, by spending some time in the open-air corridor extending behind her screen along the pavement of the Fairplex parking lot.

Over the course of this walk through DARPA’s magic kingdom, I was presented with a history of the agency’s accomplishments, printed on tarps affixed to chain-link fences. DARPA’s impact on the development of the internet, smartphones, videogames, and self-driving cars got top billing, though mention was made of less-friendly projects like shoulder-mounted anti-tank missiles, autonomous military ground vehicles, and Predator drones. There was no mention whatsoever of more unsettling projects like the drive to program living rodents or the development and deployment of the TOR network.

“What’s with all the marketing?” I asked DARPA’s Director of Strategic Communications Rich Weiss. When I bumped into him, Weiss was leading Prabhakar to her position behind a TV sports desk from which she would deliver color commentary on the robot action throughout the weekend.

“Humanitarian aide is one of the Department of Defense’s ten core mission areas, which a lot of people don’t realize.” he said. “And secondly, we found out during the Grand Challenge that people walked away and still didn’t understand what DARPA was. So that’s why we developed the DARPA walk. Did you see it? I think it came out really well and we’ll be able to use it again.”

Unlike Battlebots, airing that evening on ABC, DARPA’s Robotics Challenge is focused on the humanitarian potential of robots. Which, of course, sets it aside from DoD’s other nine core missions, all of which have to do with missiles and cyberwarfare, surveillance and counterterrorism.

The Robotics Challenge is a successor to DARPA’s Grand Challenge -- a similar cash-prize-incentivized program from a decade ago, a contest of autonomous cars. That challenge was triggered by a 2001 congressional mandate to automate a third of the Armed Forces fleet of air and ground vehicles by 2015. That mandate will not be met, of course, but at a media briefing the day before the robot event began, Gill Pratt, the DARPA program manager behind the DRC, said that the Grand Challenge had been a success in sparking the development of autonomous vehicles, referencing several popular commercial efforts to build driverless cars.

The foremost of these, on most minds, is that of Google. In fact, many of the teams and technologies involved in the Grand Challenge have since gone to work in Google’s self-driving vehicle projects. The DARPA Director from that time, Regina Dugan, now heads Google’s Advanced Technology and Projects Group. ATAP was spun out of Motorola and is apparently focused on developing the next generation of “mobile”. Fortune called it Silicon Valley’s “coolest skunkworks.”

In a Q&A last year with the blog robohub, Pratt discussed his reaction to Google’s then-recent robotics purchasing spree.

“...what’s going on in robotics is that we’ve seen some interest in the commercial world in taking this to the next step. And they have resources far beyond what DARPA has to further develop a technology and, most importantly, to drive down cost.

[robohub] “So these big deals are a good thing for DARPA?

[Pratt] “Yes, they are a wonderful thing for DARPA and, more broadly, for our nation.”

(Pando’s Yasha Levine has written before about the close relationship between Google and the DOD. I’ll leave it to you to contemplate Pratt’s remark that Google has access to “far beyond” the resources of the US Government.)

At the media briefing in Pomona, Pratt also announced that he would be stepping away from his role as Program Manager after the finals. He couldn’t say, when asked, where he was stepping away to. If I were Hunter Thompson I’d tell you the odds of the bet I just laid on Google, though.

During the briefing, Pratt told some 200 assembled media that there were many great stories to be had at the DRC, and offered his staff’s assistance in getting access to the teams and DARPA staff.  

First, Pratt described the excruciating parameters of the DRC competition, in which competing robots must remain untethered to communications and power links, and would be forced to cope with frequent wireless communications scrambling. He explained that the insufficient disaster response at the Fukushima Daiichi nuclear plant was the spark that had ignited the DRC plan.

Back in the DARPA walk, a darkened geodesic tent displayed images and stories of the Fukushima tragedy on its interior walls, a dark, string-heavy orchestral score setting the mood.

“The idea of the government sparking innovation by offering cash prizes is not new,” Pratt continued. He then narrated a series of slides about Napoleon Bonaparte’s 1795 food preservation challenge, proving that tempered Napoleonic aspirations survive at the Department of Defense in the new century.

Then, while tens of thousands of drinking Americans gathered at a horse-racing track on the East Coast to watch American Prophet win the triple crown, at the DRC finals a smaller crowd of sober dads and families gathered to watch 23 robots attempt to navigate a simulated disaster zone on a disused horse-racing track outside of LA.

The futuristic race in California was notable for its lack of speed and adrenaline. Only three teams managed to complete the course in under an hour. The crowd erupted in cheers when a robot would spend fifteen minutes opening a door and stepping through it. They groaned when, more frequently, a robot toppled over. The robots might look like Terminators, but their performance gave little cause for immediate fear. This stuttering pace of action gave Prabhakar plenty of gaps to fill with commentary and throw to DARPA correspondents reporting from the robotics expo outside the stadium.

The teams hailed from prestigious institutions like MIT, Carnegie Mellon, Lockheed Martin and NASAs Jet Propulsion Laboratory (that same day, three billion miles away, a JPL aircraft had moved a vehicle into orbit above the sunny side of Ceres, the largest asteroid in the asteroid belt, drawing closer to the mysterious bright spots on that body’s surface. It took eight years to get there. In Pomona, it took JPL’s Robosimian 47 minutes to drive a Polaris Ranger XP 900 about 100 feet, open a door, twist a valve, flip a switch, and walk over some rubble.)

Teams also came from university programs in Germany, Italy, Japan, Hong Kong and South Korea. A Korean robot, Kaist, walked away with the $2 million grand prize.

Carnegie Mellon took home half a million bucks, having come in third. In “the garage”, a large quonset structure a quarter mile from the racetrack, the teams issued commands to their competing robots and made last-minute tweaks to their entries. In Carnegie Mellon’s quarters there, Google’s name was displayed on a list of sponsors. It was the only place at the Fairplex I saw it.

But seven teams in the competition, including IHMC Robotics which took home $1 million for second place, were using modified versions of a robot built by Boston Dynamics, the robot-maker acquired by Google in 2013. After the DRC trials in 2013, Google bought the top-performing team, Japan’s SCHAFT, and dropped it from the competition. Boston Dynamics maintained its involvement but says it has done so independently of Google, which has vowed not to take military contracts. Boston Dynamics maintained a space in the garage for servicing the teams using their products.

The greatest trick that Google ever pulled was to convince the world it wasn’t at the DARPA DRC. In fact, the company is all over the agency’s turf, hiring from its ancestral pools and outpacing its investments. Google has even appropriated to itself the top-secret, sexy-scary publicity approach. And unlike Google, DARPA has to regularly petition Congress to approve its budgets. Hey kids! If you like the robots you saw here today, call your local representative.

There are some areas where the military might maintain its decades-long lead in technology seeding, out at the bleeding edge of national security needs.

“Why isn’t there anything out there about the cortical modem?” I asked Rich Weiss.

“Because that cortical modem is not really a project. It’s just an idea,” Weiss retorted. He then excused himself to rejoin the Director and walk with her up the stairs. The escalator was broken.