The drone strikes of remote warfare have already changed how Americans think about the wars being carried out in their name (if they think about them at all). On election day 2018, for instance, journalist Ezra Klein made the claim that "America isn't at war." Although he was subsequently dragged on Twitter for the statement, it points to a bigger problem: America's wars have become so invisible that even people who theoretically might report on them seem to be forgetting about them.
Across the Greater Middle East and Africa, bombings, raids, missile attacks, and drone strikes have only risen, while Americans here at home are left with the impression that, as President Trump put it, "our boys, our young women, our men" are coming home and the endless wars are finally ending. In fact, all of the plans that purport to essentially draw down, if not end, the global war on terror have resulted so far in anything but that. And given the Trump administration's new national security and artificial intelligence strategies, this is likely to be just the start of a process that will make it easier for Washington to enter a war that no one here really notices.
Pentagon officials recently made it clear that they want to expand the use of artificial intelligence in warfare in order to maintain America's "strategic position and prevail on future battlefields." Unfortunately, it's not just future battlefields that the public needs to worry about. The U.S. military has already begun making its existing weapons systems more autonomous. This process is only likely to accelerate (in a largely undetectable fashion) thanks to the lack of transparency surrounding the development and application of AI, as well as the fact that private companies, with no commitment to public accountability, will be deeply involved in creating the technology.
Military Tech: The Ultimate Black Box
In popular culture, conversations about the dangers of advanced military technology tend to revolve around killer robots and other dystopian scenarios straight out of movies like The Terminator series, iRobot, and more recently Netflix's Black Mirror. For those already in Washington's foreign policy "blob," however, discussions are more focused on just one terrifying possibility: that America's enemies will harness this technology first. Though no one in Washington is likely to deny that killer robots could pose a genuine global threat, the focus on them and especially on what the Chinese and Russians might do with them ignores all the intermediate technologies already being operationalized that are likely to lead the world into the eye of the storm.
In fact, the Pentagon's Defense Innovation Board points out the obvious -- that "the impact of AI and machine learning (ML) will be felt in every corner of the Department's operations." Indeed, the military is already using artificial intelligence on the battlefield, most notoriously through "Project Maven." Since early last year, Maven has already been deployed to a half-dozen locations in the Middle East and Africa. It employs artificial intelligence to comb through drone surveillance footage and select targets for future drone strikes in what is advertised as a faster, easier, more efficient manner.
Maven quickly gained notoriety because the employees of Google, which the Department of Defense had partnered with to produce it, went public with a petition expressing their outrage over the company's decision to be in the business of war. Many of its top AI researchers, in particular, are worried that the contract with the Pentagon would only prove to be a first step in the development and use of such nascent technologies in advanced weapons systems. That petition was signed by almost 4,000 Google employees and ultimately resulted in the company (mostly) not renewing the contract.
That, however, hasn't stopped other companies from competing for roles in Project Maven. In March 2019, in fact, a startup founded by Palmer Luckey, a 26-year-old ardent supporter of Donald Trump, became the latest tech company to quietly win a Pentagon contract as part of the project.
The concerns of those Google employees are made all the more urgent by the inherent limitations of complex AI systems, which leave them effectively immune to human oversight. Image recognition systems like Maven make mistakes. They can misidentify images in ways that baffle even their programmers. They are also potentially vulnerable to adversarial attacks by hackers capable of manipulating images in ways the humans receiving the AI system's output are unlikely to detect.
In reality, complex machine learning still remains a remarkably opaque process even for AI experts. In 2017, for instance, a study by JASON researchers -- an independent group of scientific advisers to the federal government (whose contract was reportedly recently terminated) -- found testing to ensure that AI systems behave in predictable ways in all scenarios may not currently be feasible. They cautioned about the potential consequences, when it comes to accountability, if such systems are nonetheless incorporated into lethal weapons.
Such warnings have not, however, stopped the Pentagon from moving full steam ahead in developing and operationalizing artificial intelligence systems, while the AI arms race heats up around the world. As AI is increasingly integrated into the Pentagon's everyday operations, the lack of safeguards will only become more glaring.
New tech companies large and small are already winning military contracts that previously went only to old guard contractors like IBM and Oracle. For instance, Palantir Technologies, a Silicon Valley software company founded by billionaire investor (and sometimes adviser to President Trump) Peter Thiel, recently beat out Raytheon for a large Army intelligence contract, making it the first Silicon Valley company to win a major one from the Pentagon. The intelligence community has also been integrating AI into its operations for years, thanks in large part to a 2013 cloud-computing contract with Amazon. Very little has, however, been made public about how the new technologies are being used by the CIA and other intelligence agencies.
Risking Forever War
Perhaps more dangerous than AI's technological vulnerabilities is the detrimental impact it could have on the accountability of a Washington now eternally at war. The secrecy surrounding the military and national security uses of AI is already profound and growing. Aside from already vastly overclassifying information of all sorts, especially as it relates to artificial intelligence, the use of private companies to do much AI developmental work for the country's national security agencies will only add an additional layer of secrecy. After all, private government contractors are not subject to the same transparency requirements as government agencies. They are not, for example, subject to the Freedom of Information Act.
Of the 5,000 pages documenting Google's work on Project Maven, for instance, the Pentagon refused to release a single page, claiming that every last one of them constituted "critical infrastructure information." If federal agencies are going to claim that all information on AI is too sensitive to be released to the public, then we are entering a new age of government secrecy that will surpass what we've already seen.
This is increasingly true abroad as well. The Pentagon, for instance, brought Project Maven to the battlefield in Iraq and Syria in 2018. Shortly thereafter, the U.S. military suddenly stopped releasing any information about airstrikes in those countries, leaving the public here largely ignorant about the surge in civilian casualties that followed. What role, if any, did artificial intelligence play in those killings? Who's at fault if a computer misidentifies a target (and no one ever knows)? The national security state is clearly not interested in providing answers to such questions.
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
