SPEAKER_00: This Week in Startups is brought to you by EquityZen, the premier marketplace allowing private investors to access proven startups. Head to equityzen.com slash twist now to get started for free and get your minimum first investment cut in half. Walker Corporate Law, specializing in the representation of entrepreneurs. Visit walkercorporatelaw.com and Salesforce Essentials, jumpstart sales and support by leveraging the world's number one CRM at a startup price point at just $25 a month per user. Go to salesforce.com slash twist for an additional 50% off and a free onboarding call. SPEAKER_05: Hey everybody, welcome to This Week in Startups. Our topic today, self-driving cars, autonomous robo fleets. When are they coming? When will they be here? Well, my guest today, James Peng, SPEAKER_08: is the co-founder and CEO of Pony.ai, which is one of the recent members of the Unicorn Club here in California. Welcome to the program, James. Thank you, Jason. So tell me about your company, Pony.ai. What do you do? SPEAKER_11: Pony.ai, we founded it at the end of 2016. We focused on developing level four, level five autonomous driving technology and eventually running an autonomous driving fleet to provide rides to all the passengers. SPEAKER_14: Got it. So you want to kill my number one investment, Uber. Got it. Not kill. Collaborate. Collaborate. Okay. So that's defined for everybody. SPEAKER_17: You said level four and level five autonomy. For people who don't know what that is, explain what level four and level five are. SPEAKER_11: Sure. So for the vehicle, the autonomy was divided into different levels. You can think of level one is a very simple, say your cruise control. Okay. SPEAKER_22: So we have level one now. SPEAKER_11: You have level one. Level two, we have it. Like Tesla's autopilot is sort of level two. You can, in a confined environment, you can let the vehicle drive, but you supervise. SPEAKER_05: Right. So you're still have to have your hands on the wheel. You're giving multiple functions over in level two, like. SPEAKER_25: Like adaptive cruise control. SPEAKER_05: Right. Which means I, if the car in front of me slows down, I'll slow down proportionally, but try to hit my optimal speed as set in my cruise control. SPEAKER_11: You can even notify it to do a lane change. SPEAKER_05: Right. And so I use that every day. The autopilot. It's great. That's level two, but level two requires that I have the same commitment to driving as if I didn't have any self-driving techniques on the car. Right. SPEAKER_15: Exactly. Essentially it's still, the safety responsibility is still up to the driver. SPEAKER_05: The same. You cannot be on your phone texting, et cetera. Exactly. Okay. So now we get to level three. It gets interesting. SPEAKER_11: Level three. Then it's in certain situations, you can completely be off the driving wheel. SPEAKER_30: Oh. SPEAKER_11: And then you insert some other conditions than you driving. Got it. For example, in a highway, instead of like a typical autopilot, you can completely off the wheel and start looking at your phone. SPEAKER_33: Got it. So level three, we do not have in the United States yet, correct? SPEAKER_16: That's not widely used yet, but there are some prototypes and there are certain products out there already. SPEAKER_05: That are in private beta, I can't use them as a consumer. Not yet. So that's Waymo when they're driving or Uber's cars. No, SPEAKER_36: not yet. No, they're level four. SPEAKER_05: They're level four. SPEAKER_11: Yeah. SPEAKER_33: Okay. So level four is? SPEAKER_11: So level four is in most of the areas, you will be able to drive completely without a human intervention. SPEAKER_05: Got it. So level three is in some areas I can use my phone and not pay attention and self-driving. Level four, all areas. SPEAKER_15: Not all areas, most of the areas. Most of the areas. You have predefined areas. Got it. You can completely off the wheel. SPEAKER_36: Got it. Level five means everywhere. SPEAKER_41: Everywhere. SPEAKER_17: Level five means we don't even have a map. SPEAKER_42: There's no GPS of this location, but the car will know what to do. If I put it on a private road, it would just work. SPEAKER_15: Supposedly. All weather conditions, all geolocations. Okay. You're supposed to just do it. SPEAKER_05: Basically be as good as a human, an alert human. SPEAKER_45: Or better than a human. SPEAKER_05: Or better than a human, obviously, because they wouldn't make mistakes. So you're attempting level four or level five. Level four, SPEAKER_33: level three will be wonderful because people on highways will not have to text and drive. They'll be able to text or sleep. That's going to be a huge game changer. When is level three going to happen SPEAKER_17: for, let's say, Tesla's autopilot or anybody else's, Volvo's, here in California or America, do you think? SPEAKER_08: Consumers will be able to sleep on the highway or do texting? Yes. SPEAKER_36: I think with the current technology, continue to advance SPEAKER_11: and the safety level goes several notch up, probably in three to five years. Okay. We'll be able to do, to have actually consumer product, consumer vehicles that be able to provide level three out there. SPEAKER_17: Wow. So in as few as three, four or five years, Yes. on the 280, on the 101, we might be able to, hey, do your email or if you're stuck in traffic, it's not that big of a deal SPEAKER_32: because you can watch a movie. SPEAKER_16: Yes. And the vehicle actually will notify you if you say something dangerous and let you take it over. SPEAKER_08: And who is driving this? Is it the industry? Is it companies like yours, Pony AI and Tesla and Waymo? Is it the government? I think right now, SPEAKER_11: Who's driving it? I think right now, it's mostly the, the private companies that's driving, driving the, SPEAKER_15: because the thing that we do see the market out there, we do see the huge benefit