Artificial intelligence in the Chinese automotive industry Sora – Hype or opportunity?

Related Vendors

FAULHABER Drive Systems

dataTec AG

Wuxi InfiMotion Technology Co., Ltd.

No sooner had Sora seen the light of day, than she already had fans in the Chinese automotive industry. There is a consensus: The new AI tool, which can transform texts into videos, would have great potential for autonomous driving.

Henrik Bork, long-time China correspondent for Süddeutsche Zeitung and Frankfurter Rundschau, is Managing Director at Asia Waypoint, a consulting agency specializing in China, based in Beijing.

One of those predicting a “major impact” of Sora on the automotive industry is Zhou Hongyi, founder of the cyber-security company 360. Similar to Elon Musk in the USA, he has publicly commented on Sora. Unlike Musk, however, he sees their future positively. "The original technology for autonomous driving has overemphasized the level of perception, at the expense of the cognitive level," Zhou is quoted on the auto professional portal Shouxi Zhixingguan. "Without an understanding of the world, it is really difficult to develop autonomous driving."

Training algorithms

The hope of Zhou and other industry insiders in China is that Sora will dictate a new direction for the development and application of autonomous driving technologies. While Sora is just a beginning, it already clearly demonstrates the capabilities of Large Language Models (LLM) to not only simulate the real world with moving images but also to understand it. Because the cognitive capabilities of artificial intelligence with Sora are lifted from the world of two-dimensional inputs to the world of 3D, this enables new ways of training algorithms for autonomous driving and interaction with drivers in the smart cockpit. "Sora represents a revolutionary leap in the area of AI-generated content," explains Professor Shen Yang from Beijing's Tsinghua University, who researches on the topic of AI. Sora surpasses all previous models "in terms of their spatial imagination and accuracy," says Shen.

New boost for autonomous driving in China

Sora could convince many researchers and entrepreneurs in China's automotive industry that realistic and efficient training methods for autonomous driving are possible - and at commercially realistic prices. This will give the development of autonomous driving in China a "new boost" - so the common opinion that has prevailed in China since Sora was introduced in mid-February. "The simulator will become the central engine for the iteration of the autonomous driving system," says Pan Yifeng, Deputy Chairman of PhiGent Robotics, quoted on the automotive professional portal Shouxi Zhixingguan.

Already today, video generation is an important part of the development of autonomous driving, for programmers in companies like Tesla or Wayve. If it is now possible to combine real vehicle data with new technology for generating videos, this could "essentially solve the problem of autonomous driving from one end to the other," believes He Xiang, a data expert at Haomo.ai.

Look into the future

However, Sora is still relatively simple. It will take a lot of time and work before it can truly shake up the automotive industry, warn many experts. Yet, they are also convinced that the young woman, who is seen strolling along a street in Tokyo in a roughly one-minute video, is an inspiration for how AI can be used in the future on different levels for autonomous driving.

One of these layers is scenario analysis. Sora or similar models could help in the future with the analysis of visual signals – such as gestures from traffic officers – or the analysis of complex situations, for example, unforeseen accidents in traffic, the driving behavior of ambulances, or animals on the road.

AI must learn

Until Sora or equivalent counterparts are truly capable of it, they still have a lot to learn. Currently, the model generates too many fictional scenes that cannot be used as a basis for driving decisions, where safety plays the most important role.  (se)