As Google quits controversial Project Maven, mystery deepens over role of other tech firms

By Steven Melendez

June 02, 2018

In the wake of revelations about Google’s role in a military image recognition project that led to the resignation of about a dozen employees and a petition signed by thousands more, the company has decided to end ties with the program when the current contract expires next year, Gizmodo reports.

 

Google has also reportedly pledged to unveil new principles guiding its ethical use of artificial intelligence technology.

That promise has already been met with skepticism by the Tech Workers Coalition, a group calling for Silicon Valley companies “to stay out of the business of war” and develop ethics standards for AI.

“Google management is finally recognizing that their workforce will not let this issue slide, but TWC is skeptical that internal rules will substantially alter their position in regards to military contracts,” a TWC representative said in an email to Fast Company. “They’ve already shown a propensity for hiding these contracts from their workforce, so having internal rules would not fundamentally change the decision-making structure, power we think should be in the hands of the workers there.”

It remains unclear exactly what role the tech industry continues to play in the Pentagon initiative, known as Project Maven, that seeks to use machine learning technology to quickly analyze images captured on the battlefield, such as from aerial drones. The Maven technology has reportedly been used at least since last year in the fight against ISIS. Lieutenant General John “Jack” Shanahan, the Defense Department’s director for defense intelligence for warfighter support, has said the Pentagon hopes to extend the algorithms to a wider range of drones and to other types of data, streamlining the work of human analysts.

The project was awarded $131 million in the spending bill signed by President Trump in March, and the Army has hosted an “industry day” and a demonstration event seeking to understand the capabilities of modern commercial AI technology. While Bloomberg reported that more than 100 companies attended the industry day last October, the Pentagon hasn’t publicly identified which companies are working on the project, and many companies in the industry have been closemouthed about whether or not they’re even involved with the project.

Nvidia, the chipmaker that’s seen its hardware snatched up for use in AI applications, hosted Shanahan at its GPU Technology Conference last fall, where he gave a keynote address. And the company has announced a deal with government contractor Booz Allen Hamilton to offer training on modern deep learning techniques within the federal government. But Nvidia has declined to comment on whether it’s participating in Project Maven.

 
 

“This isn’t a subject we’re discussing at this time,” a spokesperson wrote in an email to Fast Company.

DigitalGlobe, a geospatial data provider that’s been reported to provide data to the project, didn’t reply to a request for comment. Nor did Figure Eight, an AI startup formerly known as CrowdFlower, that’s been linked to the project. And while Gizmodo reported last week that IBM has been approached by the National Geospatial-Intelligence Agency to potentially work on the project, a spokesperson for IBM tells Fast Company that “IBM does not have and has not had a contract to work on Project Maven.”

Amazon, which The Intercept recently reported was claimed in leaked Google documents to host some Maven operations on its Amazon Web Services cloud environment, also didn’t reply to an inquiry from Fast Company.

Google, for its part, has said its role in the project involved its open source TensorFlow AI framework and that it was used for “non-offensive uses only,” The Guardian reported in March.

If tech workers, especially in the competitive AI arena, continue to agitate for their employers to avoid ties with the military, that could mean changes for an industry that’s long enjoyed a healthy relationship with the Pentagon. The internet itself evolved from Arpanet, funded by the Defense Advanced Research Projects Agency, and major tech companies like IBM and Microsoft have provided software to the military for decades. And cloud providers including Amazon, Microsoft, and Google are reportedly vying for a potential multibillion-dollar contract to provide services to the Defense Department.

Selling the military AI software, which could be used to help pick out targets for military strikes or even eventually to drive autonomous weapons, may be seen as more ethically fraught than renting the Pentagon cloud servers or providing software for procurement databases. Experts from Elon Musk to human rights activists have warned about the dangers of “killer robots,” and Google insiders reportedly worried about a backlash if the company was linked to AI weapons.

 

But with military interest and investment in AI continuing to rise— “The Department of Defense should not buy another weapons system without AI,” Shanahan said at the Nvidia conference last year—the relationship between the Pentagon and the tech companies and workers most accomplished in the field will likely remain controversial.

(18)