and potential benefits to all the society. SPEAKER_17: And does the government agree with this framework, the level 1, 2, 3, 4, 5, did they come up with that? SPEAKER_58: Yeah, it's come up from the, from the highway authority. SPEAKER_17: Got it. And so, do you get the sense that the government is very pro this innovation happening or are they concerned? SPEAKER_05: Where do you think the government sits on that? Because we're all very enthusiastic. We believe that this will save lives ultimately. What does the government believe? Do they have a lot at risk and they're conservative or are they aggressively SPEAKER_62: want to see this future? SPEAKER_61: Well, SPEAKER_11: apparently, I don't have a complete picture yet. But it's so far from our, my personal experience, I think most, most of the government agencies that we deal with, it's, it's very cautiously optimistic. SPEAKER_65: Right. SPEAKER_11: So they, they will, they will let it out a little bit, see what happens. If they, if all the companies showed SPEAKER_36: safe track record, I think they'll open a bit more. That's certainly the case for California. SPEAKER_05: Let's take a look at a video here. You brought some video of your product in market. Now you're adding your technology to existing cars. You're not making your own Teslas like Waymo. Well, actually Waymo makes their own cars now. Are you making your own cars or you're adding your technology to cars that exist in the market? SPEAKER_11: Actually, same as Waymo. We are not making the vehicles yet. We are retrofitting SPEAKER_36: an existing vehicle by putting the sensors and also developing advanced AI algorithms so that to make the vehicles fully, SPEAKER_11: be able to handle all the traffic scenarios by itself. Got it. So this is level four. What do you use? SPEAKER_39: Prius? Volvo? SPEAKER_11: This one is Lincoln MKZ. SPEAKER_39: Lincoln. Oh, okay. So like the ones you would get picked up in an Uber Black. SPEAKER_11: Yeah. SPEAKER_39: Let's take a look. SPEAKER_11: So here we go. SPEAKER_36: So this is the level four, meaning that there's no human intervention. And then this is in the urban scenarios. SPEAKER_75: Okay. Urban scenarios, level four. SPEAKER_36: So let's take a look. Pony.ai. This is a mix of clips. So this is doing in a busy road, doing an unprotected left turn means there's no light, no left turn lights. You have to yield the incoming traffic from the other side before you can make SPEAKER_76: And this is not America because I can see the way people are driving SPEAKER_36: and the way people SPEAKER_76: are crossing. SPEAKER_16: Busy road in Beijing and you have to wait for the motorcycles and pedestrians. SPEAKER_77: The motorcycles driving the pedestrian crosswalk you're referring to. Yeah. And then- SPEAKER_17: Just take a pause here for a second. I have a question for you. In China, do people jaywalk like this still and are roads still very dangerous? Because I haven't been to mainland China in 10 years SPEAKER_05: and it was crazy when I was there. People were making left turns 100 feet, 200 feet before the intersection SPEAKER_14: to try to get a jump on the oncoming traffic. It was terrorizing. SPEAKER_11: I would definitely say it's a lot better now but we still have this once in a while. To make the car safe means we have to handle all the extreme cases. SPEAKER_81: Ah. SPEAKER_36: Regardless whether it's a once in a million or once in a ten situation, we still have to handle it. SPEAKER_08: So the driving conditions in a major city in China are how much more challenging than say San Francisco or New York? SPEAKER_86: I would say at least equally challenging. SPEAKER_17: Okay. Because it seems more challenging if people are driving motorcycles on the crosswalks. I mean, I don't even see that in New York. SPEAKER_36: San Francisco has a lot of pedestrians to deal with too. That's true too. Jaywalkers. SPEAKER_08: There are jaywalkers here. Yeah. Is part of the solution going to be that we're going to have to get pedestrians to behave differently SPEAKER_05: like put them into pens and put fences up like I see some places in Hong Kong or in some areas of New York now they've put fences so people can't jaywalk. Is that part of the solution to all this, do you think? SPEAKER_61: That definitely will increase the safety. SPEAKER_11: But as a technology player, SPEAKER_16: our job is to make it safe regardless of what's surrounding traffic out there. SPEAKER_39: All right. Let's keep watching here. SPEAKER_77: All right. Rush hour we saw. And you made that left turn very elegantly and you didn't kill any of the pedestrians or the motorcycles. Okay. Congratulations. SPEAKER_36: Thank you. So the next one we show is the case of dynamic lane change. How you can notice on the left side the blue line is the one we the path that we will take. You can see we try to make a left turn but because there's incoming traffic on the adjusting lane we opportunistically choose the place SPEAKER_16: where we make the turn. SPEAKER_42: Got it. And that was a pretty aggressive lane change there. When I'm using my autopilot on my SPEAKER_16: This is rainy days. Yeah. Rainy day here. Very heavy rain. SPEAKER_96: I mean that looks like a tropical storm. Here we have two trucks. SPEAKER_36: Two big trucks. Notify on the left side how we sweep to the left a little bit to keep the distance from the right truck. Right. And then we sweep to the right to keep distance to the right. SPEAKER_97: All while staying in your lane. Exactly. Just to be clear. SPEAKER_36: Just to sweep a little bit. SPEAKER_98: Here you are going underneath a highway underneath a highway whipping around a turn at about 30 miles an hour. SPEAKER_100: And the pedestrian is minding his own business out of nowhere. SPEAKER_05: That pedestrian is walking in the middle of the street trying to commit suicide I think is what you're saying. SPEAKER_101: Here's a bunch of school kids