These are the stocks posting the largest moves before the bell.Market Insiderread more
China wants to have another round of talks with the U.S. before signing phase one of a trade deal, a source tells CNBC's Kayla Tausche.Marketsread more
The Treasury secretary Steven Mnuchin expressed optimism Monday that the U.S. and China have a workable first-phase agreement.Economyread more
"The Champagne should probably be kept on ice, at least until the two presidents put pen to paper," said state-owned media China Daily.Traderead more
U.S. stock index futures turned lower after China said it needed to have further discussions before it would sign off on the so-called phase one trade deal President Trump...US Marketsread more
Analysts say the partial U.S.-China trade deal doesn't touch on thorny issues plaguing both sides, and warn talks could break down again.World Economyread more
U.S. stock futures turn lower after word that China wants more talks before signing the "phase one" trade deal.Marketsread more
Economists polled by Reuters had expected Chinese exports denominated in the U.S. dollar to fall by 3% and imports to decline by 5.2% in September, compared to a year ago.China Economyread more
Economists Abhijit Banerjee, Esther Duflo and Michael Kremer won the 2019 Nobel Economics Prize for their work in fighting global poverty, the Royal Swedish Academy of...World Newsread more
Boeing's board removed CEO Dennis Muilenburg as chairman amid the fall out of two 737 Max crashes that killed 346 people.Aerospace & Defenseread more
The U.K. and EU are gearing up for what could be the busiest week in British politics since June 2016.Europe Politicsread more
Below is the transcript of a CNBC interview with Harry Hu, COO, poni.ai and CNBC's Everett Rosenfeld. The interview took place at CNBC's inaugural tech conference, East Tech West, in Nansha, Guangzhou.
ER: Great, thank you-, uh, thank you, Mandy. As all of you have heard, this discussion will be about autonomous driving, and, as you heard from ( earlier today, poni.ai is a company that is, uh, doing work in Guangzhou on self-driving technology, especially on the software side-,
ER: You know, that is a space that a lot of companies are doing work in, giant tech conglomerates, small startups, what is different about poni.ai?
HH: Yeah, I think it's from our genes, because we-, our founders, primarily we are from Baidu, and Google, who previously have been through-, like, the difference, with those two companies, basically, they are-, the one is best in US, the one is best in China. Um. So, we know the pros and cons, and what they did before, and probably, if they want to scale up, in to, like, tens of thousands of vehicles in the future, what are the things they are missing, and what are the things we can do better? Do we have a chance to actually catch up, in a technology roadmap? So, that's basically-, we're seeing some of the, you know, things that they didn't, you know, foresee that clearly. For example, for Baidu, they use this system called ROS, R-O-S, Robotic Open System, it's-, it's actually an open source system, especially built for robotics, and when we start our, you know, autonomous driving startup, we are thinking that this-, this system is not really ideal for autonomous driving, because that system is kind of a (inaudible), you know, in robotics, you have-, it's basically a robotic arm, you have several parts, connected in a serial way, but autonomous driving is actually a centralized system. So, we rebuilt the system, and we rewrote the system, and we are adding some of the elements that we think it's potentially to be the-, so we don't really have any bottlenecks when we scale up, in to, like, tens of thousands of vehicles, uh, which-, we are targeting around 200 vehicles next year in Nansha for operation. Yeah.
ER: You also mentioned Google.
ER: Where are some mistakes that you think Google has made, that you're trying to improve upon?
HH: [Laughter]. Yeah, Google's definitely still the world's number one in autonomous driving, because I think Google started in 2005. Well, that's pretty early-, a pretty early year. At that time, they don't really have AI, or they have AI, but they don't really do practical AI things added in to the autonomous driving field. I think, in 2013, they hired Geoffrey Hinton, to be the AI leader in their sector. But I think, for poni.ai, when we start, we think AI elements, because, uh, not only computer vision, but how you make decisions in contact situations, how you make decisions under, kind of, you know, rule-based scenarios, because we-, uh, for transportation, it's highly regulated, you have a lot of rules. How you make decisions within those rules, it's-, it's-, it's-, it's very challenging. So, we-, when we started, we actually add that element in to our decision-making process, so when we-, when we actually are heading in-, when we are actually driving to a crossroads, we-, we do-, we do this judgment better than Waymo. I think-, I think, for-, kind of, a lot of people complain, to (inaudible), um, when they try to turn to the-, uh, what we call unprotected left turn, they are usually just, you know, stuck there, so-, but when we do autonomous driving, we will handle that situation much better than Waymo does.
