DARPA Needs Cylons—Just not from Google.
As I was asking last week, how ever will it build its Cylons (er, rescue robots) if Google keeps buying all the promising suppliers? Boston Dynamics and Tokyo’s SCHAFT looked very promising, so Google hoovered them up. Suppose Houston’s TRAC Labs looks good in the next round—will Google simply pay whatever is necessary to get that outfit too? It’s not unimaginable. So, to salvage what it can of a military robotics industry, what is DARPA to do? few extreme but possible tactics come to mind: none of the options are all that appealing, but the government could attempt
…to steal Google’s designs. Google’s cyber defenses are probably formidable, but long-term security is a challenging prospect, particularly against the vaunted NSA. The legality of such action outside wartime expediency would be dubious, of course, unless somehow justified by the DPA (see below). More importantly, though, whatever book knowledge the government might steal would come without the production knowhow of actual staff. After all, Chinese Army hackers can steal all the blueprints they want; they still can’t figure out how to build a decent jet engine over there.
…to legally require Google to build its robot army. By invoking the Defense Production Act (DPA) of 1950—which former Senator Phil Gramm once called “the most powerful and potentially dangerous American law”—the president or his designee (the defense secretary) could order Google to “allocate materials, services, and facilities… as he shall deem necessary or appropriate to promote the national defense.” The DPA has been rarely invoked over the past sixty years, if perhaps notably for MRAP subcontractors in 2007, during the California electrical crisis of 2000, and on a West Wing episode in 1999. Apart from the silliness of this idea, the trouble is that robotics research is not a commodity like steel armor, kilowatt-hours, or even long-haul trucking. That which isn’t readily described isn’t readily expropriated: it’s one thing to attempt to order up scientific advances with bags of money, but trying to summon them up through executive orders is a non-starter. The whole thing would be a barely friendlier version of grabbing German scientists out of Peenemünde: what geek would go to work at Google Robots after that kind of mandate were put in place?
…to threaten Google’s business elsewhere unless it cooperates in defense. This is playing hardball, but as my colleague Jason Thelen reminds me, it’s a not-uncommon federal tactic. Just this week, Vladimir Putin accused Washington of attempting to blackmail Paris into not delivering those two helicopter carriers to the Russian Navy, in exchange for a reduced fine against BNP Paribas for dealing in dollars with Cuba, Iran, and the Sudan. If that had been attempted, it was singularly unsuccessful. However, it’s easy to imagine greater regulatory or tax problems for Google if it didn’t cooperate with the Defense Department in a suitable way.
…to outbid Google with its research budget. Google may have cosmic amounts of money, but the US government has the printing press. In theory, it can bid what it wants. In reality, though, the budget is not expanding, so DARPA can only offer a few million here and there as contest prizes or in research contracts. As Davida Isaacs of the University of Kentucky was recently reminding me, few entrepreneurs go into business specifically to be defense contractors. Between its monopsonistic buying power and overarching legal authority, the customer-and-regulator will never pay enough to justify banking on government contracts for the big payoff.
…to bring an antitrust action against Google. While this may seem straightforward, the state-of-the-art for robotics is so rapidly advancing that the market may be hard to define. If the market is hard to define, the motions and counter-motions in an ex ante action could go on for some time. And catching Google ahead of time might not be so easy: few of its transactions in this field will be large enough to trigger a Hart-Scott-Rodino filing, and as Google is an American firm, none will trigger a CFIUS filing.
In contrast, if Google is unwilling to play along, I imagine that the best course of action will be simply to reinforce its own contractors’ robotics businesses with more funding. DARPA may not have many millions to spend on robotic mules and stair-climbing androids, but the Air Force has already spent a few billion on drones. Google has its self-driving car program, but Oshkosh Defense and Lockheed Martin have theirs too. Google at best has a monopoly on advertiser-supported Web searches; it has no monopoly on good ideas. Indeed, for military applications, it is the military that has that monopsony in the markets for some good ideas—those that involve heavy weaponry. If Google doesn’t want to build warbots, someone somewhere does, and within a reliable organization, can be reasonably invented to make that happen.
James Hasik is a non-resident senior fellow at the Brent Scowcroft Center on International Security.