Until recent years, people thought autonomous cars existed only in their imagination. George Shao is an Investment and Strategy Investment Manager at Pony.ai, a level 4/5 autonomous driving tech unicorn that aims to deliver autonomous mobility everywhere by building the safest and most reliable self-driving technology. Pony.ai is currently testing its self-driving system in multiple geographies across the U.S. and China. The company was founded in late 2016 with a current valuation of $3B and its investors include Toyota, Sequoia Capital China, IDG Capital, Legend Capital, Comcast Ventures, NIO Capital, among others. The company is also collaborating with Toyota, Hyundai, Bosch, FAW, and GAC to integrate autonomous solutions in the manufacturing process.
We got in touch with George to know more about autonomous cars and how they work.
-Could you please tell our readers about yourself and your hobbies?
I was born and raised in rural areas of China, moved to Shanghai, and attended my middle and high school there. Never have I planned to study in the US until I graduated high school and completed China college entrance exam. Then I studied at Lehigh University in Pennsylvania in the US for both Computer Science and Finance. Struggled in my first semester in the US college as I never spoke English, but gradually adapted to the culture and language and even got into a social fraternity in my second year at college. Struggled again in my US internship and full-time job-hunting experience for 3 years until I landed my Investment Banking M&A job on the Wall Street. I worked in NYC IBD for 4 years before I headed back to China to work at Pony.ai (a level 4/5 autonomous driving tech unicorn) for Investment & Strategy.
Barely have I had time to have any hobbies during my IBD career in NYC, but now I am back to enjoying my hobbies. I have been lifting for 10 years and I’d often couple it with running. I enjoy chilling with my friends and know new friends at bars & restaurants – sharing each other’s fun story is quite exciting to me, and you never know who you might know along the way that might help your career or ignite interesting ideas. I am also trying to learn new stuff like wine tasting, do some easy DJing stuff, etc. as there are so many more young and fun stuff in China. I also enjoy outdoor trip – just finished 14km climbing on the Great Wall recently and am going to Dreamland (similar to Burning Man in the US with EDM and some desert outdoor activities like ATV) in Inner Mongolia during China’s 8-day long National Holiday on Oct 1st.
-Will the autonomous vehicle necessarily be electric?
Not necessarily but easier with Electric Vehicules (EV). There are gas cars that some Autonomous Vehicle (AV) tech companies are testing with. Given EV has better Drive-by-Wire system that enables the AV software system to drive the vehicle, EV will be a global trend going forward (almost all taxis, buses, ride-hailing vehicles are already EV in China, and fast charging parking lots are super common in China cities such as shopping mall, residential areas, outdoor parking lots, etc.), EV has longer lifetime and more environmental and user-interface friendly, it will most likely be EV that will be equipped with AV system.
As the whole car becomes electric, it will be easier to manufacture (don’t know the exact number of components but EV has significantly less components than gas car and thus easier to manufacture in mass production and less issues and costs associated with maintenance). It will also enable better infotainment system on EV as the whole car is redesigned to be EV, because traditional gas car’s upgrades almost likely will not be totally redesigned (too much costs and there are other complicated factors that I can’t expand there). There are hiccups like battery issues with EV sometimes, but over time I think they will be resolved in the next few years.
-What is the difference between an autonomous vehicle and a semi-autonomous one?
There are 5 levels in AV, with level 1 being cruise control, level 2 – 2.5 being Tesla’s autopilot, level 3 is a little more advanced than level 2 but still not fully autonomous, and level 4 and 5 being fully autonomous (no human driver at all with level 4 being able to drive on most roads that human will drive on and level 5 will expand in more road conditions). At this point it’s pretty self-explanatory what an autonomous vehicle and a semi-autonomous mean (level 1 2 3 being semi and level 4 5 being full autonomous).
Most car manufacturing companies (Tesla, Audi, Toyota, NIO, etc.) are doing semi-autonomous because it is easier and quicker to integrate into the current vehicles without too much disruption.
My current company Pony.ai along with more well-known ones such as Waymo and Cruise are working on level 4 and 5 autonomous technology. You can think of our level 4 and 5 tech as Windows, and they can be installed on many vehicles during the manufacturing and production phase. But the actual installing will be coupled with not just the car itself but also lots of sensor such as Lidar, Radar, camera, and processing units such as CPU and GPU.