running in front of traffic. Because we're driving right along elementary school. SPEAKER_104: Oh my lord. It's elementary school right there. You have to yield them before you can drive. SPEAKER_105: Where are the teachers? What is going on in China? SPEAKER_106: Well I guess it's out of school. SPEAKER_105: Oh my god. SPEAKER_17: It really is amazing to me I guess when we were in New York we would jaywalk a lot but this is a level SPEAKER_105: of jaywalking in China that I it just seems crazy. Are those children crossing against the light on their own? SPEAKER_36: Yeah. Because that's actually a school zone so they actually allowed you see what's marked on the street. Markings on the sidewalk. SPEAKER_11: They're marking so they actually Oh they're allowed to SPEAKER_110: run in front of traffic. SPEAKER_11: Yeah we have to notify we have to notice it recognize it and avoid avoid any accidents. SPEAKER_17: Alright when we get back I want to get your estimate on when level 4 you said level 3 3 to 5 years now level 4 which is hey that's when SPEAKER_08: you're going to really be competitive and maybe be selling some really cool equipment and cars to our friends at Lyft and Uber when we get back on This Week in Startups. Chamath Palihapitiya: Have you ever wondered how to get on the cap table and buy and sell shares in private companies? I'm sure you have. Well there's an easy place to do that right now it's called Equity Zen Equity E-Q-U-I-T-Y Zen Z-E-N and Equity Zen is a premier online marketplace where you can invest in private tech companies backed by top tier venture capital firms and angel investors like me and you can do it long before they IPO Equity Zen is also for shareholders who want to get a little liquidity maybe you've been working at a company like Robinhood or previously Lyft and Uber and you want to just get a little liquidity before those companies go public Facebook etc well you can sell some of your shares on Equity Zen and get cash and help out the private investors who are trying to get shares in these coveted world changing companies some examples of companies that are available right now on Equity Zen before they IPO'd include Spotify Sonos Glassdoor MongoDB Cloudera and PillPack all of these companies did exceptionally well I was able to use this a couple times in my career I sold a little bit of my com in a secondary I sold a little bit of my Uber in a secondary it's a great idea if you're an angel investor or shareholder in these companies maybe to sell 10% 20% of your position as they grow so you know listen you can dollar cost average and it's great for private investors who want exposure to this area to buy a little bit of those shares go to equityzen.com and for twist fans only Equity Zen is offering half the minimum investment that's right you can now invest just 10k as opposed to the previous minimum of 20k but only if you go to this URL equityzen.com so you can dip your toe you can get your feet wet you can get your beak wet just for a taste this is for people who obviously are investors you know there's risk involved but hey no risk no reward that's the story in my life thanks again to Equity Zen for supporting my podcast and for lowering the minimum so people can get a taste equityzen.com slash twist equityzen.com slash twist let's get back SPEAKER_05: to this amazing episode of twist hey everybody welcome back my guest today James Peng SPEAKER_08: he is the co-founder and CEO SPEAKER_114: of pony.ai SPEAKER_08: previously SPEAKER_114: you were at Baidu and before that Google is that right SPEAKER_115: that's right what did you do at Google I was working mostly on the infrastructure side SPEAKER_36: ads back end SPEAKER_116: the ads back end SPEAKER_36: advertisement the ads AdWords AdSense the back end system SPEAKER_17: got it so you were figuring out how to get people to click on ads and make 17 cents SPEAKER_08: and now you're figuring out how to not kill people who are jaywalking in the streets of Beijing SPEAKER_115: that's a good summary SPEAKER_08: wow so not being facetious here and I know I had another question I was going to ask you is the background that you learned at Google SPEAKER_05: and the internet revolution how did the internet and the mobile revolution SPEAKER_118: lead to the self-driving revolution SPEAKER_119: one of the key SPEAKER_11: challenge for autonomous driving is how do you efficiently handle that huge amount of data because all the sensors in the vehicle collects all the surrounding information so so in that regard how do you design a system that can efficiently process all the data extract useful information from the data I think all those backgrounds are very interconnected yeah SPEAKER_05: so if you were going to index the world's information and allow people to search for their favorite book or comic book character and then instantly give them an answer you have to take you have to have some tool to take a bunch of inputs which would be web crawlers or in this case with self-driving the inputs are cameras put that data into a database or into some sort of framework and then make sense out of it with an algorithm Google made sense of web pages and documents and images and videos you're making sense of videos and cameras and I'm assuming LiDAR SPEAKER_08: in the real world SPEAKER_122: yes SPEAKER_08: so essentially self-driving is the Google search engine at play in real time indexing the world SPEAKER_123: index the world around you the world around you yes not the entire world SPEAKER_08: but in a way we will have almost the entire world if all these cars and all these phones were actually indexing it it's kind of interesting to think about we'd have the entire simulation in another simulation that's right we could build a simulation of the simulation you think we're living in a simulation right now you know this theory simulation SPEAKER_126: yeah yeah sure sure of course SPEAKER_08: what do you