Is There a Yellow Light for Autonomous Robotaxis?
Recently, a Tesla Cybertruck running the latest FSD 13.2 failed to recognize a thick pole standing at the edge of the road (positioned inside the roadway) and crashed into it, resulting in severe damage. This was an incredibly simple accident—one that a human driver using their own vision would never have caused.
This serious incident may cast a yellow light on the operation of robotaxis using the next version of FSD, which is set to launch after June 2025 in traditional vehicles equipped with steering wheels and brakes.
This ultra-basic accident vividly highlights the uncertainty of AI-driven autonomous driving. Today’s AI can instantly answer specialized questions in fields like law and science, easily solve complex quizzes, and even pass the University of Tokyo entrance exam. However, at the same time, even the latest AI, like ChatGPT, sometimes makes simple arithmetic errors—an unexpected weakness.
It suggests that AI training is still a work in progress. Is this just a bug? Or does AI sometimes fail at even simple tasks when it gets stuck in certain patterns? Since AI operates as a "black box," it’s impossible to determine exactly why it makes such mistakes. In a way, it's similar to the occasional blunders humans make due to misunderstandings.
That said, the fact remains that AI-driven vehicles have, on average, about ten times fewer accidents than human drivers.
Meanwhile, China’s rapid AI advancements are putting even Tesla on edge. On February 10, 2025, BYD, a major Chinese EV manufacturer, announced plans to equip nearly all its vehicle models with AI-based autonomous driving systems. This includes 21 models, such as the compact EV "Seagull," which starts at approximately 1.45 million yen.
A key aspect of this announcement is that BYD will integrate technology from AI development company DeepSeek into its advanced driver assistance system, "Tian Shen Zhi Yan" (Eye of the Gods). This upgrade is expected to enhance the precision of autonomous driving, offering a more comfortable and personalized driving experience.
"Tian Shen Zhi Yan" already includes features such as remote parking and highway autopilot navigation, which were previously limited to higher-end models. Now, this technology will be introduced in more affordable vehicles. For comparison, Tesla's equivalent feature costs at least 5.76 million yen (4.86 million yen for the vehicle plus 900,000 yen for FSD), whereas BYD plans to offer similar capabilities for less than one-third of that price—around 1.45 million yen plus additional costs.
During a live-streamed event, BYD Chairman Wang Chuanfu emphasized, "Autonomous driving is no longer a future concept; it is already a necessary technology."
Beyond BYD, Chinese tech giants such as Xiaomi, Huawei, and nearly a dozen other emerging automakers are ramping up production. Their manufacturing capacity, currently at several hundred thousand units per year, is expected to multiply by 2025–2026.
By 2025, battery electric vehicles (BEVs) are expected to become cheaper than internal combustion engine (ICE) vehicles. AI-powered autonomous driving will likely become a standard feature across all new economy automakers’ vehicles. By 2026 at the latest, this transition will be inevitable.
AI will be embedded in all manufactured products, from Tesla’s AI-equipped homes to white goods, black goods, smartphones, and beyond.
In this rapidly evolving landscape, what will happen to Japan’s legacy automakers, which still rely primarily on ICE vehicles despite calling for a "multi-pathway" strategy? What do you think? What will become of Japan's sluggish AI industry?
The years 2025–2026 will undoubtedly bring drastic change. It’s concerning… Will Japan, the Galápagos of the automotive world, be left behind?