Why drone surveillance is not like TCP/IP
Last week Business Insider published an interview with Vint Cerf in which he defended Google’s involvement in Project Maven [1], a US DoD/industry partnership to analyse aerial drone footage using AI technology. I am a former Google engineer who left the company this year due to my concerns about this project.
The headline for this article names Vint Cerf as the ‘father of the Internet’, but what it downplays somewhat is the fact that Vint Cerf is a VP at Google, and has been since 2005. Google pays much of its compensation for senior employees in the form of stock grants, which vest over a 4 year period. This means that Cerf almost certainly has a substantial financial interest in Google’s share price, and therefore its reputation. It’s wrong to paint Cerf as a neutral party here. Defending Google’s reputation is part of his job.
Cerf developed the TCP/IP protocol, which underlies the Internet, with funding from DARPA, the US Defense Advanced Research Projects, so its origins are military. Perhaps this has left him well-disposed to military projects, but there are many differences between TCP/IP and Maven. TCP/IP is a general-purpose protocol, as useful to industry and private citizens as it is to the military. TCP/IP, in and of itself, isn’t a surveillance technology, or a weapon — at best it can be used as a building block.
Maven is quite different — it is not general-purpose infrastructure, it is specific applications of technology in direct support of warfare. From the DoD’s memo establishing the project [2]:
“The AWCFT’s first task is to field technology to augment or automate Processing, Exploitation, and Dissemination (PED) for tactical Unmanned Aerial System (UAS) and Mid-Altitude Full-Motion Video (FMV) in support of the Defeat-ISIS campaign. This will help to reduce the human factors burden of FMV analysis, increase actionable intelligence, and enhance military decision-making.
The AWCFT will initially provide computer vision algorithms for object detection, classification, and alerts for FMV PED. Further sprints will incorporate more advanced computer vision technology. After successful sprints in support of Intelligence, Surveillance, Reconnaissance (ISR) PED, the AWCFT will prioritize the integration of similar technologies into other defense intelligence mission areas.”
Cerf states that Google’s involvement in the project was primarily to detect ‘harmful’ things, like people placing IEDs, but not to ‘automatically detect and target individuals’. It seems to me difficult to achieve the first goal without tracking individuals, who are likely to be targeted if they are considered to be placing IEDs. ‘Object detection and classification’, as mentioned in the memo above, sounds much like tracking people and vehicles.
Even if Google’s project was indeed limited to detecting IED placement, software with the ability to classify and track objects and detect patterns of behaviour required to perform that task would likely be easily modifiable to support a broader range of behaviour patterns. This brings to mind the concept of ‘pattern-of-life analysis’ [3] in which ISR is used to find suspected enemies and to target them, without knowing their identities. These ‘signature strikes’ have been called war crimes by Amnesty International [4].
Signature strikes notwithstanding, helping the US military to analyse drone footage more efficiently and cheaply is likely to lead to the deployment of more surveillance drones. Their presence negatively affects people in an area under long-term surveillance [5]. People living in drone-affected areas experience extreme stress, avoid movement outdoors, and avoid gathering in groups. Drone surveillance is inherently harmful in itself.
When Vint Cerf worked for DARPA in the 1970s he had the privilege of knowing what it was he was building, and for whom. The thousands of engineers and researchers who developed Google’s computing and artificial intelligence infrastructure over the past decades did not have the privilege of making an informed choice. This part of the interview with Cerf I agree with: Google’s leadership was not transparent.
2018 has been a landmark year for worker activism in the technology industry. I don’t know what 2019 will bring, but I hope greater transparency will be part of it. Companies should be open about the kinds of work they do, particularly in controversial areas like military work, surveillance work, and harvesting of personal information. Companies should publish their ethical standards and they should appoint external, independent boards of ethics to make sure they are enforced.
[1] Vint Cerf interview: https://www.businessinsider.com/vint-cerf-defended-googles-project-maven-defense-pilot-program-2018-12?r=US&IR=T
[2] DoD memo establishing AWCFT (Maven): https://www.govexec.com/media/gbc/docs/pdfs_edit/establishment_of_the_awcft_project_maven.pdf
[3] Targeted killing and pattern-of-life analysis: weaponised media, by Nina Franz:
https://journals.sagepub.com/doi/10.1177/0163443716673896
[4] The Obama administration may be guilty of war crimes: https://www.theguardian.com/commentisfree/2013/nov/05/obama-administration-drone-strikes-war-crimes
[5] Amnesty report on drone warfare in Pakistan: https://www.amnesty.nl/content/uploads/2016/08/will_i_be_next_us_drone_strikes_in_pakistan_091013_final.pdf?x32866