At the 2022 World Artificial Intelligence Conference (WAIC) held in Shanghai in early September, Baidu CEO Robin Li stated in his keynote speech, "L4 autonomous driving, not L3, is likely to enter the commercial operation stage after L2." The rationale of this message takes "clear attributions of responsibility for accidents" into account. In L2 vehicle autonomy, the responsibility is attributed to the driver, whereas in L4/L5 full self-driving, the responsibility is attributed to the auto factory and the distributor. In L3 autonomy, however, the attribution of responsibility is obscure. Furthermore, road traffic safety is at greater risk when the control of the vehicle changes hands between the AI system and the human driver. Baidu, one of the global leaders in self-driving technology, has spoken up. I think its argument is powerful enough to awake major automakers that are still touting L3 vehicles to the fact that the substance of L3 introduction is more about "showing off technology" than about improving road traffic safety.
Perhaps some people would say Baidu had L4/L5 fully autonomous technology in mind right from the start, so it belittles the commercial value of L3, no surprise there, but I don't think so. If we look at L2 or so-called L2.5 vehicles (L2.5 has never existed in vehicle autonomy levels. It is just a marketing gimmick automakers take advantage of), we find that car accidents occur frequently. Isn't that because drivers mistakenly hand over all the responsibility of road traffic safety to the vehicle? Likewise, while automakers are obsessed with winning the race of specific technologies for "assisted" driving, pushing vehicle autonomy to L3 without taking care of overall road traffic safety, drivers are forced to assume greater responsibility for emergencies arising from the immature technology. Such a chaos is the responsibility of automakers, but I think it is the relevant government agencies in each country that should be even more responsible.
At the end of the day, should we kill R&D for L3-related technologies simply because of the above concerns? Technologies such as Mercedes-Benz's ALKS (Automated Lane Keeping Systems) in fact benefit road traffic safety, but I think they should be categorized as L2. In other words, the driver should not depend on the carputer for the control of the vehicle. Therefore, the government must establish and execute rigid regulations and penalties, starting from the content of catalogs and marketing advertisements created by automakers. Moreover, from the technical point of view, consumers would have wrong expectations that L3 is safer than L2, and that L3 automakers can develop L4 or even L5 in the next phase. Two wrong, fatal expectations! Firstly, L3 has been proven to cause more troubles and risks to drivers. Secondly, the L3 technology is developed from L2, but L4 is not developed from L3. Otherwise, how come those automakers still cannot bring their self-driving technology to the top of the world? The technology industry's full self-driving technology and the automakers' assisted driving technology are on different paths to commercialization. The latter, coming all the way from L1, has been installed in vehicles and experienced by consumers for years, but the former is still being rigorously tested in the hope of a breakthrough soon. Once full autonomy reaches the commercialization stage (which may be divided into several stages), automakers will have to partner with technology giants and even release some of their design leadership in order to introduce L4+ fully self-driving vehicles step by step. As for L3, the commercial value will have been minimized by then (the star product will be either L4 or L2).