The Israeli military has used a new artificial intelligence (AI) system to create a list of tens of thousands of human targets for possible airstrikes in the Gaza Strip, according to a report released last week. The report comes from +972 Magazine, a nonprofit media outlet run by Israeli and Palestinian journalists.
The report cites interviews with six anonymous sources in Israeli intelligence. Officials said the system, known as “Lavender,” was used in conjunction with other AI systems to target and assassinate suspected militants, many of them in their own homes, resulting in numerous civilian casualties. claims.
According to another Guardian report based on the same sources as the +972 report, one intelligence officer said the system “facilitated” mass attacks because “machines attacked in ruthless fashion.” said.
As militaries around the world race to use AI, these reports show what its use could look like. That means machine-speed warfare with limited precision, little human oversight, and high costs for civilians.
Military AI in Gaza is not new
The Israel Defense Forces deny many of the claims in these reports. In a statement to the Guardian, it said: “We do not use artificial intelligence systems to identify terrorist operatives.” Lavender is not an AI system, but “simply a database intended to cross-reference information sources,” the company said.
But in 2021, The Jerusalem Post reported that Israel had just won its first “AI war” (an earlier conflict with Hamas), using a number of machine learning systems to sift through data and generate targets. An intelligence official said that there was. That same year, a book called The Human-Machine Team outlining a vision for AI-powered warfare was published under a pseudonym by an author recently revealed to be the head of Israel's main secret intelligence unit. .
Last year, another +972 report said Israel was also using an AI system called Habsora to identify militant buildings and facilities that could be bombed. According to the report, HabSora generates targets “almost automatically,” with one former intelligence officer describing it as a “mass assassination factory.”
In a recent +972 report, a third system called “Where's Daddy?” monitors targets identified by Lavender and, when they return home, often alerts their families and alerts the military. I claim that.
death by algorithm
Several countries are turning to algorithms for military advantage. The U.S. military's Project Maven provides AI targeting used in the Middle East and Ukraine. China is also rushing to develop AI systems to analyze data, select targets, and support decision-making.
Proponents of military AI argue that it enables faster decision-making, greater accuracy, and fewer casualties in war.
But last year, Middle East Eye reported that Israeli intelligence said it was “totally unfeasible” for humans to inspect all AI-generated targets in the Gaza Strip. Another source told +972 that they privately said “he's going to spend 20 seconds on each target” but it's just a “rubber stamp” of approval.
The Israel Defense Forces' response to the latest report states that “analysts must conduct independent inspections and ensure that the identified targets meet the relevant definitions under international law.”
In terms of accuracy, Lavender claims to automate the identification and matching process to ensure potential targets are Hamas military leaders, according to the latest +972 report. The report said Lavender made errors in “approximately 10% of cases” by relaxing its inclusion criteria to include lower-ranking employees and those with weaker evidentiary standards.
The report also claims that one Israeli intelligence officer blamed “Where's Daddy?” Under this system, targets would have their homes bombed “without hesitation as a first option,” potentially resulting in civilian casualties. The Israeli military says it “completely rejects claims regarding a policy of killing tens of thousands of people in their homes.''
Military AI rules?
As military uses of AI become more commonplace, ethical, moral, and legal concerns have largely taken a backseat. So far, there are no clear, widely accepted, or legally binding rules regarding military AI.
The United Nations has been discussing “lethal autonomous weapons systems” for more than a decade. These are devices that can make aiming and shooting decisions without human input, and are also known as “killer robots.” The past year has seen some progress.
The United Nations General Assembly has voted in favor of a new resolution ensuring that algorithms “must not have complete control over decisions related to killing.” Last October, the United States also released a declaration on the responsible military use and autonomy of AI, which has since been endorsed by 50 other countries. The first Summit on the Responsible Use of Military AI was also co-hosted by the Netherlands and South Korea last year.
Overall, international rules for the use of military AI have struggled to keep up with the enthusiasm of nations and weapons companies for high-tech AI-powered warfare.
Facing the “unknown”
Some Israeli startups producing AI-enabled products are reportedly making their use in Gaza a selling point. However, reports about the use of AI systems in Gaza suggest how far AI falls short of dreams of precision warfare, and is instead creating serious humanitarian harm.
The industrial scale at which AI systems like Lavender can generate targets also means that they can effectively “replace humans by default” in decision-making.
And by actively accepting AI suggestions with little human oversight, they widen the range of potential targets and do more damage.
set a precedent
The reports on Lavender and Habsora demonstrate what current military AI is already capable of. The future risks of military AI could be even higher.
Chinese military analyst Chen Hanhui, for example, envisions a future “battlefield singularity” where machines will make decisions and take actions faster than humans can. In this scenario, all we are left with is spectators or victims.
A study published earlier this year issued a new warning. US researchers have conducted experiments in which large-scale language models such as GPT-4 play the role of nations in wargaming exercises. The models almost inevitably became embroiled in an arms race and escalated conflicts in unpredictable ways, including the use of nuclear weapons.
The world's reaction to the current use of military AI is likely to set a precedent for future technology development and use, as seen in Gaza.
Natasha Kerner, PhD Candidate, International Studies; RMIT University
This article is republished from The Conversation under a Creative Commons license. Read the original article.
