The steering wheel of Inceptio Technology’s self-driving truck in Jinan, Shandong province, China, Thursday, April 18, 2024.
Bloomberg | Bloomberg | Getty Images
BEIJING — While AI updates make headlines every few weeks, advances alone won’t be enough to get self-driving cars on the roads faster.
Chinese self-driving truck companies say large-scale language model improvements, from Anthropic’s Claude to China’s DeepSeek, have little impact on vehicle deployment schedules.
“The world’s best linguistics” [expert] Pony.ai CEO James Penn told reporters last week: “It doesn’t mean he’s a good driver. AI is a very broad term. They’re completely different things. Absolutely…zero relevance.”
“We all use different skills when processing language, playing sports, and driving,” he said.
Autonomous driving uses artificial intelligence to imitate a human driver using a combination of sensors, chips, and algorithms. However, the required real-world training data is very different from that which powers large-scale language models such as OpenAI’s ChatGPT, and requires a so-called world model.
Self-driving truck startup Inceptio is unaffected by broader advances in AI and remains on track for a mid-2028 commercialization milestone, CEO Julian Ma told CNBC.
He predicts that by the third or fourth quarter of 2028, Inceptio will have amassed 5 billion kilometers (3.1 billion miles) of truck driving data in China. This is enough to allow fully autonomous large trucks to drive on public roads.
According to Ma, with 5 billion kilometers of driving data collected, the AI can extrapolate it to 50 billion kilometers of global model experience. This amount would be enough for a large truck to drive completely on its own. He expects that in certain parts of the country, trucks will be able to start operating without people inside.
Reaching the goal of about two years is already quite fast, he said, noting that widespread adoption of driverless trucks will require not only technology but also partnerships with manufacturers and regulatory approvals.
Self-driving cars rely heavily on data about driving on the road. Similar to robotaxi companies, self-driving truck operators conduct manned tests to safely collect training data.
According to ARK Invest’s January Big Ideas 2026 report, Inceptio outpaces its U.S. rivals and records the most miles driven by commercial self-driving trucks in the industry. At the time, the company had driven 250 million miles, far surpassing the second-place Chinese self-driving company Pony.ai’s 4.2 million miles.
U.S.-based rivals Aurora, Kodiaq and Gatik rounded out the top five with a total of 8.9 million miles, according to the report.
Inceptio’s Ma said in late April that the company’s trucks had traveled 700 million kilometers (434.96 million miles) and was targeting 1 billion kilometers (621.4 million miles) by the end of the year. He said the company can use AI to identify which specific scenarios to focus on to collect test data.
At the Beijing Auto Show, Pony.ai also announced upgrades to its PonyWorld 2.0 AI model to improve its ability to collect specific data and train models more efficiently. The company, which already operates robotaxis in China and other countries, announced a fully driverless light truck developed in collaboration with battery giant CATL.
Regulatory challenges
Read more electric car stories
Although China has a five-year development plan that increasingly focuses on technology goals, it is often companies that take the lead in driving innovation, Ma said.
“We’ll make it happen,” he said, before regulators see the technology in action and have enough confidence to provide policy support.
But it’s clear we have a long way to go before we see trucks and cars driving around the country without drivers.
“Automotive is actually the most challenging area for AI, and to some extent even exceeds the difficulty of embodied AI because it involves safety,” Ma said. Embodied AI includes humanoid robots.
Make CNBC your preferred source on Google and never miss a moment from the most trusted names in business news.
Source link