think the percentage is that this is a simulation right now a simulation SPEAKER_127: of a simulation SPEAKER_08: or a simulation that you're making a simulation of a simulation that you don't know you're in a simulation what do you think the chances of that are SPEAKER_129: is it zero so far I'm not a believer of that but nothing is zero so I don't know but I would say very slim SPEAKER_105: very slim all right SPEAKER_08: and in a way if we hadn't had the graphics cards and the sensors that mobile phones brought and video games brought we would never have gotten the ability to process this amount of data would we no when the first Waymo cars came out the cost of that sensor array I think it was SPEAKER_77: almost 10 years ago now that they started SPEAKER_136: 10 years ago SPEAKER_77: yeah my understanding is it was over a hundred thousand dollars for all the lidar and sensors is that ballpark correct SPEAKER_11: used to be more than that SPEAKER_77: 150 or something SPEAKER_08: yeah what is what do you have it down to now what is the industry because you're a lidar solution not a camera based solution like Tesla we'll get into that what is your package of sensors cost today what will it cost in five years to outfit a car SPEAKER_11: now it's already halved so 75 we're about 70 everything including all the sensors computer graphics card everything I mean everything when we put in the vehicle and the computer SPEAKER_05: that's in that vehicle would be the same as the highest end video game machine or more SPEAKER_44: it's actually less less what we always trying to do is to optimize SPEAKER_11: the whole system so we've actually put in not too powerful of a computer because the thing that if you put too powerful of a computer that means you draw too much power from the vehicle you'll reduce the range SPEAKER_144: you'll reduce the stability of the vehicle SPEAKER_05: got it okay and so that's only a couple of thousand what is all the other cost then what does the 70,000 break down into today SPEAKER_11: it's sensors will be maybe 60-70% of the total cost SPEAKER_146: and what kind of sensors are there and how many SPEAKER_15: we have it's the center main lidar 32 line lidar that's that's about 20, 22-ish got it SPEAKER_33: so that's the big piece that's the big piece that's that big circle people see on top that's spinning SPEAKER_11: no we have a smaller one not the 64 old big one right it's much smaller but it's still we're not purchasing by bulk so the price is still relatively high SPEAKER_148: five years what will it be down to SPEAKER_11: my prediction with so many companies start working on this as so many startups and companies start manufacturing the lidar I'll say easily in three five years we're gonna be half again easily wow SPEAKER_77: and $35,000 seems like a lot SPEAKER_08: until you realize that these cars could run SPEAKER_05: almost all day long probably 18 hours a day so it would be the equivalent of having three full-time people working SPEAKER_33: or four full-time people working around you know working full-time for SPEAKER_150: and then you could amortize that cost over what would be the life cycle five ten years something to that SPEAKER_36: even less than that right you think about the uh if it's even single driver you amortize it the payback period would be months SPEAKER_39: it'd be under a year SPEAKER_36: well under a year yeah SPEAKER_39: so when I went to commercial break with you I was asking you when do you think level four let's pick a city of note Beijing San Francisco Los Angeles you pick the city um when do you think that will occur this is the hard question SPEAKER_118: isn't it SPEAKER_11: yeah um I think the SPEAKER_118: and take me through your thinking SPEAKER_11: of path to commercialization yeah will take a step-by-step approach got it so we'll get to that level three we'll get to not level three level four as I mentioned already level four is without human intervention driving in certain areas right for example in the Silicon Valley in Mountainville in Sunnyville in those areas probably we'll be able to completely handle them in five years got it um then we start maturing start learning and and also making the uh mass production even better so go then probably another three five years going to cities like San Francisco SPEAKER_05: got it so you see working in small towns where you can define the problem set in a very narrow way SPEAKER_156: exactly SPEAKER_08: and why is that so much easier I mean we're watching you drive through Shanghai Beijing Guangzhou why is that going to be the path because it will just be people because is that because of human nature people are conservative SPEAKER_62: and they want to see it work SPEAKER_36: no it's because it's like how you uh prepare for the test right you you do SPEAKER_11: hard problems you want to do practice on much harder problems but when during the test if you want a perfect score you want the test problems to be easier SPEAKER_159: got it SPEAKER_11: right now we we can handle San Francisco we can handle New York we can handle Beijing right but not perfectly SPEAKER_31: got it SPEAKER_11: but going there will help us to make uh handling suburban area much faster SPEAKER_05: so it's a you're forcing yourself to solve the hardest problem and then you'll go back to easier problems after that SPEAKER_36: exactly so that you can get perfect score in your test SPEAKER_08: right and mountain view there's going to be less pedestrians you're if there's a construction site or something changes you'll know about it exactly very quickly um and so it'll be very easy to do but you could also have SPEAKER_05: a safety driver in there is anybody doing tests without safety drivers right now is the Waymo test there's a Waymo test in Arizona right SPEAKER_15: Waymo has a small scale testing without a safety driver in Arizona in Arizona SPEAKER_33: and they're using minivans SPEAKER_08: and