Pentagon’s AI Surge On Track, Despite Google Protest
In the long term, large government contracts and cutting-edge projects will be hard for tech companies to resist.
Google made headlines earlier this month when it pulled out of the U.S. Defense Department’s flagship artificial intelligence program known as Project Maven, which leverages sophisticated algorithms to analyze drone footage.
Until then, the project had been so secretive that few people knew Google was involved — not even the former executive chairman of Google’s parent company, Alphabet, who now sits on the Defense Department’s Innovation Advisory Board — let alone what it actually is.
“We believe that Google should not be in the business of war,” more than 3,000 Google employees wrote in an April letter to company CEO Sundar Pichai that prompted the move.
The growing resistance from Silicon Valley to working with the government, particularly the Pentagon, raises questions about the viability of Defense Secretary James Mattis’s ambitious plans to leverage cutting-edge commercial technology for military purposes. But all indications are that the department is moving full steam ahead on AI — with or without Google.
“I’m confident that we will have access to good talent and to people that have a strong desire to help defend the United States,” said Gen. Mike Holmes, the commander of Air Combat Command, during a June 28 roundtable in Washington. AI “is a big part of our future, and you will continue to see that expanded, with Project Maven being one of the first steps.”
The controversy over Project Maven reflects a fundamental disconnect between Washington and the tech world, as the Pentagon looks to significantly ramp up its investment in key areas such as AI, hypersonics, and laser weapons. Google employees are worried that the military will use AI to kill people, but current and former officials argue that it will actually be used to save the lives of U.S. citizens, allies, and innocent civilians.
So what exactly is Project Maven? Around the world today, intelligence analysts spend their days sifting through millions of hours of video captured by drones flying over sensitive areas, looking for particular objects or threats. Commanders use this information in myriad ways, such as tracking Islamic State transport routes and scouring proposed targets to making sure a strike will not cause civilian casualties.
The idea behind Project Maven, which since its launch in April 2017 has been deployed to at least five clandestine locations in Africa and the Middle East, is to train computers to rapidly mine raw surveillance data and turn it into usable information. This not only speeds up the process of getting critical battlefield information to commanders but also frees up human operators to focus on higher-level tasks, said Lt. Col. Gary Floyd, deputy chief of Project Maven, during a May event in Washington.
“Already, we are seeing what some would call the tidal wave of data and that there is this growing disparity between the amount of intel we can collect and the amount we can process,” Floyd said. “More people is not going to be the solution. So what comes next?”
To this end, Project Maven is building a comprehensive hub of labeled data that is then used to train the sophisticated algorithms needed to autonomously identify objects from huge amounts of data. This is where Google came in — the tech giant is one of a few industry partners selected to build, test, and deploy these algorithms.
But Google is not the only company that can do this work. Its decision to pull out of Project Maven creates a market opening for other companies such as Amazon, Microsoft, and IBM — which all originally bid for the work — or even new entrants, said Michael Horowitz, a professor of political science and the associate director of Perry World House at the University of Pennsylvania.
“I think that in the short term there are other contractors ready and willing to step to the plate, and so the work on Project Maven should continue,” Horowitz said. “The risk for DoD is that something like this spirals into a broader rejection.”
And there is a lot of interest. Project Maven hosted demonstrations at the annual Trident Spectre exercise near Virginia Beach, Va., for an audience of 1,000 attendees, according to the Pentagon.
The project currently involves partnerships with Silicon Valley startups, the biggest AI and data companies in the world, and the nation’s top AI academic institutions, according to the Pentagon. ECS Federal is the primary contractor.
In the long term, lucrative Defense Department contracts — not to mention the chance to work on cutting-edge national security challenges — will be hard for tech companies to resist, Horowitz argued. In fact, Google’s new ethical AI principles leave the door open to working on military programs like Project Maven in the future, he said.
Matthew Colford, a former Obama administration official and former partner at Andreessen Horowitz, a California-based venture capital firm that has backed many defense start-ups, said the recent backlash is not an “existential threat” to the relationship between the Pentagon and Silicon Valley. The Google employees who support wholesale bans of military work are in the minority, Colford stressed, noting that the company has tens of thousands of employees, but only 3,000 signed the letter.
But the Pentagon clearly has a messaging problem.
“At the end of the day, the things their products are used for often times are not fully understood,” Colford said. “They are very easy to portray simplistically and overgeneralize that these are products to help kill women and children. The answer is much more complicated; it’s one tool in the proverbial arsenal.”
Former Deputy Defense Secretary Bob Work, a crucial player in launching Project Maven, went so far as to accuse Google of indirectly aiding China’s AI research while refusing to help the U.S. government.
“Google has a center in China, where they have a concept called civil-military fusion,” said Work, who recently joined the board of advisors of big-data firm Govini, at a June 26 event in Washington. “Anything that’s going on in that center is going to be used by the military,” he said.
Officials stress that partnering with commercial industry on AI is a national security priority, particularly as potential U.S. adversaries such as Russia and China ramp up investments in that area. China, in particular, is dedicating $150 billion to AI through 2030, Floyd said. Meanwhile, the Defense Department is spending $7.4 billion on AI, big data, and the cloud in fiscal 2017, according to Govini.
The Pentagon only just announced a plan to establish an official AI hub for the U.S. military, the Joint Artificial Intelligence Center. The department’s new research chief, Michael Griffin, is currently in discussions about the new center’s structure and mission, which is supposed to help the United States maintain its competitive edge.
Project Maven is just the beginning of the Defense Department’s foray into AI. While officials are not yet worried about a real-life Skynet — the fictional AI system in the Terminator franchise that becomes self-aware and tries to destroy humanity — Project Maven is now moving beyond full motion video exploitation to other areas.
The next step is the Defense Department’s move to a cloud-based system, which is crucial to the department’s ability to fully leverage advances in AI. This work includes several separate initiatives, like the hotly contested Joint Enterprise Defense Infrastructure program, a contract that could be worth $10 billion over the next 10 years. Amazon, Microsoft, Google, Oracle, and others are expected to bid for the work.
Amazon’s cloud storage unit is already under contract to provide a cloud-based service designed to handle classified information for U.S. intelligence agencies, the Amazon Web Services Secret Region.
As for the Google backlash, Holmes acknowledges that the military needs to do a better job of communicating its intent to Silicon Valley and the general public. He hopes to convince the public that “we are all in the business of avoiding major war.”
“Americans have expectations about what their government does and whether the government uses technology and tools to infringe upon their rights,” Holmes said. “So we are going to have to work through, as Americans, our comfort level on how technologies are used and how they are applied.”
Humza Jilani contributed to this report.