Read about adapting arms control to military AI. Received this from a friend? SUBSCRIBE [[link removed]] CRITICAL STATE Your weekly foreign policy fix. If you read just one thing …

… read about adapting arms control to military AI.

Artificial Intelligence (AI) is an awkward grouping of technologies, conjuring images of autonomous humanoid robots while mostly referring to vast software systems that can, with training, distinguish between an apple and a hand grenade. The promise of AI for militaries is to improve the detection, recognition, and targeting of enemies. The danger inherent in such tools is that the machine makes errors that human supervisors could overlook because of their trust in the machine or because they are too distracted by the battle to double-check the work of a weapon. Writing in the Texas National Security Review, Megan Lamberth and Paul Scharre [[link removed]] outline what makes arms control for AI difficult and what might make it possible. “It is a general-purpose technology; it is an emerging technology; and verifying the compliance of any AI-related agreement will pose unique challenges. This does not mean that AI is uncontrollable, however,” the authors write. After exploring several options, from functional norms to inspections and tests, the authors suggest a focus on large-scale computing resources, or “compute,” as a viable way forward. “Restricting access to large-scale compute is a particularly attractive approach because it would work even for a somewhat ‘leaky’ regime, since prohibited actors must assemble large amounts of compute to be effective.”

virtual crimes

Over the weekend, Norway’s Amnesty International ran a campaign attempting to remind people of the harm Colombia’s police did to protesters in 2021. However, instead of using photography from the protest and police response, Amnesty Norway opted for an AI-generated image of the violence.

“The images contain hallmarks of AI generation. Besides the Colombian flag’s colors being in an incorrect order, in the same image there are plenty of unrecognizably warped human faces and features like hair, officers’ names are gibberish, and hands in the image are distorted,” reported Jason Koebler in VICE [[link removed]].

The errors alone are disqualifying, to say nothing of privileging an AI tool over real and available photographs. More troubling is the idea that a human rights organization would choose to illustrate its point with an easily provable false image. While Amnesty Norway claimed the action was taken to protect the identity of protesters still in the country, it’s, at best, an uncomfortable precedent for distorting reality in order to address real harms.

FORWARD TO A FRIEND [[link removed]] Content Moderators

Before an algorithm can tell a video of a beating of a horse from a video of a piñata, that task falls to human moderators. Even when there’s a trained AI tool to analyze such images, its work is often double-checked by the same human workers that informed its training data. These workers, often concentrated abroad and paid pittance wages by foreign tech giants, suffer from both the trauma of screening horrific uploaded videos and from an industry that sees such workers as inherently expendable.

“For too long, we, the workers powering the AI revolution, were treated as different and less than moderators,” Richard Mathenge, a former ChatGPT content moderator, told Time magazine [[link removed]]. “Our work is just as important, and it is also dangerous. We took a historic step today. The way is long, but we are determined to fight on so that people are not abused the way we were.”

Mathenge’s remarks came following a meeting of moderators [[link removed]] and AI data workers in Nairobi, who gathered to pledge the creation of an African Content Moderators Union. Such a union could protect the workers as a category against the abuses of distant tech companies or at least guarantee a fair wage for enduring them.

FORWARD TO A FRIEND [[link removed]] DEEP DIVE Rebel Reliance: Part I

Battles are a curious event, requiring two participants but whose timing is decided almost entirely by the planning and malice of one half of the equation. This is especially tricky when it comes to rebel or insurgent groups, who have extra incentives to carefully pick battles given their inherent vulnerability. One possible explanation for the timing and frequency of battles has to do with the resources a rebel group has on hand. If the money is there to fund a fight, a once-vulnerable group may become bolder, but there’s a major catch. If the money from controlling and exploiting resources directly is too good, rebels may lose interest in rebellion, instead choosing stable profit over political contestation.

In “ Rebel Resource Efficiency and the Escalation of Civil Conflict [[link removed]],” authors Bryce W. Reeder, Dongjin Kwak, John R. Smith, and Michael D. Wales examine rebel propensity toward battle as a function of rebel control of resources.

“We begin from a basic, noncontroversial premise: rebellion costs money,” the authors write. Securing, generating, and sustaining that income is difficult work, and it leads political rebels to follow paths and patterns more common among armed criminal enterprises than rival political factions. This means looking to the black market for resources, whether illegal, like narcotics, or commercial, like timber. By nature of rebel exploitation, the resources in insurgent hands pass through black markets before arriving at legitimate ones.

