SenseTime executive touts advance of AI-based facial recognition technology at EmTech China
Since outperforming human eyes for the first time in 2014, AI-based facial recognition technology has advanced to the point of negligible deviation, Xiaogang Wang, managing director of SenseTime Research Institute told an audience at the recent EmTech China conference, DIGITIMES reports.
Wang said that the advance is equivalent to advancing from four-digit passwords to eight-digit ones in four years, while delivering a keynote speech on “Now and Future under AI Empowerment” at EmTech China. The conference was hosted by MIT Technology Review in Beijing.
Deep learning is the source of outstanding achievement in rapid AI over the past decade, Wang said, driven by big data, GPU and AI chip-powered cloud computing, and constant algorithm innovation. He also noted that neural networks have advanced from 5-layer architectures to 1,200-layer ones, which have improved AI facial recognition to the point of accurately identifying individuals among hundreds or thousands of images.
Wang noted that the improved algorithms make cities safer by enabling police to identify long-wanted criminals who have changed names and used fake IDs, as well as by helping to locate missing elderly people or children. He said that SenseTime’s “face unlock” technology has been integrated with over 100 million smartphones through Qualcomm chipsets.
As previously reported, Alibaba recently became the largest backer of the Hong Kong-based unicorn with a $227 million investment.