how is that going to the best of your knowledge what has the reporting on it been because it's public the reporting on it SPEAKER_145: I think so far SPEAKER_37: it's at least it's it's a perfect safety track record nobody's been killed SPEAKER_05: um have there been instances where remote people have had to take over because they have a remote person watching each car is that correct SPEAKER_11: no no it's like a remote person watching a bunch a fleet of vehicles SPEAKER_05: got it so there's one person at a desk somewhere watching 10 cars at the same time and watching the video of it is that SPEAKER_164: live streaming SPEAKER_67: live streaming it and if there's an issue with one they can take over and drive it SPEAKER_25: not drive it to lead SPEAKER_11: to safety SPEAKER_166: to lead it to safety SPEAKER_11: for example if it's a construction that the vehicle was not able to recognize it then the one behind the desk can give it an instruction to to bypass it SPEAKER_08: really wow they can see a route around it or they can pull the car over and then send somebody out exactly have their they have to document all the times that occurs right and publish it every year all the data from all these safety has to be safety trials SPEAKER_77: have to be published right SPEAKER_11: that's the law in California I'm not sure if that's the one in Arizona got it SPEAKER_08: so what do you think when we get back from this break I want to know what bet you're making as a business now because it feels like there are four or five players who are very close to level three and four and have different ranges of pilots and tests going on what is going to be the business here will it all as some people have told me be commoditized and everybody will have access to the technology at about the same time so then Uber Lyft and other services will have it at the same time or do you believe that someone like Tesla Elon claims he's going to have next year his own fleet out there and he's going to kill Uber and Lyft and have half price or one third price SPEAKER_17: or is he going to win when we get back thanks again to the Walker Corporate Law Group a boutique law firm specializing in startups for supporting this week in startups for a decade thank you Scott from me Jason Calacanis to you Scott Walker thank you not only for supporting the program but for supporting all the startups I've sent to you over the years you know the Walker Corporate Law encourages fixed fees they believe billable hours are inefficient and they will give you a fixed fee so that when you get that bill at the end of the month boom you know what you're going to pay you're not going to have that sticker shock when you open the PDF additional services mergers and acquisitions licensing arrangements terms of service privacy policies all that stuff that you need to do to make sure your startup is protected and done right and their lawyers have decades of experience 10 20 years or more there are no junior associates getting on the job training with your startup so if you want to talk to the founder himself Scott Ed Walker go directly to walkercorporatelaw.com or email scott at walkercorporatelaw.com scott at walkercorporatelaw.com or call 415-979-9998 415-979-9998 okay let's get back SPEAKER_77: to this amazing episode all right SPEAKER_08: we're talking about self-driving with somebody who's working on it every day and has raised 264 million is that correct SPEAKER_58: that's right SPEAKER_08: and you're valued at over a billion SPEAKER_58: 1.7 SPEAKER_08: 1.7 billion dollars for pony.ai you have offices here and also in China SPEAKER_58: that's right SPEAKER_08: are you a Chinese company an American company SPEAKER_11: global company we started in the Bay Area okay we expanded into China SPEAKER_08: got it and so what is the regulatory framework in China like for self-driving and do you hope SPEAKER_77: to play in it there SPEAKER_36: I think similarly as I mentioned it's cautiously optimistic SPEAKER_11: so the regulation opens a little bit for example several big cities come up with regulations for autonomous driving testing then you can apply for license once you get the license you are allowed SPEAKER_15: to do the open road testing but you have to have SPEAKER_08: a safety driver SPEAKER_15: you still have to have a safety driver SPEAKER_08: should there be two safety drivers because there was a horrible accident that Uber was involved in where somebody was jaywalking in the middle of a street they were obviously at fault and then also at fault the driver who was looking down at their phone and there was only one safety driver so the technology was at fault the safety driver is at fault the pedestrians at fault you have a triple fault going on here which is typically how errors happen right with computers you have multiple or when an airplane or some complex system fails it's usually multiple people at fault should there always be a second safety driver in there so that the first person doesn't or feels the social pressure to not look at their phone because what they found was when there was two one person was taking notes one person was prepared to take over you didn't have this issue of people SPEAKER_92: going and playing Candy Crush or looking at Twitter or texting SPEAKER_36: I think like anything else it always has a very small SPEAKER_11: probability of failure so I think to start with the autonomous driving system within the vehicle needs a backup system sure SPEAKER_187: so two computers SPEAKER_11: at least it has one main computer and a backup computer two sets of different sensors maybe maybe the backup system does not need to be as powerful SPEAKER_15: as the main system but it at least can do the safety job so if the SPEAKER_188: main system is at fault the other system could pull you over or tell you to take over exactly SPEAKER_08: so you have to build you have to duplicate every system SPEAKER_15: I