Is AI integration feasible in automobiles? What opportunities and transformations could it bring to the automotive industry?

Legal Perspectives on Autonomous Vehicles and AI Integration

Cautious optimism prevails in the legal aspect.

The ultimate implementation of AI in vehicles primarily focuses on user interaction.

The expansion of large AI models into various application scenarios has always been of great interest in the industry.

With the increasing significance of vehicle intelligence in recent years, the demand for AI technologies in automobiles, including autonomous driving, intelligent route planning, and cabin systems, continues to grow.

Turning back to autonomous driving.

China’s existing road traffic laws are based on human driving patterns. Current laws such as the “Road Traffic Safety Law” and its implementation regulations have almost no provision for the legal regulation of autonomous driving.

For instance, Article 8 of the “Road Traffic Safety Law” states, “The state implements a registration system for motor vehicles. Motor vehicles can only be driven on the road after registration by the public security traffic management department.” Article 19 requires “a motor vehicle driver to obtain a driver’s license according to law.”

Thus, two conditions must be met for a motor vehicle to legally drive on the road:

First, the motor vehicle must legally complete relevant registration procedures, such as obtaining traffic accident liability insurance, a vehicle travel permit, displaying a vehicle license plate, and ensuring the vehicle passes inspection.

Second, the driver must possess a driver’s license for the corresponding vehicle type.

Regarding registration procedures, there is no consensus on whether autonomous vehicles should be governed by the existing registration methods. Autonomous vehicles' inspections must consider not only the vehicle’s mechanical performance but also the functioning of software, sensors, radars, cameras, and other hardware, which current motor vehicle inspection standards do not cover.

In terms of having a driver’s license, the concept is reversed for autonomous vehicles. The “driver” gradually shifts to being a passenger, and controlling an autonomous vehicle is more aptly described as “operating” rather than “driving.”

Therefore, to legally allow autonomous vehicles on the road, it is essential to expand the motor vehicle registration system to accommodate the characteristics of autonomous vehicles and to change the legal requirement that a motor vehicle must be driven by a licensed driver.

Additionally, let’s discuss the legal liability issues concerning AI in vehicles.

Article 76 of China’s “Road Traffic Safety Law” stipulates, “In the event of a traffic accident involving a motor vehicle causing personal injury or property damage, the insurance company shall compensate within the liability limit of the motor vehicle third-party liability insurance; the insufficient part shall be borne according to the following provisions: In traffic accidents between motor vehicles, the party at fault shall bear the compensation responsibility; if both parties are at fault, they shall share responsibility proportionate to their respective faults. In accidents between a motor vehicle and a non-motorized vehicle driver or pedestrian, if the latter is not at fault, the motor vehicle party shall bear the compensation responsibility; if there is evidence of fault on the part of the non-motorized vehicle driver or pedestrian, the motor vehicle party’s compensation responsibility shall be appropriately reduced; if the motor vehicle party is not at fault, they shall bear up to ten percent of the compensation responsibility. If the accident loss is caused intentionally by the non-motorized vehicle driver or pedestrian colliding with the motor vehicle, the motor vehicle party bears no compensation responsibility.”

Thus, China’s principle of attributing responsibility in motor vehicle accidents adopts a fault liability principle between motor vehicles and combines no-fault and presumed fault responsibilities in accidents between motor vehicles and non-motorized vehicles or pedestrians.

However, as mentioned earlier, China’s current road traffic laws are based on the human driving model, and it is inappropriate to directly apply existing provisions to determine the responsible party in autonomous driving scenarios. Directly identifying the owner or actual user of an autonomous vehicle as the responsible party for traffic accidents is unfair and not conducive to the future realization and popularization of autonomous vehicle technology.

For example, in the (2022) Lu 1121 Min Chu 2301 traffic accident liability dispute case, the defendant collided with another vehicle while using adaptive cruise control. The court held that the defendant “did not pay constant attention to traffic conditions and the road environment during driving. Adaptive cruise control is an auxiliary driving function, not an automatic driving function. The accident occurred due to negligence in mistaking that the vehicle would automatically avoid obstacles, and the defendant should bear corresponding compensation responsibility.”

In this case, the court viewed assisted driving as different from autonomous driving, thus requiring the driver to maintain a duty of care for safety. However, if the case involved autonomous driving, still requiring the vehicle owner to maintain a similar level of care would be unjust.

Additionally, some opinions suggest considering autonomous vehicles as electronic robots. In case of traffic accidents, these could be handled under product defect liability or product flaw liability.

This view has merits, but treating it solely as product liability ignores the fundamental nature of autonomous vehicles as motor vehicles, and product liability does not apply to motor vehicle insurance systems.

In AI scenarios, how to better integrate motor vehicle traffic accident liability with product liability still requires further exploration.

When it comes to AI-driven vehicles, we can’t help but mention Huawei’s Yun Pan Gu autonomous car model!

From remote urban areas to mining ports, they operate day and night. Some say these are “elf-driven” cars, while others believe it’s extraterrestrial technology. What secrets lie behind this? Huawei’s Yun Pan Gu autonomous car model takes you into the world of autonomous driving with its mysterious “old driver”!