ER: Mm. So, you talked about poni.ai catching up to some of those tech giants.
ER: Do you see the autonomous driving sector as a winner-take-all environment? Is this a world where multiple companies can coexist? Or is a network effect, where just Google, or just Baidu, or just poni.ai can be the winner?
HH: Oh, that's a very interesting question, because, uh, we've been thinking of this, the paths, actually there are two ways to-, to get ahead. There's one where you do the whole system, and you sell to the OEMs, and you are hoping that one day, you know, (inaudible) is mature enough, so everyone-, so all the vehicles can be pre-installed with this system. And the other way is what Waymo, and I think Uber, does, so, because now, the OEM system, it's-, it's an old, traditional, you know, manufacturing system, where they can't really allow you to actually pre-install this kind of thing, before you pass any, you know, like, auto grade testing. So what they do is, like, they will kind of partnership with OEMs-,
HH: To actually launch a big fleet, and they do a lot of testing, they do a lot of simulation, and make sure that the technology is mature enough to handle, you know, maybe the city level, uh, autonomous driving hailing, you know, services. So, um, we are-, we are actually trying these two routes, at the same time, basically, because the base, it's a system, it's a-, it's a-, it's a technology, so you can try to partner with the OEM, on the one side, and-, and the other side, you do a fleet operation, at the same time. So, um-, so the experience that we've got, when we operate in Nansha, is that, um-, because Nansha, it's just a-, it's such a great-, it's a-, it's a really ideal place for, you know, to launch this autonomous driving services, as Nansha has pretty nice weather. Um. Not as good as comparing to California, but it's-, it's pretty good, for-, for-, compared to other cities in China. Secondly, it's not really that crowded, but it's still got enough people, you know, on the streets, so you meet a lot-, you will know a lot of scenarios. And, thirdly, I think, Nansha, they have pretty nice cities, and pretty nice streets, and everything is very clear on the roads, so you don't have to do actual work to-, like, some of the rural areas in China, you have to do actual work, to make sure the road is-, is clear. So, Nansha-, the infrastructure in Nansha it's-, it's very ideal.
ER: So, what kind of testing-, what kind of scale are you operating at within Nansha? I know there are a lot of autonomous driving operations that are experimenting with ride hailing applications-,
ER: Is that something that you're looking at?
HH: Yes, and, technology-wise, we are actually-, our system is kind of stable, you know, we think we can scale up to 1,000 vehicles, at the same time, but actually, we are gradually building up the-, the vehicle fleet, uh, we are getting close to 50 vehicles, um-,
ER: And with this fleet, are you doing ride hailing? Just within the Nansha testing environment?
HH: Yes, just within Nansha, ride hailing. And on the other side, we are thinking a lot on the product side, because eventually it's going to be the service, it's going to be client-facing, so you need to think a lot of, um, what are the corner cases, or edge cases, the client might face, when they actually use this app, to actually call your taxi? And there are a lot of things, small things, you need to actually break in to, like, smaller pieces. So, think about, what if some of the-, you know, if the customers, they throw up on the taxi, how are they going to clean up? And how are you going to know that you are the guy who was calling this taxi? So, you need to think in to that, deeper in to scenarios, make sure that this is actually a product that can be faced to the client, and-, and the company, so it's very important, as well.
ER: So, you talk about anticipating problems, whether it's-,
ER: Someone throwing up in a taxi, or fighting over a car. Autonomous driving has had some very public problems, some bad headlines-,
ER: Over the last year.