These sensors are still expensive from a mass deployment and commercialization point of view but they are already decreasing by 30-50% in price per year or per half year, not to mention in the actual mass deployment phase (they will be very affordable if you look at how expensive PCs used to be but now they are quite common now; I know it took a long time to get PCs that cheap but since tech is developed much faster and IT related consumption iterates at much faster pace now, AV related sensors and CPUs and GPUs prices will drop at an accelerated speed as well)
-At what conception stage are we of autonomous vehicles and when are we going to see the result on the roads?
Lots of people are asking this and thought they’d see fully autonomous vehicles (level 4 and 5) in their 60s. In fact, there are defined engineering steps and timeline that AV companies agreed upon and are taking to realize commercialized AV tech. Don’t want to give a definitive answer to this and make it sound too aggressive.
Similar to vaccine to Covid-19, there are defined steps to make the actual mass-deployed vaccine, but it takes time to make sure the vaccine really works in every corner scenario. There is generally agreed timeline to tackle the Covid-19 but no scientist would say we will issue the vaccine at which date because they need to be super cautious – they would only make it public when it’s actually going to work. This is a good way to explain when we are going to see level 4 AV, we don’t want to make a big promise until it’s really going to work safely and intelligently. However, it will be sooner than the general public thought, probably in the next few years.
– What are the limits of autonomous vehicles (some of our readers are very curious about the accuracy of autonomous vehicles in handling crowded streets where people do not always respect priorities)?
As mentioned in level 4 and 5 before, level 4 will be able to drive in what human beings are mostly driving in. Probably the best impression that readers have for autonomous is from Tesla, though it’s level 2. On level 2, the current hardware setup is only using camera and not using Radar or Lidar. Camera for the most part is 2D vision, which might not be able to tell the depth of the view and some certain definitions that Lidar and Radar can catch. For example, camera only based solution might not work well in heavy rain due to water and light reflection. I use “might not” because some scientists might say they can add more cameras or use more sophisticated algorithms to filter out certain noises.
However, practical application is about “better approach”. For example, with Lidar, it can tell very minuscule objects and their depth. With Radar, it can expand car’s vision really far. With camera, it can tell the color of traffic lights and different traffic instructions on the road. There are also limitations to each, for example, Radar cannot distinguish two people side by side when they are really close, Lidar might be reflected by water or blocked by fog. Therefore, the better approach would be to fuse all sensors together, camera, Lidar, and Radar, and let AI algorithms integrate these three different perception signals to best catch everything on the road.
To summarize, my experience with our company’s AV vehicle is that it works pretty well in crowded streets where there are cars and people violating traffic rules and there are cars forcibly cut into the lane all the time. If people have been to China they will know the traffic here is so much more complicated than in the US and I have been commuting to work on our AV vehicles in rush hours pretty well and smoothly.
– Do autonomous vehicles have any positive impacts on the environment (less pollution …)? If so, what are they?
There are 3 positive impacts:
- Definite less pollution: more EVs are used in the mass deployment phase of autonomous vehicles
- Less traffic: as the AVs are rolling out, Robotaxi will be the top choice in daily commute as it will be a much cheaper solution than the current taxi or ride-hailing option given driver accounts for 80% of the Uber and Lyft and Didi expense that passengers pay (which is why these ride-hailing giants are trying to develop their AV tech as well). The reason for less traffic is that lots of ride-hailing drivers are often cruising on the road not knowing where to best find passengers, and as such also contributes to unnecessary pollution as well. Since robotaxi will be dispatched from a centralized fleet management system, the system knows exactly where and when passengers need cars and can dispatch accordingly. Second reason is that because robotaxi is so much easier and cheaper, people will tend to not own cars anymore due to complications surrounding getting a driver license, cost of driving cars vs. Robotaxi, parking expense, paying insurance, and many other inconveniences and costs of owning a car. Third reason is that many new ride-hailing drivers might not know the city as well and as such they sometimes drive on the wrong lane and miss entry and exit of some roads, which unintentionally take the passengers elsewhere for extra trip time and if many drivers do that it will create lots of traffic. With robotaxi, everything is standardized and it will know exactly how to drive the passengers to the destination without extra miles which saves lots of traffic.
- More safety: the driving behavior of robotaxi will be standardized and when lots of cars on the road are autonomous, there won’t be too much of cutting in the lane or angry drivers. In addition, because there won’t be any driver in the robotaxi, people don’t have to worry about harassment caused by unethical drivers.
A special thanks to George for his insightful answers. His inspiring journey truly is admirable. His expertise is shown in this article and the valuable information shared provide a good perspective of the autonomous vehicle industry.
If you want to share your expertise and become a contributor of Ideas Growth, get in touch with us.