Japan launches AI software to identify thieves in advance

According to foreign media reports, a Japanese startup called Vaak claims to have developed an artificial intelligence software that analyzes surveillance videos in stores to detect suspicious behavior involving theft and identify thieves before they "take action", thereby achieving the purpose of preventing crime.

The software is called "Vaak Eye". It is said that it can use algorithms to analyze whether people captured by surveillance video have discrete movements and behaviors suspected of theft, such as looking around or fidgeting. If the algorithm determines that someone has a high probability of shoplifting, it will send an alert to store staff through a mobile phone program to prevent crime. According to the American Fun News website, this complex algorithm deeply learns more than 100,000 hours of surveillance video, from people's faces, clothing, movements, overall behavior, to macro data such as crime rates in the area where the store is located, and even weather conditions, and analyzes more than 100 different factors related to theft to make judgments.

The Japanese IT Media website reported that Vaak Eye helped detect at least seven shoplifters in 10 days during testing. Recalling the first time Vaak Eye helped detect a shoplifting attempt, Vaak founder Ryo Tanaka said, "We have taken an important step towards a society where crime can be prevented through artificial intelligence systems." Currently, Vaak Eye is being tested in dozens of stores in the Tokyo area. Vaak plans to expand it to 100,000 stores in Japan in the next three years.

The author believes that "Vaak Eye" mainly detects and judges the actions and expressions of people, which is also the future development direction of AI , from passive prevention to active prevention. However, the application of this kind of technology is also very controversial, because the more accurate the technology is, it means that it needs to collect more comprehensive and rich data for algorithm training, and this process is offensive to users who value human rights and privacy. Whether this kind of technology can be widely popularized depends on the acceptance of the user market.