快猫短视频

Russia appears to be using anti-personnel mines in Ukraine that are equipped with artificial intelligence that is claimed to be able to distinguish between soldiers and civilians. While the mines supposedly comply with international law, experts doubt whether they are any less dangerous to civilians.

Human Rights Watch, a New York-based non-governmental organisation, , on 29 March. Separately, truck launching mine-laying rockets. The vehicle has 50 rocket launch tubes, and each rocket scatters five POM-3 mines. A salvo can cover a 250,000 square metres area with mines from 15 kilometres away.

Once launched, a POM-3 mine parachutes down, then orients itself upright on six legs and pushes a seismic sensor into the ground to detect footsteps. When someone gets within 12 metres, the mine launches an explosive device upwards, which detonates at waist height, firing shrapnel horizontally in all directions.

The seismic detection system has AI said to be able to distinguish humans from animals or other moving objects. In a Russian TV interview in 2019 Sergey Bachurin, head of the design bureau which produces the POM-3, .

at the Carnegie Council for Ethics in International Affairs in New York, is highly sceptical of claims made for the mine鈥檚 AI, which echo claims made for other autonomous systems. Michel says that such systems inevitably fail when faced with confusing real-world situations and lack the contextual understanding of humans.

鈥淭hese failures will inevitably cause unintended harm, especially if the autonomous capabilities have anything to do with distinguishing civilians from military personnel,鈥� says Michel.

The mines might comply with international law, though. 鈥淩ussia is not part of the antipersonnel mine ban treaty,鈥� says at Human Rights Watch. 鈥淏ut it is party to 1996 Amended Protocol II of the 1980 UN Convention on Certain Conventional Weapons. This agreement regulates the use of remotely delivered antipersonnel mines and contains technical specifications for them.鈥�

The protocol only permits mines with a built-in self-destruct mechanism, which must ensure that 90 per cent of deployed mines are destroyed within 30 days. The POM-3 has a self-destruct designed to that operate after 8 or 24 hours but Hiznay notes that the photos seen by Human Rights Watch show mines that have failed to deploy properly, so it is unlikely that the self-destruct will work.

鈥淚鈥檓 interested in what type of power-source is used and how many redundancies are built into it,鈥� says Hiznay, but Russia hasn鈥檛 revealed this information.

While militaries generally seek to incorporate AI in weapons to increase the speed of operations or reduce operator workload, they may also claim AI makes the weapon safer with no real justification.

鈥淲eaponised AI is often framed as a means of reducing civilian harm,鈥� says Michel. 鈥淭his story points to how militaries might claim that their indiscriminate weapon has 鈥楢I鈥� in order to lend it an air of legal legitimacy. And the unfortunate thing is that people are so inclined to believe that 鈥楢I鈥� is as capable as a human in distinguishing friend from foe, the ruse might just work.鈥�

The Russian Ministry of Defence didn鈥檛 respond to a request for comment.