Police State   
Article 3 of 416 | Back to Police State Previous | Next

San Jose uses AI to scan streets for homeless people's cars


San Francisco Gate
March 31, 2024

Area: San Francisco, Oakland, San Jose

In southern neighborhoods of San Jose, cameras attached to a city-owned car have been recording footage of streets and sidewalks. The purpose, in part, was to evaluate whether artificial intelligence software could automatically detect which parked vehicles might have homeless people living in them.

A Monday report from the Guardian outlined the pilot program: Using AI, tech companies’ cameras mounted on the car scanned vehicles for signs that they were being used as lodging.

Chelsea Palacio, the spokesperson for San Jose’s information technology department, confirmed to SFGATE that the city was using the technology and said the program had run three tests from December to February. She said the pilot was launched to “test detecting potholes, trash, graffiti, parking violations, lived-in vehicles/RVs and tent encampments.”

Though the last drive to scan for lived-in vehicles was Feb. 8, even using such surveillance as a test concerns local advocates for homeless people. Gabriela Gabrian, who was once homeless and is now the chair of the Lived Experience Advisory Board of Silicon Valley, told SFGATE that the pilot project doesn’t seem aligned with finding “humane and livable” solutions for homeless people.

Living in a car is already incredibly difficult, she said, even without invasions of privacy from a camera. “If you live in a car, there’s no safe place; you have to keep moving,” Gabrian said. “To keep moving all the time and know you’re being watched, it kind of terrorizes a person.”

After evaluating “testing data and community feedback,” San Jose will “reassess” the part of the pilot focused on detecting lived-in vehicles before sending the car out for another test, Palacio said Wednesday. She added that data associated with lived-in vehicles and RVs was deleted within 30 days of each test drive and that they haven’t yet scanned for encampments.

The Guardian reported that during the final drive and scan for lived-in vehicles Feb. 8, the software used by Australia-based SenSen.AI picked out 10 vehicles as lived-in in a two-street stretch. The outlet found a report from SenSen.AI that included photos of vehicles partially covered in tarps.

SenSen.AI CEO Subhash Challa said in an email to SFGATE that the company already knows how to classify vehicles and various curb objects and hadn’t used the San Jose data to train its algorithm. A published paper describes the company’s camera-enabled AI ambitions: Information arrives from cameras and sensors, and then algorithms take that information and trigger automated actions or notify people that work is needed.

San Jose information technology director Khaled Tawfik gave the Guardian more statistics from the companies. He said their software detected lived-in RVs with between 70% and 75% accuracy and lived-in cars with between 10% and 15% accuracy. City staff followed the camera car’s route to check the software’s success, per the Guardian. If the pilot’s tech were deployed, the city could respond to a scan of a tent by sending outreach workers before the site becomes a large encampment, Tawfik told the outlet.

Pending San Jose’s reassessment, it’s unclear if the companies will keep looking for lived-in vehicles. But much of the use of AI-powered scanning software is set to continue in San Jose’s District 10. Arjun Batra, the councilmember for the district, told SFGATE that if tech can successfully detect potholes and biowaste, the city may be able to more efficiently dispatch maintenance workers to clean up streets. He stressed that the pilot project is just a test of the AI tech’s scope and said the companies involved aren’t being paid.

However the pilot unfolds, Gabrian and other advocates for homeless people and for privacy will closely watch cities’ experiments with automatic surveillance tech. Thomas Knight, a colleague of Gabrian’s, said in a text to SFGATE, “It is beyond disturbing that the city is using AI to target the most vulnerable members of our community, as if they were potholes and graffiti.” The organization has been advocating against the adoption of a policy that would let San Jose police tow lived-in cars and RVs if they were parked near schools.

SenSen.AI CEO Subhash Challa said in an email to SFGATE that the company already knows how to classify vehicles and various curb objects and hadn’t used the San Jose data to train its algorithm. A published paper describes the company’s camera-enabled AI ambitions: Information arrives from cameras and sensors, and then algorithms take that information and trigger automated actions or notify people that work is needed.

San Jose information technology director Khaled Tawfik gave the Guardian more statistics from the companies. He said their software detected lived-in RVs with between 70% and 75% accuracy and lived-in cars with between 10% and 15% accuracy. City staff followed the camera car’s route to check the software’s success, per the Guardian. If the pilot’s tech were deployed, the city could respond to a scan of a tent by sending outreach workers before the site becomes a large encampment, Tawfik told the outlet.

Pending San Jose’s reassessment, it’s unclear if the companies will keep looking for lived-in vehicles. But much of the use of AI-powered scanning software is set to continue in San Jose’s District 10. Arjun Batra, the councilmember for the district, told SFGATE that if tech can successfully detect potholes and biowaste, the city may be able to more efficiently dispatch maintenance workers to clean up streets. He stressed that the pilot project is just a test of the AI tech’s scope and said the companies involved aren’t being paid.

However the pilot unfolds, Gabrian and other advocates for homeless people and for privacy will closely watch cities’ experiments with automatic surveillance tech. Thomas Knight, a colleague of Gabrian’s, said in a text to SFGATE, “It is beyond disturbing that the city is using AI to target the most vulnerable members of our community, as if they were potholes and graffiti.” The organization has been advocating against the adoption of a policy that would let San Jose police tow lived-in cars and RVs if they were parked near schools.

ForumShare your thoughts in the Forum