The Significance of AI Large Models in Automotive Industry

AI large models are not just a hype when it comes to automotive applications. Tesla’s FSD (Full Self-Driving) V12 version is trained using AI large model technology, enabling end-to-end AI autonomous driving.

In fact, intelligent cockpits and autonomous driving are two major application scenarios for AI large models, which have already become industry consensus.

However, due to the use of ChatGPT, there’s a misconception that large models primarily apply to intelligent cockpits. In reality, AI large model technology has been utilized in autonomous driving for quite some time.

Furthermore, because ChatGPT is a natural language large model, it excels in human-machine interaction. Therefore, after the integration of AI large models into vehicles, intelligent cockpits will significantly enhance capabilities such as conversational interaction, logical reasoning, strategy planning, knowledge retrieval, and may even develop emotional perception and empathy, serving as efficient AI vehicle butlers and cloud assistants.

Former President of Baidu’s Intelligent Driving Business Unit, Li Zhenyu, once summarized the significance of AI large models in reshaping the intelligent automotive industry:

Firstly, with the integration of language large models, human-vehicle interaction will shift from “command-based” to “conversational.” Secondly, new technologies like Transformer and BEV (Battery Electric Vehicle) will completely reshape the autonomous driving technology stack, improving perception capabilities and accelerating the maturity and adoption of pure visual solutions. Thirdly, future large models will become multimodal, shaping fully autonomous car robots.

Currently, it’s not just an ideal scenario; numerous automakers, including Geely’s Galaxy E8, Polestar 1, and many others, have already adopted AI large models.

Applications like ChatGPT, which are Internet killer applications, are undoubtedly crucial in the path toward automotive intelligence. Electric vehicles, as another type of Internet-connected smart terminal, need access to such killer applications.

Moreover, in the fiercely competitive new energy vehicle market, it’s not a matter of whether you’ll adopt AI large models or not, but rather, if your vehicle lacks AI large models, you may face the risk of being eliminated from the competition.

Indeed, just last night at CES (Consumer Electronics Show) in Las Vegas, Volkswagen, Mercedes-Benz, and BMW all showcased their integration of AI large models into vehicles.

This is the trend.

However, the current improvement in human-machine voice interaction due to AI integration in vehicles is just the beginning. With technological advancements, we can expect to see the emergence of more functionalities and applications.

I believe that with the advancement of technology, there will inevitably come a day when AI driving becomes feasible. After all, many things in this world have transitioned from impossible to possible.

It is certain that AI will bring opportunities and transformation to the automotive industry. I truly think AI in vehicles is excellent as it frees tired drivers.

I anticipate that there will be new advancements in autonomous driving, and the in-car devices will become even more intelligent. I look forward to the development and changes in the automotive industry.

Currently, many artificial intelligence applications in cars are essentially just bringing Xiao Ai classmates into the car, which actually doesn’t have much impact. The most useful applications are improving the driving experience, such as intelligent driving assistance, automatic alarms, road condition analysis, and route planning. We look forward to more new artificial intelligence applications in cars in the future.

AI in the Car vs. AI Autonomous Driving

Getting AI into the car and achieving AI autonomous driving are two completely different things.

Getting AI into the car is not difficult at all, not difficult in the slightest.

In my previous company’s artificial intelligence department, we implemented a speech-to-text interface from iFlytek. Whenever there was a hot topic related to AI, we would integrate this interface into our existing products. Then, we’d have the design department create a 3D character. During demonstrations, the leaders would speak to the 3D character, and the iFlytek interface would convert their speech into text input. The 3D character would then provide a voice output. In reality, the whole process was just about converting speech to text and had nothing to do with AI.

However, this trick was quite effective in fooling our leaders and the general public. They would say, “Oh, we have AI now,” or “We can have a conversation with AI.” Funding would be approved, and we could continue for another year.

But as for AI autonomous driving? Legally, that doesn’t seem feasible, right? A collision with an elderly person vs. AI tech giants—I’d love to see these two groups go head to head.

As for AI-assisted driving, didn’t it exist 15 years ago?

There are currently two popular directions for AI in vehicles: autonomous driving and intelligent car systems. I think this is a good development, especially for new drivers. It’s idealistic, but if everyone can use it together, it can avoid many conflicts and dangers. I hope this can be realized in the future, so that everyone can travel safely and smoothly. Some people find it very difficult to get a driver’s license, and some are only confident in driving when there’s an experienced driver around. With the introduction of intelligent systems, this becomes much easier.

Two Key Preconditions for Advancing Vehicle AI

Currently, vehicle AI systems are at a basic level, similar to summoning Siri on a smartphone.

If we want to make them more intelligent, I believe there are two rigid prerequisites:

1. The AI in smartphones and computers can afford to be less precise and even make mistakes, like GPT generating nonsense that you can ask it to “re-generate.” However, AI in vehicles must be more stringent because any issues could immediately jeopardize road safety. The margin for error is much lower, if any. If your smartphone freezes, you can simply restart it or take it to a repair shop. But if the AI in your car malfunctions, you might panic while driving, especially on a highway, and your foremost concern might be whether you’ll have a chance to get it fixed.

