Was an Iranian Scientist Really Assassinated With an A.I. Weapon?

12 min readDec 13, 2020

A.I.-assisted weapons are proliferating quickly

A funeral ceremony for Mohsen Fakhrizadeh in Tehran, Iran, on November 30, 2020. Photo: Anadolu Agency/Getty Images

OneZero’s General Intelligence is a roundup of the most important artificial intelligence and facial recognition news of the week.

Inlate November, Iran’s top nuclear scientist, Mohsen Fakhrizadeh, was assassinated on a highway outside of Tehran.

Iranian military and state-owned news outlets blame Israel for the attack but also claim that Fakhrizadeh was killed by an A.I.-controlled machine gun mounted to a Nissan truck. A deputy commander of Iran’s Revolutionary Guards described the machine gun as “equipped with an intelligent satellite system which zoomed in on martyr Fakhrizadeh.” Little other information is known.

Eyewitnesses and the scientist’s family contest claims that A.I. technology had anything to with the assassination, according to the New York Times. Instead, they say, the story of an A.I.-powered boogeyman is an attempt to save face after Iran’s failure to protect one of its top scientists.

Surprising as it may be, this internet column about A.I. research doesn’t have the inside scoop as to whether international assassins used a robot. But we can shed light on how far-fetched this claim really is based on what we know about military robots.

Unlike most of the A.I. research community, eager to post its latest work in conferences and public-facing repositories like arXiv, defense contractors are notoriously secretive about their R&D projects. In the United States, these projects can be branded as national security secrets to shield themselves from public records laws. This handy loophole is used to hide how sophisticated our military systems have become — in fact, $76 billion was spent on classified defense projects in 2020 alone.

But we do know that hundreds of autonomous military systems already exist. A 2017 report from the Stockholm International Peace Research Institute (SIPRI) surveyed publicly available information to catalog 381 autonomous military systems, 175 of which were armed.

“Autonomous military system” is a vague term that encompasses everything from a self-flying drone to record intelligence footage to a robotic gun. Self-guided missiles, autonomous submarines, and automatic missile defense systems all fall into this category. The word “autonomous” is also a gray area. For instance, the U.S. military is making an “optionally manned” turret that can allegedly identify and aim at an enemy while a human pulls the trigger. On the other end of the spectrum, suicide drones armed with explosives can be equipped to find their own target.

Some of these weapons are used to guard military bases. The SIPRI report identified three stationary autonomous weapons used to guard tactical positions: a Samsung device called the SGR-A1, another device made by Israeli defense contractor Rafael called the Sentry Tech, and a third made by South Korean company DoDaam called the Super aEgis II. These automated turrets are equipped with cameras and infrared sensors that allow them to see and recognize the heat of human bodies. The Super aEgis II can allegedly detect and track human-sized targets from nearly two miles away, according to the SIPRI report.

Other autonomous military systems are being deployed as mobile weapons, as detailed in a 2019 report from Pax, a Dutch humanitarian organization. An Estonian company called Milrem has been building a kind of autonomous mini-tank called THeMIS since 2014. The THeMIS isn’t a weapon itself but a mobile robot like Boston Dynamics’ robot dog except with tank treads instead of legs. Other companies like Raytheon, Lockheed Martin, and ST Engineering build autonomous and remotely operated weapons made to be carried into battle on top of a THeMIS robot.

One of the first automated machine guns for the THeMIS was made by ST Engineering in 2016, formerly known as Singapore Technologies. The company has expanded its “remote weapon stations” to include seven kinds of weapons.

All publicly available information says that A.I.-enhanced turrets for both defense and attack can be sold with some semblance of human control, whether that be literally controlling the machine from afar or simply designating what to attack. But much less is known about which weapons have full autonomous capabilities and how those systems function.

In July 2020, Israeli company Smart Shooter unveiled a portable and autonomous weapon mounting system called Smash Hopper, which can aim and fire a gun either autonomously or controlled from a distant tablet computer. The whole thing weighs around 50 pounds, and there’s even a smaller version that folds up like a camera tripod.

The existence of these kinds of autonomous weapons doesn’t mean they were involved in the killing of Fakhrizadeh, and there’s no hard evidence to suggest the Iranian military’s claims. The truck that allegedly held the automated weapon exploded at the scene.

But the era of autonomous weapons has already begun, and countries like Britain are already sketching what an army of robots might look like in the future.

WRITTEN BY

Dave Gershgorn

Writing on AI at OneZero. Previously Qz, PopSci, and NYTimes.

335

4

Sign up for Pattern Matching

By OneZero

A newsletter that puts the week’s most compelling tech stories in context, by OneZero senior writer Will Oremus. Take a look

Get this newsletter

Emails will be sent to davidmiller58945@gmail.com.

Not you?