would say it's duplicate but make make a much smaller system but redundant as a redundant got it SPEAKER_05: so it's not like a raid array where you have to have the same exact hardware profile because that would get awfully expensive what happens if the LiDAR breaks up top SPEAKER_36: then you still SPEAKER_11: have other sensors that can SPEAKER_191: the other camera sensors SPEAKER_11: can allow you to actually safely pull over I think that SPEAKER_15: needs to be put in place instead of two drivers because then that defeats all the purpose of going for autonomous SPEAKER_138: yeah SPEAKER_39: did you look at that instance where Uber had the fatality and what was your assessment of that SPEAKER_11: well I think your assessment was absolutely right it's the some three things went wrong at least at least and then it's a very unfortunate event SPEAKER_08: what was the fault of the self-driving technology was that ever unpacked or explained I haven't seen an update on that because something went wrong with the self-driving technology that it didn't see the person in that low light which it's lidar supposed to see at night this was a completely dark street right it was between street lamps or something and that video of it SPEAKER_11: of course without knowing all the nitty gritty details it's hard to make the right assessment but something must went wrong otherwise the cars supposedly should make a stop SPEAKER_39: got it what is SPEAKER_08: your business model how are you going to make money do you think or you're going to push that out and figure it SPEAKER_77: out once you get the technology done SPEAKER_198: we start at the stage SPEAKER_11: we need to figure it out of course we iterate and try to change it as we go we start with by working with the car manufacturers so it's sort of like a licensing model as the stakeholder in the mobility business all the OEMs the car manufacturers they need to own or at least to know this technology so I think licensing definitely is a very valid revenue stream SPEAKER_15: the other is as you said it's to own the fleet and to use it as a robot taxi SPEAKER_08: what do you think of Elon's claim that next year he will have a robot taxi fleet in a city in America up and running SPEAKER_11: depends on the scale but I think next year probably is a bit too aggressive but some point I think we will SPEAKER_40: you said three to five years SPEAKER_11: we'll have certain maybe in the order of thousands SPEAKER_39: of cars on the road somewhere will those cars have safety drivers or not SPEAKER_16: we're talking about real autonomous so no SPEAKER_05: safety driver I SPEAKER_16: think with the safety driver is already out there right SPEAKER_05: so yeah because Waymo has it without the safety driver so he believes or he said that a couple of SPEAKER_39: months ago so he believed he needed let's call it 20 months you believe 36 to if you are average in aggressiveness in terms of your estimate which I think you are he's probably being a year more aggressive but it's possible in small cities small deployments to do this SPEAKER_202: I think it could be SPEAKER_39: yeah SPEAKER_08: and you just geofence it and the cars won't be able need for public transportation and you see governments becoming the group that then provides the service because what would be the need of buses if we had buses right now cost $2 or something and right now Ubers and Lyfts cost $4 or $5 typically $6 or $7 what's the point of even having a bus if people could take SPEAKER_77: point to point and it SPEAKER_11: even buses can be autonomous that's true too right and then there are subways as subways take you to a longer ride I'm just SPEAKER_05: thinking specifically about buses like buses are people have to transfer and then walk in between them it just seems like SPEAKER_11: see the thing that why the ridership of buses are not very high because it's not convenient and also it's very expensive to run the way it's expensive because the drivers are expensive with expensive drivers so the wait time there's not that many buses you have to wait for five minutes if you have autonomous bus imagine with only 10 SPEAKER_16: seats or five seats then you can go every minute wow so you SPEAKER_211: or they just sit there and wait for somebody and then when they get on then it starts taking them SPEAKER_11: then ridership will go up it will be more convenient and it will also be more energy efficient SPEAKER_212: yeah that makes total SPEAKER_08: sense why is this not open source and is there an open source movement for self driving yet because I count a lot of open source technology SPEAKER_39: is part of the framework of open databases of all SPEAKER_114: of the maps out there and all the world world data or the algorithms is it too complex to be open source is it required too much data to be open source or is there SPEAKER_77: an open source project that's got any kind of traction I SPEAKER_214: think they're definitely open sourced out there but the thing that the system is so SPEAKER_11: complicated and also it's so mission critical at this stage so mostly everyone is trying to develop it I think at certain stage at certain time there might be open sourced version but I think at this time they're put too much emphasis on safety say for example there's an open source platform and there's accidents SPEAKER_100: who's responsible for it SPEAKER_29: yeah it's the same thing with databases right now if you use open source SPEAKER_08: database the person who implemented it would be responsible right but a database isn't necessarily going to kill somebody exactly so that does make it slightly more complicated would this be moving faster if it was an open source project or there were more open source initiatives do you think or does it move faster with venture money and closed systems SPEAKER_15: it seems SPEAKER_39: like a lot of duplicate work right is going on SPEAKER_15: there not