As a founder of $ 5 billion, the autonomous taxi company Pony.ai, James Peng is a leader of the Autonomous Driving Revolution. Trained in Silicon Valley in 2016, Pony.a initially focused on the Peng’s internal market in China, but has since developed in Europe, in East Asia and the Middle East. Pony.a is currently improving its technology fueled by AI with large -scale deployment, targeting 2025 as “mass production year”, in order to build a fleet greater than 1,000 vehicles by the end of the year. In May, the company signed a strategic partnership to offer its driverless taxis via Uber in the Middle East, in order to set up deployments via the application on additional international markets in the future.
A graduate of Stanford who cut his teeth to Google and Baidu, Peng is optimistic about the Pony’s position. In August 2024, Chinese public security authorities issued 16,000 test licenses for autonomous vehicles and 20,000 miles of roads had been opened for tests.
“Strategically, we certainly have the ambition to globalize,” explains Peng, which is included on this year’s Time100 list. “Because mobility needs are everywhere. The use of technology to have a positive societal impact should be our ambition. ”
This interview was condensed and published for more clarity.
You recently unveiled your seventh generation robot taxi at the Shanghai automobile show. What improvements and advantages does it bring?
It has many advantages compared to the latest generation, but I think that the most important are all the sensors, and the computer systems will all be autonomous, which means that they will have a much higher quality and will last longer. And, for example, the lidars we use are already widely used for the [Level 2] Systems, we can therefore share the supply chain and the ecosystem. As they are widely used components, the cost drops considerably. I think that compared to the latest generation, the overall cost fell 70%.
Why do you take your time to develop the business?
One of the main reasons why we have not yet increased aggressively was that the existing model has always lost money. The new generation, according to our forecasts, will have a margin of positive contribution, which means that the life cycle of the vehicle should have a positive gross margin. This means that when we add a vehicle at least, it does not lose money. This is one of the main criteria we use for a large scale. I see that we think that in this new generation, we should be able to achieve it.
Pony.ai was founded in Silicon Valley, but you focus on expansion in cities in China. What is the justification behind this decision?
One of the main reasons is that the Chinese market itself, in terms of mobility, is already the number one in the world. For the journey, including taxis, I think that the daily journeys of China represent almost 40% of the world total. Another partial reason is due to my inheritance, so I want to contribute to China.
In which other markets do you seek to develop afterwards?
We already have efforts in South Korea, Singapore; Last year, we established a presence in Luxembourg, thanks to the support of the local government. And we also look at the Middle East.
How does the regulatory framework differ in China and the United States? Is it easier to obtain official approval to test and deploy in China?
Each country has permanently advantages and disadvantages. It has a lot to do with customs and legal systems. In most countries of the East, and probably even in Europe a little, the regulations are more complete. It’s step by step. So, at a high level, it is much more difficult to obtain a license to do anything, including autonomous driving. But the right thing is that once you have the license, the government has more latitude to support you. In the United States, it is easy to get a license. But if you are in an accident and it is your fault, they will penalize you strongly. In China, it is a longer process but in a way that the government and private companies work more hand in hand.
If you are going to deploy thousands of robotaxis, there will be accidents. How to manage legal responsibility if a robotaxi is involved in a fatal collision?
First and foremost, he currently uses the existing regulatory framework, which means that the one who is at fault is always judged by the existing transport law. As for who is responsible, it is in fact the owner of the vehicle as the first line of response. Because in some cases, we work with our partner who has acquired the vehicle. But of course, thereafter, you can go further to say: “What group is one who is responsible?” But our design principle has always tried to avoid accidents as much as possible. And if an accident is inevitable, we actually have different ways to try to reduce the severity of accidents.
What makes your system of unique owner?
First of all, we have sensors that cover 360 degrees, we have no blind spot. It’s much better than humans [field of view]which is around 120 degrees. Another thing is that we use in -depth training to really do the AI driver or the virtual driver, much better.
When a human driver has a quasi-failure or puts himself in a difficult situation, he experiences it once and leaves it. For us, once we have collected the data, we can generate hundreds or thousands of permutations – maybe I can add another vehicle to come the other way, and put our vehicle in slightly different situations and form it many times. Thus, many scenarios can be played in large -scale simulation, such as a game engine where we play different combinations. It is therefore a much better way compared to a human driver, who can only do it once. We probably do it billions of times. This is how it can learn driving behavior much faster than typical humans.
Obviously, AI is a massive subject in China at the moment, with the success of Deepseek sending the markets that are skyrocketing. What is your assessment of AI Boom in China at the moment? Is it a real or a little exaggerated thing?
I think a little of both. Ok, it’s really real. It is here in the economy. It evolves very quickly, probably faster than what we expected several years ago. But at the same time, the global state of real intelligence is still very emerging. Because like any revolutionary technology – even the Internet, even the mobile revolution – it takes a while to mature. What you have seen, even in depth, is even more or less a potential basic model. You can see a lot of potential, but it takes time to be done and to really revolutionize our daily life.
Your business has been founded in the United States, but you focus on China. How do current trade and technological tensions between the two countries affect your business?
He has definitely [effects]. First and foremost, we cannot change anything, so we have to face the existing environment. We try to minimize influence as much as possible. For example, we must comply with the law. This means that we will not transfer any data between our Chinese office and the American offices. We must comply with all regulations. But the good news is that we have already been developing for many years. Technology has started to mature, so we can almost feel like China and the United States can operate independently, and we can continue different markets respectively.
Do the United States blocking access from China to high-end fleas such as those produced by Nvidia affect you?
We do not use these types of high -end data center fleas as [Nvidia’s] H100. Rather in the vehicle, we use autonomous quality fleas, which are already widely used and not on the list of sanctions. On the training side, we use old -fashioned chips and we use software optimization to compensate for the limitation of hardware. We are therefore not as assigned.
You were born in China and studied in Tsinghua then Stanford, then worked in Silicon Valley. But it seems that the new generation of Chinese technological entrepreneurs is more trained locally. Most Deepseek developers studied in China. Do you fear that there is not so much collaboration between China and Silicon Valley?
Personally, I have the impression of being a beneficiary of this collaboration and this open communication. The open dialog can widen the horizons for many people. So, I am quite absolutely for opening, communication and collaboration. But it is also a will not only for technology itself, but also for the formation of talent itself that China has already made a long way by catching up fairly quickly. Because I think China has a lot of raw talents, which was missing was this full ecosystem. But I think that even on this forehead, China catches up a little.
Today, many large companies – SoftBank, Alibaba, etc. – plowed hundreds of billions of dollars to reach general artificial intelligence, or acted. Do you believe that Aga is on the horizon?
It has a lot to do with your definition of AG. If the act is defined as passing the Turing test, where you cannot make the difference between a human bot and AI, it is certainly an achievable objective. We will probably be able to see [that] in five to 10 years. But if you talk about act as more TerminatorSo self -aware, where it is super intelligent, which will probably be far away.
So when does the day of judgment arrive?
Probably not for a decade or two!
