GabyJapanese police hope new AI tools will help prevent serious crimes. Officials say the technology will initially be used to protect high-ranking VIPs. The idea emerged a year after the assassination of former Prime Minister Shinzo Abe. Police are worried about new attacks by “lone criminals,” and they want to use AI to identify such individuals in advance.
Artificial intelligence against lone criminals
The AI-powered security monitoring test system will deliberately avoid facial recognition technology. Instead, it will focus on recognizing three types of machine-learning patterns. Cameras using artificial intelligence will detect suspicious behavior and objects that may pose a threat to people or property, as well as potential intrusions into restricted areas.
Officials say the system’s test launch in Japan will take place by March 2024. Some security experts say the latest AI-based cameras could boost vigilance and make police work more effective against terrorist threats. But not everyone agrees—some people worry about implementing these technologies because of hidden algorithms.
Following last year’s unexpected assassination of the prime minister and a failed attempt on a newly elected official this year, Japanese police are making every effort to prevent high-profile crimes often carried out by so-called “lone offenders.” In Japan, these individuals are also referred to as “otaku,” a term that can mean “reclusive.”
Both of these terms are used to describe lonely young people who make up a disgruntled segment of Japanese society. These individuals are not all involved in crime, but some have shown violent behavior.
The system is currently in the testing phase.
Proponents of the new method for combating individual criminals say the cameras will be able to monitor suspicious people while AI is trained to recognize genuine threats. For example, people who glance around frequently and nervously, or show other signs of a “guilty mind,” may come under closer scrutiny.
Experts say AI surveillance systems will be able to identify potentially threatening individuals in crowds or other distracting environments and detect suspicious objects. Those tasks are very complex even for experienced security personnel. Automating the process would significantly lighten specialists’ workloads.
Currently, these technologies for detecting potential offenders are still being tested. Experts say it is crucial at this stage to thoroughly assess the accuracy of the system’s performance before deciding whether to implement it officially.
Other similar technologies in the field of security
Japan is not the only country applying AI to security. Security expert Isao Itabashi says that, as of 2019, these technologies were being researched in 176 countries and were already in use in 52 of them.
Last year Japan introduced a similar AI-based system on the railway to detect suspicious activity. A Japanese startup, Vaak, also developed a humanoid robot designed to identify potential thieves based on their movements and gestures.
There are suspicions that Vaak’s software is the foundation of the latest technology used by the Japanese police. However, there is no official confirmation of this.
The Daily Mail reports that the AI Vaak is programmed to recognize suspicious activity in security footage. According to the technology’s developers, the system can distinguish between normal and potentially criminal behavior. When it detects something suspicious in video surveillance, it sends an alert through a special app.
However, researchers warn that Vaak’s theft-detection systems show racial and other biases. A 2018 study found that many popular AI systems exhibit sexist and racist tendencies, and researchers urge attention to this flaw in artificial intelligence.
Predictive systems with AI in the service of security
AI that uses data from previous reports can predict where and when new crimes are likely to occur. The database used by Palantir Technologies includes a wide range of information—from geography to criminal records and social media posts. Based on that data, the AI identifies individuals or locations that may be involved in a crime.
There are many other predictive systems that differ significantly from one another. For example, the Chicago police use a so-called “urgent list” created by an algorithm that identifies people most likely to be involved in shootings.
But many experts point to problems with that AI approach. Critics say data on past crimes is not a reliable indicator of future criminal activity.
