Watch Now

World’s biggest AI startup SenseTime is a Chinese surveillance company

CCTV cameras in Tiananmen Square. ( Photo: Shutterstock )

Mass surveillance and autonomous vehicles rely on some of the same technologies

An artificial intelligence (AI) startup called SenseTime Group Ltd. that performs facial recognition for China’s surveillance state has raised $600M from Alibaba and other investors at a $3B valuation, making it the most valuable AI startup in the world. SenseTime is just three years old, and has now doubled its valuation in a matter of months. 

Li Xu, SenseTime co-founder and CEO, said, “SenseTime has established an AI ecosystem anchored with robust research, deep industry collaboration, and diverse partnerships. Our Round C funding will maximize these advantages by accelerating the development of a global footprint with a larger ecosystem incorporating both domestic and overseas partners. The funding will also help us widen the scope for more industrial application of AI, thus increasing the value of SenseTime’s global ecosystem.”

“SenseTime is doing pioneering work in artificial intelligence. We are especially impressed by their R&D capabilities in deep learning and visual computing. Our business at Alibaba is already seeing tangible benefits from our investments in AI and we are committed to further investment,” said Joe Tsai, Alibaba’s executive vice chairman.

SenseTime specializes in the intersection of computer vision and deep learning—training algorithms to recognize images correctly and make predictions based on new images—and is developing applications for autonomous driving and augmented reality products, but one of their biggest sets of training data comes from state-controlled surveillance systems in China. There are over 170M CCTV cameras in China, and police officers in Henan are using smart glasses to identify people traveling by train and plane in Zhengzhou, capital city of Henan (population 9.4M).

As reported by Bloomberg, “if you’ve ever been photographed with a Chinese-made phone or walked the streets of a Chinese city, chances are your face has been digitally crunched by SenseTime software built into more than 100 million mobile devices.”

SenseTime’s ‘Viper’ service parses facial images from thousands of live camera feeds (everything from ATMs and office face-scanners to traffic cameras): eventually SenseTime wants Viper to handle 100,000 feeds simultaneously. SenseTime also claims to work with 40 separate city authorities in China. SenseTime already has hundreds of clients and partners, including Qualcomm and Nvidia, and is working with Honda Motor Co. on autonomous vehicles. 

What do mass surveillance and autonomous driving have in common? They both depend on the ability of new technology to map environments and reliably identify objects. Recognizing the face of a person who has a warrant out for their arrest seems at first glance very different from, say, distinguishing between a dog on the road and a plastic bag. But the problem is in principle the same, and face recognition and autonomous driving algorithms learn from training data in similar ways.

SenseTime’s SenseDrive product line is oriented toward autonomous driving. SenseTime has already built outward-facing warning systems for lane departure, pedestrians, and forward collisions, and the company is also working on inward-facing cameras that monitor the driver. There are two levels of products that SenseTime is working on with inward-facing cameras: a level 3 autonomy solution that can transfer control of a vehicle to an autonomous system when a driver’s attention is directed away from the windshield (the camera monitors eye and head movements), and a level 4 autonomy solution that allows a driver to control the interior cockpit displays and entertainment interfaces through gesture alone.

In the United States, technology companies like SmartDrive have developed inward-facing cameras and video analytics software to help trucking companies enhance safety—the system can wake up drivers who are falling asleep, for example. SmartDrive counts Knight Transportation, Penske Logistics, and Gordon Food Services as customers.

Stay up-to-date with the latest commentary and insights on FreightTech and the impact to the markets by subscribing.

John Paul Hampstead

John Paul conducts research on multimodal freight markets and holds a Ph.D. in English literature from the University of Michigan. Prior to building a research team at FreightWaves, JP spent two years on the editorial side covering trucking markets, freight brokerage, and M&A.