Once China’s Facial Recognition Titan, SenseTime Pivots to Generative AI

SenseTime Group, a Chinese AI company renowned for its facial recognition and computer vision technologies, has announced a strategic restructuring to focus on generative AI. The pivot aims to adapt to the rapidly evolving AI landscape and capitalize on opportunities presented by the AI 2.0 era.

The restructuring introduces a new “1+X” organizational structure. The “1” represents SenseTime’s core business, now centered around building AI models, cloud technology, and developing a general vision model for various applications. The “X” refers to a series of independent ecosystem enterprises focusing on specific industry solutions, including smart automotive (SenseAuto), home robotics (SenseRobot), smart healthcare, and smart retail.

The strategic shift is driven by several factors. SenseTime has faced challenges in recent years, with its traditional AI business segment experiencing a significant decline. The company’s “traditional AI” revenue dropped by 50.6 percent in the first half of 2024, and its stock price has fallen 61 percent since its IPO three years ago. By pivoting to generative AI, SenseTime aims to drive future growth, enhance profitability, and ensure financial stability.

As part of its transformation, SenseTime has been investing heavily in generative AI technologies. The company launched its large language model, SenseNova, in early 2023, claiming capabilities comparable to OpenAI’s latest GPT version. SenseTime has also invested in substantial computing power resources, with over 10,000 GPUs dedicated to its daily new large model development. The investment aligns with the broader AI industry trend, where companies like Nvidia are making significant profits from AI chip sales, despite geopolitical tensions affecting technology exports.

The company’s new strategy focuses on achieving seamless integration of large computing devices, foundational models, and AI applications. This approach is designed to drive model development through applications and optimize computing power through model iterations. By establishing independent CEOs for each ecosystem enterprise, SenseTime hopes to foster a more market-oriented development approach and quickly capitalize on opportunities in vertical industries.

SenseTime had initially established itself as a leader in facial recognition technology. The company’s journey began with a groundbreaking algorithm developed by co-founder Tang Xiao’ou and his students at the Chinese University of Hong Kong, which achieved a 98.52 percent accuracy rate in face recognition, surpassing human capabilities and outperforming existing algorithms at the time.

The company’s facial recognition technology saw rapid adoption and widespread application across various sectors. By 2017, SenseTime’s technology had scanned and authenticated the faces of 400 million consumers in China’s fintech sector, including 300 million for China Mobile alone. Its software was integrated into China’s ten largest security surveillance manufacturers, and the company developed face-based identity verification systems for China’s peer-to-peer microlenders, significantly reducing processing times.

SenseTime continued to improve its facial recognition capabilities, pushing the boundaries of technological advancement. By August 2017, their technology could process data from 240 points on a human face, a significant increase from the previous 106 points. The company built an internal training database of 2 billion faces and claimed access to 10 billion more images and videos, further enhancing the accuracy and capabilities of their systems.

The company’s facial recognition technology found extensive use in government and security applications. SenseTime worked with 40 Chinese localities, including major cities like Guangzhou, Shenzhen, and Chongqing. In 2016, police in Chongqing used SenseTime’s technology to identify 69 suspects and apprehend 14 fugitives within 40 days. The company also collaborated with China’s Ministry of Public Security to develop the nation’s citizen network identity infrastructure, solidifying its position as a key player in national security technology.

While SenseTime’s pivot to generative AI shows promise, with its generative AI business revenue reaching 1.051 billion yuan in the first half of 2024 (a 255.7 percent year-on-year increase), the company still faces significant challenges. The AI industry is highly competitive, and the substantial investments required for AI development continue to outpace commercial revenues in the short term.

Source: South China Morning Post

–

December 6, 2024 – by Cass Kennedy