“Rebels holding these types of nonlootable resources often resort to the sale of booty futures,” the authors write. “A booty future is a promise to provide future access to a particular resource to an outside financier after the rebel group has achieved its political goals and overthrown the government. In exchange for future access to a major resource, such as an oil refinery, wealthy external patrons will provide funding to rebel groups in the present.”

Looking at the complex nature of present and speculative assets controlled by rebels, the authors divide their model of rebel resourcing into three stages. There’s a “vulnerability phase,” when rebels control few resources and are most at risk from state repression. This can be followed by an “emboldened phase,” when rebels with a steady income and power base turn those resources into an attack on the government. Then there’s an “exploitation phase,” where rebels focus on the profit and wealth from resources over the further pursuit of political violence.

“While many factors contribute to a group's decision to rebel against the government, this paper suggests that the ‘greed’ element plays a different role than the ‘grievance’ element in predicting insurgency,” the authors write. “Rebels accumulate resources in order to field capable fighting forces. At a high enough level of resource portfolio efficiency, greed overcomes grievance for the rebels; the perceived injustice the group initially organized may remain, but groups with highly productive resource holdings choose to pursue profits over justice, vengeance, or governance.”

Rebellion is a common-enough phenomenon that understanding the material underpinnings behind how groups chose to fight and not fight can help policymakers understand if their foes are desperate and starving, well-fed and hungry for battle, or living off the fat of the land, and perhaps open to a political settlement.

LEARN MORE [[link removed]]

FORWARD TO A FRIEND [[link removed]] SHOW US THE RECEIPTS

Raksha Kumar documented [[link removed]] the removal of Muslim figures from textbooks for grades 6 through 12 in India. The changes to the texts have been implemented while students and teachers are on vacation, but when students return, they’ll find truncated sections on the Mughal era of Indian history, the erasure of Muslim anti-colonial figures in Indian history, and narrowed definitions of democracy and dissent. “Students will also see no mention of the anti-Muslim riots of 2002, under the watch of Prime Minister Narendra Modi, when about 2,000 people (mostly Muslims) were killed in the violence,” wrote Kumar.

Ambika Vishwanath attended [[link removed]] the UN Water Conference in March, the first since 1977. “Having such an important event on a water-secure future without sufficient representation of the communities that are most affected by water stress either due to financial constraints, visa problems, or the extremely difficult deadlines on registration, will negatively impact the potential outcomes of the conference,” wrote Vishwanath, who highlighted the absence of representation from water-stressed and Indigenous communities. The work holds promise, but it will require governments to live up to their commitments and move beyond posturing. “Given the catastrophe of climate change and drought, business as usual could prove fatal,” she concluded.

Halima Gikandi spoke [[link removed]] with people fleeing the violence in Sudan. As the civil war continues with open fighting, people living in the city turned to a battlefield have made the hard call to escape however best they can. “Trying to figure out how to get from Khartoum to a location where we could have an ease of migrating was extremely difficult,” Mohamed el-Mobarak Kibeida told Gikani. Bus drivers “should be labeled as war profiteers because they’ve been doing the worst they can do,” he said. Kibeida and most of his family made it to a bus terminal to catch a ride toward Egypt, but family back home reported that shelling continued, despite a promise of a ceasefire.

FORWARD TO A FRIEND [[link removed]] WELL PLAYED

Before there was Powerpoint, Powerpoint was there [[link removed]].

Is the UK okay? Part 1: Swan song [[link removed]].

Hegemony’s just another word for favorable readings of history you get to pick and choose [[link removed]].

Is the UK okay? Part 2: Air Chief Marshal Ord Peach and the Sword of Mercy [[link removed]].

Rebranding the American Revolution as World War Twee [[link removed]].

Is the UK okay? Part 3: Between a rock and a soft place [[link removed]].

Workers of the world, unite: two tickets to the hottest movie of the summer, please [[link removed]].

FORWARD TO A FRIEND [[link removed]] Follow The World: DONATE TO THE WORLD [[link removed]] Follow Inkstick: DONATE TO INKSTICK [[link removed]]

Critical State is written by Kelsey D. Atherton with Inkstick Media.

The World is a weekday public radio show and podcast on global issues, news and insights from PRX and GBH.

With an online magazine and podcast featuring a diversity of expert voices, Inkstick Media is “foreign policy for the rest of us.”

Critical State is made possible in part by the Carnegie Corporation of New York.

Preferences [link removed] | Web Version [link removed] Unsubscribe [link removed]