M Cube Conference in China "The era of the AI-defined car is inevitable"

Related Vendors

Asia Waypoint

NVIDIA GmbH

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

Wuxi InfiMotion Technology Co., Ltd.

Car manufacturers have only just begun to accept that the future of the automobile is defined by software. However, developments in China are already advancing this concept further: according to visionaries like Nvidia's new Head of Automotive, Wu Xinzhou, the vehicle of the future will be defined by artificial intelligence.

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

This statement might be somewhat exaggerated since artificial intelligence (AI) is, in essence, still software, but only slightly so. The era of the AI-defined car is "inevitable" and will occur "within the next five years," according to Wu Xinzhou, as reported by the Chinese tech portal Qiche Zhixin. The Nvidia executive made these remarks at the M Cube 2024 Man Machine Mobility conference during the Greater Bay Auto Show in Shenzhen earlier this month. Wu, educated Electrical engineer at Tsinghua University in Beijing and in the United States, is currently one of the most influential thinkers in the field of autonomous driving. He previously led autonomous driving projects at Qualcomm, was the head of autonomous driving at the electric vehicle startup Xpeng, and is now the Vice President of Automotive at Nvidia.

Hungry algorithms

In his much noted presentation at the M-Cube conference in the Chinese auto industry, which can now also be seen online on the Nvidia website, the Chinese expert explains why he believes the automotive industry is facing a new, decisive paradigm shift. In the era of the software-defined car, which has just begun, Wu says, massive road tests are required where all possible scenarios are captured over thousands and millions of kilometers, which then feed the corresponding algorithms.

All of this is immensely labor-intensive and yet only results in a "slow identification of corner cases". In other words, capturing all conceivable exception scenarios, which endanger the safety of drivers on the roads, is a huge hurdle for the industrialization of autonomous driving.

Nothing works anymore without the cloud

However, in the era of the AI-defined car, which is about to begin, exactly that is changing. The now available, stronger computing power of large computers, big data and generative artificial intelligence together ensure that the model described above quickly becomes obsolete, Wu explains in his lecture. "The gravitational center of autonomous driving is moving to the cloud," Wu says. The model training for autonomous vehicles will take place in the future, as well as the validation of the models. The constantly improved AV stacks then only need to be transmitted over WiFi into individual vehicles, so over-the-air.

What has changed is the ability of artificial intelligence to link data points across the dimension of space and time, explains the Chinese thought leader. With the help of LLM and VLM, this increases the ability of intelligent vehicles to understand complex situations.

From the brain of a fish to that of a human

In his talk, Wu Xinzhou compares the intelligence of current, AI-enhanced systems in the industry to the brain of a fish. The attention span is about seven seconds. However, in the era of AI-defined car driving, intelligent vehicles get "the ability to think logically over longer time horizons".

So, vehicles will be able to make their decisions more like a human being who can decide within milliseconds based on their experience to run over a plastic bottle lying on the road, instead of making a dangerous full stop.

This "end-to-end" approach is the only way forward for autonomous driving, believes the Nvidia engineer. Several speakers at the M-Cube conference in Shenzhen agreed with him, including Gu Weihao, the CEO of Haomo.ai. Similar to Nvidia, the Chinese startup is developing autonomous driving solutions based on generative AI.

Also in the background: Wayve, the British provider of autonomous driving solutions, recently completed a major financing round. The Chinese startup Xpeng recently introduced a AI-powered In-Car OS. The debate about how much AI is changing the auto industry is certainly becoming increasingly intense in China. (se)