Google is providing TensorFlow APIs to the Pentagon for Project Maven, which aims to leverage AI to increase drone effectiveness.
Building a slide deck, pitch, or presentation? Here are the big takeaways:
- Internal emails reveal that Google is providing TensorFlow APIs to the Pentagon's Project Maven, a computer vision program for military drones.
- It's unclear what specifically Google is providing outside of APIs, and the company states that its contributions are non-offensive in nature.
The DoD is using TensorFlow to analyze drone footage in order to "extract objects from massive amounts of moving or still imagery," which Gizmodo reports angered Google employees who found out about Google's involvement throughdue to information shared in an internal mailing list.
A spokesperson for Google told our sister site ZDNet that TensorFlow is only being used to flag images for human review, and for non-offensive uses at that.
Reassurances aside, news that Google is working with the DoD to improve drone object recognition is likely to anger privacy advocates, as well as those concerned with military adoption of machine learning technology, something that Google has acknowledged.
"Military use of machine learning naturally raises valid concerns. We're actively discussing this important topic internally and with others as we continue to develop policies and safeguards around the development and use of our machine learning technologies," a Google spokesperson told ZDNet.
Project Maven is a DoD initiative that focuses on the computer vision aspect of machine learning. Computer vision is dedicated to helping a computer pick out specific objects from still photos and videos, a topic of particular interest to the Pentagon since the advent of drones for surveillance and other military applications.
SEE: IT leader's guide to deep learning (Tech Pro Research)
Google's involvement in Project Maven can be seen as unsurprising, especially considering Eric Schmidt, executive chairman of Google's parent company, Alphabet, is also the chairman of the Defense Innovation Board. The board specifically looks at military applications for technologies like machine learning, which Google is highly invested in.
In the initial announcement of Project Maven in July 2017, Marine Colonel Drew Cukor even mentioned Google by name when arguing that the United States was involved in an AI arms race, saying, "Many of you will have noted that Eric Schmidt is calling Google an AI company now, not a data company."
Gizmodo's unnamed source at Google said that many employees were outraged that Google would be contributing to a government drone surveillance project, and others stated that Project Maven raises ethical concerns about the future of AI and machine learning.
It's ultimately unclear what the scope of Google's actual involvement in Project Maven is, outside of the company's statement that it's providing TensorFlow APIs for non-offensive applications.
Whether Google's technology will be used to power self-targeting AI-controlled gunships is likely a moot point, at least for now. Cukor said that "AI will not be selecting a target [in combat] ... any time soon. What AI will do is complement the human operator."
The best we can hope for is an increase in drone efficiency that leads to less collateral damage. As for the potential worst, that remains to be seen.
- Big data in 2017: AI, machine learning, cloud, IoT, and more (free PDF) (TechRepublic)
- Eric Schmidt: AI research needs to be done in the open, not in military labs (ZDNet)
- Why AI could make the US and China the two biggest superpowers and change warfare as we know it (TechRepublic)
- Machine learning can also aid the cyber enemy: NSA research head (ZDNet)
- Mattis: Silicon Valley cyber unit, focused on AI and autonomous machines, will grow under Trump (TechRepublic)