2. When AI directly or indirectly causes traffic accidents, should the responsibility solely lie with the driver, or should the vehicle manufacturers also bear some responsibility? This raises a legal and ethical dilemma that requires corresponding regulations to deepen the use of vehicle AI.

With the current trend in computational power development, this is what future cars should look like. As AI advances, smart vehicles will become more intelligent and resemble humans more closely. As the industry progresses, there may come a day when you don’t need a driver’s license to buy a car. Many of our elderly family members who don’t have driver’s licenses can also have their own cars, leading to increased car sales and a decline in the driving education industry. There will be a broader scope, and the development of personal aircraft will also take off. The world depicted in science fiction movies is on the horizon.

The Potential of AI in Car Interactions

Certainly feasible.

At present, new cars with prominent features like intelligence and electrification are reshaping people’s driving and travel experiences. Cars are no longer just means of transportation but are characterized by multiple interaction subjects, various interaction methods, numerous computational components, large data scales, spatial and social attributes, all of which are ideal scenarios for AI large models.

Compared to buttons, touchscreens, and gestures, controlling vehicle software and hardware through voice commands not only has a low learning curve but also keeps your hands on the steering wheel during operation, enhancing safety.

However…

Based on recent customer feedback over the past couple of years, the overall adoption rate is relatively low, primarily due to poor user experience. Currently, some car manufacturers have not fully mastered voice interaction technology, failing to provide users with a better intelligent experience and even causing some inconvenience.

Furthermore, AI large models themselves are a rather costly project. Although the “hundred-model war” has heated up the market, some investors remain cautious, believing that no more than two companies will ultimately succeed. As for the integration of AI large models into cars, although some progress has been made, achieving deep integration involves many uncertainties, including algorithms, computational power, data support, safety, legal regulations, and more. There is still a long way to go.

With the continuous development of artificial intelligence technology, improvements in algorithms and computing power have made AI more accurate and efficient in handling complex driving environments, perception, and decision-making.

In the past, when watching science fiction films, there was often talk of “man and machine becoming one.” I believe that when AI takes the wheel, it will likely gradually evolve towards the integration of humans and vehicles, especially in providing significant assistance to novice drivers.

The emergence of artificial intelligence has brought significant changes and improvements to our lives.

Simultaneously, apart from its impact on our daily lives, artificial intelligence has also made substantial contributions to the field of transportation.

On the one hand, artificial intelligence has greatly advanced autonomous vehicles, making self-driving cars a reality. On the other hand, it has also transformed the design of automobiles.

Artificial intelligence is creating entirely new modes of transportation for us, and soon, it will influence the manufacturing of cars, potentially altering our current transportation models.

The AI intelligence, I believe, may become a reality in the future. This requires time to prove, just like how our predecessors once thought reaching the moon was incredibly difficult but achieved it. Implementing AI in vehicles should be relatively easier, and I think it places significant pressure on the traditional automotive industry. It’s not easy; we must always adapt to the market’s evolution. Perhaps in the future, our descendants will all be living in an era of intelligence. Every change requires a revolution and continuous development.

The answers above were all generated by AI, right?

AI in the Automotive Industry: Opportunities and Transformations

AI integration into the automotive industry has the potential to bring about the following opportunities and transformations:

Design and Manufacturing Process: Traditional automotive manufacturers require substantial research and development investments and human resources in their design and production processes. The introduction of AI large models will significantly reduce these costs. With AI large models, car manufacturers can expedite the development of new products, shorten the R&D cycle, and enhance production efficiency. Furthermore, AI large models can assist car manufacturers in better analyzing user needs, optimizing product design, and improving user experiences.

Safety: As competition in the automotive industry intensifies, car manufacturers are focusing on enhancing the safety performance of vehicles. Integrating AI large models into vehicles will contribute to higher safety levels. For instance, through technologies like machine vision and sensors, AI can monitor and analyze data during the production process in real-time, enabling timely detection and resolution of issues, thus reducing waste and losses during production.

Usage: Traditional car operation primarily relies on manual control. The application of AI technologies can enable intelligent vehicle operation through features like voice-controlled music, navigation, and climate control, enhancing convenience and safety while driving.

Business Models: Traditional car sales primarily occur through offline channels. The application of AI technologies allows for online vehicle sales and services through the internet and mobile devices. This model not only improves sales efficiency and service quality but also reduces sales costs and risks.

In summary, AI integration into the automotive industry promises numerous opportunities and transformations, potentially driving the industry towards greater intelligence.

With the advancement of technology, it should be feasible, what used to be only seen in science fiction movies can now be truly experienced in the future. Looking forward to the transformation of the automotive industry!

It works because many things are now highly technologized. I saw an autonomous car driving on the road before, and it felt quite novel. The entire journey was driverless, operating in autonomous driving mode. Technology has made significant advancements, and it will continue to improve in the future.