Chinese researchers for the first time swept a competition to develop AI systems that monitor urban traffic.
What’s new: Chinese universities and companies won first and second place in all five categories of the 2021 AI City Challenge, beating hundreds of competitors from 38 nations. U.S. teams dominated the competition in its first three years, but Chinese contestants started overtaking them last year.
What happened: 305 teams entered at least one of the competition’s five tracks. All teams used the same training and testing data for each track. Here’s a summary of the challenges and winners:
- Counting the number of vehicles turning left, turning right, or going straight through an intersection. Winner: Baidu/Sun Yat-sen University.
- Tracking individual vehicles across multiple cameras. Winner: Alibaba.
- Tracking multiple vehicles across multiple cameras scattered around a city. Winner: Alibaba/University of China Academy of Sciences.
- Detecting car crashes, stalled vehicles, and other traffic anomalies. Winner: Baidu/Shenzhen Institute of Advanced Technology.
- Identifying vehicles using natural-language descriptions (a new challenge for this year’s contest). Winner: Alibaba/University of Technology Sydney/Zhejiang University.
Behind the news: Nvidia, QCraft, and several universities launched the AI City Challenge in 2017 to spur the development of smart city technology.
Why it matters: This competition is the latest example of China’s rising profile in AI. The Chinese government has funded hundreds of Smart City programs. In contrast, U.S. funding for urban AI initiatives has been limited to a few one-off grants or competitions.
We’re thinking: Smart-city technology could make urban living more pleasant and productive, yet it also carries a risk of invasive surveillance. We call on regulators and researchers who work on such projects worldwide to lead a global debate on appropriate standards of privacy and to design their systems that protect privacy from the ground up.