Recently, the results of a pilot project for AI weapon scanning in New York City have been astonishing. The scanning devices provided by Evolv Technologies, a company, failed to detect a single gun during trials for the subway system, yet managed to find 12 knives.
Image source note: The image was generated by AI, authorized by the service provider Midjourney.
The trial, which began at the end of July, lasted only a month and was disappointing. According to CBS New York, although the device detected 12 knives, it did not lead to any arrest records, which dissatisfied Diane Akerman, a lawyer from the Legal Aid Society. She stated that this means the knives might all be legal, making it unnecessary to conduct additional searches on 118 passengers.
In fact, these 118 false alarms caused considerable distress to passengers. Many experienced unnecessary stops and searches due to the malfunctioning of these scanners. Even a CBS journalist encountered false alarms twice while passing through these scanners in 2022 and 2023, clearly raising concerns about the system's reliability.
Despite the poor performance of the project, the city government still tried to defend it. An NYPD spokesperson stated that no shootings occurred during the trial at subway stations using Evolv technology. While this is true, according to research by The New York Times, subway violence is relatively rare, typically occurring once per million rides. Moreover, this technology was deployed for only a month in 20 out of 472 subway stations, making it difficult to prove its actual deterrent effect.
The failure of this AI scanning project not only disappointed citizens but also exposed numerous issues with the application of technology in public safety. Whether the city government will continue to invest in this technology remains a topic of interest.
Key Points:
🔫 No guns were detected, but the report found 12 knives.
🛑 118 false alarms led to unnecessary stops and searches for passengers.
📉 The city government claims the technology has a deterrent effect, but there is insufficient data to support this.