enough data for me to make that assessment yet but I think SPEAKER_11: at least from Pony's point of view we're making very fast execution as the video have shown that we were already able to handle very complicated urban scenarios just based on simple just if it were just based on open source initiative I don't know if it can be this fast SPEAKER_08: when we get back from this final break I it would make sense that cars and we've seen this concept before with adaptive cruise control there was a concept that at some point we would be having the cars in the left lane go into a train and be able to go 100 miles an hour 125 miles an hour and only be two car lengths between them and the computers would get it all perfect and we see speeds on the highway what I want to know from you based on your experiences would you put your family would you put yourself in 10 years in self driving cars that were going over 100 miles per hour on safe roads when we get back on SPEAKER_17: this week in and now with Salesforce essentials you get easy out of the box tools and support at a startup price point and that's critical here are the benefits you're going to get set up instantly and you can easily scale your sales team by customizing with clicks you don't need to write any code you will get simple integrations that connect and integrate all of your data under one roof and you'll have you'll be able to automate busy work and repetitive tasks so you're not wasting your time and customers can help themselves of course with a self service support site everything you need is on one screen so you can track emails calls and meetings from your inbox get access to the once again that url salesforce.com SPEAKER_233: slash twist SPEAKER_29: 50% off all right welcome back my guest James Pang he is the co-founder and CEO of pony.ai SPEAKER_08: he's competing with the likes of cruise which is owned by GM I hear they're very far behind is that true they're the furthest behind SPEAKER_179: not really Tesla SPEAKER_114: Waymo who else has got a major Baidu has initiative SPEAKER_236: yeah it has one SPEAKER_114: has one Apple doesn't have one SPEAKER_53: now or does SPEAKER_238: it still has one Titan SPEAKER_53: is still going on I think so but are they SPEAKER_239: committed to it or not SPEAKER_11: what do you think I don't know I think they're committed it's such an important market that SPEAKER_114: who did I miss Cruise Tesla Waymo Baidu Titan at Apple yourself pony who SPEAKER_39: else has got a major initiative SPEAKER_241: Aurora SPEAKER_39: Aurora yes and who else SPEAKER_242: major ones SPEAKER_151: Neuro Neuro SPEAKER_244: is another one SPEAKER_114: so we've listed now major 1 2 3 4 5 6 7 8 players SPEAKER_39: here it's getting a bit crowded SPEAKER_245: I SPEAKER_08: was asking you before about is everybody going to get there around the same time because everybody I've had on the program talking about it says three to we could probably do it in a controlled environment and everybody says level 4 SPEAKER_17: more like 10 years do you agree with the 10 year kind of for level 4 or 5 or you think that's too far out SPEAKER_15: I think level 5 probably still going to take a while level SPEAKER_39: level level level San Francisco New York Beijing you think that's 10 7 to 10 SPEAKER_246: could be could be faster but 5 to 7 SPEAKER_05: 5 to 7 great so the range of when they start showing up on streets is 3 to 7 years if things go slower maybe it's 5 SPEAKER_08: to 10 but it's coming in the next 10 years what's it going to look like on a business basis will the person who makes the cars and the technology like Tesla and Cruise win or like an existing GM which GM owns Cruise yeah will they win or do you think it's going to be a new product like maybe if you make your own car at some point or Waymo is going to make their own cars or Waymo SPEAKER_179: put SPEAKER_77: do you think you going to make your own car is that not a good idea does it probably not probably not you don't SPEAKER_250: have to you SPEAKER_17: don't have to the existing cars are good yeah they're very good so what's the what's this business going to look like in five years is it going to be a free for all with SPEAKER_05: just everybody like the scooters where they're just flooding cities with cars and scooters and bikes everywhere I SPEAKER_27: think the whole mobility market SPEAKER_11: the industry is big enough it certainly has the room for multiple companies to thrive SPEAKER_196: yeah SPEAKER_11: but who's going SPEAKER_15: to win and who's going to lose out only time can tell SPEAKER_08: yeah and there'll be some people like Tesla which want to own fleet and make the cars and then there'll be people like Uber and Lyft where Uber SPEAKER_77: you would be selling SPEAKER_08: into Lyft I SPEAKER_180: think more like collaborating SPEAKER_08: ah so you would maybe go after a city together what about speed when we left you on that last break I SPEAKER_11: probably don't need SPEAKER_58: to get their own driver license they won't have it they don't they don't need to SPEAKER_255: yeah it's interesting what about SPEAKER_39: speed because these SPEAKER_114: things are so safe and the thing that kills people in speed is human reaction at speed is terrible right but a computer's reaction at high speed would be much better is that correct SPEAKER_258: definitely SPEAKER_114: so a human trying to hit the break a human can only see so far down the road SPEAKER_08: a computer could see further correct SPEAKER_259: that's right SPEAKER_08: and it could see through the car in front of SPEAKER_11: the straight highway safety levels at that time but definitely the SPEAKER_144: scenarios you just described totally is feasible SPEAKER_08: yeah because we already have I think they've raised the highway speed in some long straight roads in America to 85 like in Nevada and stuff like that so it seems to me that people will start driving 100 miles