Eric Schmidt Didn’t Know That Google Was Working the Pentagon's AI Project

Eric Schmidt in New York, May 5, 2016.

Eric Schmidt in New York, May 5, 2016. Dennis Van Tine/STAR MAX

The former chairman of Google was kept in the dark about the company’s outreach to the Defense Department — by design, he says.

When more than 3,100 Google employees signed a letter in April saying that they did not want the company working on one of the Defense Department’s most important artificial intelligence initiatives, former Google leader Eric Schmidt was unaware that two the entities were even working together on that project at all. That was very much by design, the former head of Google’s parent company Alphabet said Tuesday.

“I didn’t know that we [meaning Alphabet] were doing it until I read about it in the press,” Schmidt told Defense One, after a House Armed Services Committee hearing.

Google is preparing a bid for a multi-billion dollar Pentagon cloud contract called the Joint Enterprise Defense Infrastructure, or JEDI, according to multiple sources. That could present a conflict of interest for Google and for the Defense Department, or at least create the appearance of one, since Schmidt occupies board positions with both organizations.

Schmidt currently serves as a technical adviser to Alphabet and also serves on the board of directors. He’s also the chair of the Defense Innovation Board, which advises Defense Department leaders about technology and innovation.

Google and the Defense Department agreed early on to establish a rule to separate his work for the two. That means that emails and other communications are carefully screened to avoid the rise of a conflict, said the former chairman. 

“There’s a rule: I’m not allowed to be briefed” about Google or Alphabet business as it relates to the Defense Department, Schmidt said. He’s kept separated from any aspect of Google’s or Alphabet’s business that touches on the Defense Department — and that includes the multi-billion-dollar JEDI pursuit, he said. “There’s some sort of cloud contract, which I was specifically excluded from even being aware of on both sides. They told me the name of it yesterday because it slipped out.”

That’s also an indication of how serious Google is about pursuing future work with the U.S. military. The project could be worth $10 billion over ten years.

Although Schmidt has been excluded from discussions around the cloud bidding process, the Defense Innovation Board has made recommendations about the Department’s need for a single enterprise-level solution — read that to mean: one big cloud contract. The recommendations doesn't say which company is the best suited for the job. Government officials have observed that only Amazon could reach the high-security requirement since they've already built a secret and top secret cloud for the U.S. government.

One big cloud is a key requirement for the department’s artificial intelligence ambitions, said Schmidt. That's Google's strong suit. But it's also a key point of contention within the company since many of Google's AI engineers and researchers don't want their work to be used as part of military operations (hence the letter.)

“This urgent need to address DoD’s lack of computer and storage was one of the Board’s official recommendations announced in October 2016,” he wrote in testimony he submitted to Congress Tuesday. (Schmidt here was specifically referring to recommendation #10.)

“Moving to cloud services at an enterprise level will ensure that AI efforts have a common foundation as opposed to operating in silos, which is the DoD norm. A common infrastructure allows many AI projects to grow relatively quickly in an iterative learning environment and for less cost than currently constructed,” he wrote in his testimony.

Schmidt also addressed recent revelations about how Google is attempting to reconcile employee’s concerns about writing software for the Defense Department and Google leadership’s desire to do more work with the U.S. military. Last week, Diane Greene, the head of Google Cloud, told participants of a company-wide town hall hat the company was working on a set of “ethical principles” to guide decision-making on what projects to work with the Defense Department on artificial intelligence work, meeting participants told Defense One on condition of anonymity. 

Schmidt hinted that the discussion of those principles was taking place at multiple companies, not just Google or Alphabet. “My sense is that the industry is going to come to some sort of agreement sense of AI principles. My guess is that there will be some kind of consensus among key industry players on that. That process, which will take a little while, will inform how Dr. Griffin [the current undersecretary for research and engineering] and his teams leverage, work with, work against, what have you,” he said, adding that because he was somewhat cut out of the conversation, he was speaking speculatively.