an hour and that will make certain commutes more viable and with electric cars you're not going to be worrying about polluting as much what's SPEAKER_39: the Chinese government's involvement in all of this and how does that impact the SPEAKER_08: pace I've heard people say China is going to get there before America because the Chinese government can with eminent domain or SPEAKER_39: you know with regulations move much faster when you want to build a bridge or a highway they do it overnight or seemingly in a weekend in China because they don't have a lot of regulations and the government has a lot more authority to do so is it going to arrive much faster in China and also they have the need so is it going to SPEAKER_15: to tell you you are Chinese who SPEAKER_255: has more will right now today the Chinese entrepreneurs or the American in your experience SPEAKER_15: I think both we definitely see the will and the hard work and the SPEAKER_11: strong vision on both sides SPEAKER_271: yeah what do you think of the 996 culture in China SPEAKER_211: it seems people work much harder in I think the number show is it SPEAKER_33: healthy to work that hard and is it something that should be criticized or should it be something that is celebrated the work ethic in China and technology today SPEAKER_86: my SPEAKER_15: personal view I take very neutral view on this I I SPEAKER_11: just love my work there's nothing wrong with work 60 hours 70 hours a week right SPEAKER_274: yeah yeah it just seems to be the standard in China well SPEAKER_275: at least not in our company not in your company SPEAKER_274: you let people make that decision and the engineers America versus China do you see a difference similarities we SPEAKER_15: definitely see both countries have great engineers have great researchers engineers in the valleys still with many many years of accumulation it's definitely more experienced SPEAKER_150: right because they've worked at a Google or a Facebook and the number of people that have worked at a Baidu is smaller SPEAKER_15: even look at Google Google so great is all the engineers used to work at some SPEAKER_11: other companies SDI some Microsoft some microsystems and all these great companies build on each other so it's the whole ecosystem that train the mind SPEAKER_104: side of the engineers and the researchers it's SPEAKER_29: very interesting to have a Chinese you're American citizen now too SPEAKER_280: yeah SPEAKER_29: oh wow congratulations when did you become a citizen SPEAKER_280: many many years ago SPEAKER_39: congratulations wow but you go back on a regular basis you have employees in both places so it's nice to see the deep collaboration of the countries and you having employees in both places you're the SPEAKER_150: you're the vote for engagement engagement between our countries is good you believe I think so yeah SPEAKER_11: I think collaborating engaging understanding each other is always good SPEAKER_150: yeah what is it that Americans don't understand about China what should we understand and what do the SPEAKER_212: Chinese not understand about America that they should understand more of SPEAKER_36: I think the mutual sides understand each SPEAKER_11: other a lot more than they used to be when I first came to this country about 20 something years ago I SPEAKER_15: knew very little about the country but I think these days when people start growing up the younger generations they actually knew each other the culture and the things going on in each side a lot better than we used to be and I think that leads to a lot of great things SPEAKER_166: I think what's going to SPEAKER_114: solve this whole problem movies I think the fact that we're going to funding Tom Cruise movies I know it sounds silly but SPEAKER_39: if SPEAKER_114: you share those SPEAKER_39: stories and Americans started to understand some of the stories from China and some of the folklore and some of the characters if we share that you SPEAKER_114: start to share like a mythology and heroes and villains and common ground so if you meet each other and SPEAKER_39: culture and you think about how revered martial arts became in the United States it's SPEAKER_114: very interesting how that part of the culture SPEAKER_29: is now leading over I was just in Hong Kong and my book I translated to Chinese right there SPEAKER_288: what does SPEAKER_29: it say on the cover SPEAKER_288: rich man's mindset SPEAKER_29: rich man's mindset right that's what it says yeah they renamed my book angel into the rich man's mindset yeah I'm now getting emails from people in China who have read the book and they say I'm in China I'm in Beijing I'm in Shanghai I want to be an angel when can I get SPEAKER_292: autographed paper well you just take that one off SPEAKER_17: the wall and I'll sign it for you first time I've ever signed it you have to show me how to write Jason in Chinese all right listen SPEAKER_08: this has been great let's recap you would put your kids in 100 mile per hour car 10 years from now racing down the 280 you feel that confident about it if you want to work 996 and you opt into it that's your choice muscle tough that's good for you you're not going to make the cars you're going to sell the technology to other people and level three level four we're ready to go we'll have it in Mountain View or Palo Alto in the next three to five years no problem easy peasy and SPEAKER_114: then in San Francisco and Beijing and Guangzhou as soon as maybe six seven eight nine ten years SPEAKER_29: congratulations James on SPEAKER_278: I SPEAKER_08: am sitting in traffic with my autopilot on looking at my phone going oh my lord I could recapture this 20 minute commute and I could be watching the game instead of just listening to it no SPEAKER_119: you should listen to your podcast SPEAKER_08: listening SPEAKER_299: to podcasts exactly absolutely all right we'll see everybody next time on the swing service bye bye