HH: Unfortunately, yeah.
ER: To what do you attribute that? Was that hubris? Was that overconfidence, by other companies? Was that something that is naturally going to happen, with the new technology? To what do you attribute it, and what is poni.ai doing, to avoid headlines like that?
HH: Yeah, yeah, that's a great question, because I think some of the headlines which people are trying to avoid, it's because, um, there are people trying to catch up with the trend, and there is too much capital, and too much talent, in this space, and everybody's trying to get ahead. I think that's where some of the bigger companies get in to the trouble. So, they loosen their standards, when they do the testing, so they hire someone who is not supposed to be the driver, because the driver shouldn't always focus on the road, so you have to ensure the whole safety system is there, and you have to pace yourself. Um. Even-, even though, you know, like, people always try to get the number one in China, to get the get number one in US, trying to do this and do that, to try to launch the first taxi services in Guangzhou, I think we should really think back on-, on the technology, what you can do, and is there other ways you can do better? And to pace yourself, not to hurry. Because, for us, we are already a unicorn in this space, we don't really worry about getting enough money. We already have-, we already have a lot of talent, so we need to think through what is the technical roadmap we should be dealing with, in the next three to five years, and to gradually pace for that.
ER: Interesting. So, you talked about being number one in the US, you talked about being number one in China. Poni.ai operates and tests in both countries, and you have staff in both countries. But, the world's two largest economies do have quite a rivalry, in terms of tech and IP. How does that relationship affect your business model? How are you thinking about the rivalry between US and Chinese tech?
HH: Well, I think, of course, this autonomous driving starts from the US, the (inaudible) engine, everybody is very excited about. The thing is that, for China, I think this is a natural barrier, because the-, you know, in China, a driver just doesn't behave that much-, you know, behave that-, in the same way as the US driver does. I mean, they will-, uh, so, basically, transportation, I think, in a lot of countries, it's like, you have a set of rules, but sometimes, just people doesn't really follow that rule. So-, but for autonomous driver, in our code, we actually, we have to 100% obey that rule. So, how do you deal with-, there's a gap between how the normal drivers drive, and then how you drive, and-, because if we violate the rules, by ourselves, we're going to have a big trouble. So, this is like a game theory, between you, and the people-, the actual people driving the vehicles, and this behavior is different in US and in China. Um. In China-, well, um, we can't really talk that much, but-,
HH: [Laughter]. Um. Like, for example-, like, uh, last week, the German Embassy came to visit us, and they were saying, on the highway in Germany, there's no speed limit. Well, that is kind of scary, I was saying, you know, as we kind of need that speed limit, to actually-, to-, to tell our vehicle how to behave. So, there's different ways of doing things, and, you know, so we need to really think really carefully.
ER: You bring up a really interesting point. So, obviously, you're doing testing in the US and in China, but driver behavior, and rules of the road, and technology regulations are different-,
ER: Everywhere in the world. Does that make it a challenge to scale your technology? Does that make it a challenge to really see beyond Guangzhou, beyond China?
HH: Yeah, yeah. I think that's one of the good lessons that we can learn in Guangzhou, because Guangzhou is actually, it's a tier one city in China, it's got a lot of people, and Nansha-, it's-, it's also very complicated, because it's-, it's not really that close to downtown, but-, but it's-, within itself, it's a-, it's a-, basically, it's a town, and a lot of traffic is here. So, we do this, and we-, we-, we have-, we've already increased the number of scenarios that we have, we've actually tripled the number of scenarios, when we're comparing to what we do-, what we have done in California. So, this will actually help us, if we go back to the US, if we are testing in some other areas in the US, or Europe, I think we are going to have an edge on that. There's a joke in our industry saying, if you can do autonomous driving in India, you can do it everywhere-,
HH: So, actually, Guangzhou will help us to bring up the standard, so-, so we are more confident, and when we-, we still have a fleet in California, Freemont, so when we do our upgraded system there, uh, we trial-, we trial autonomous driving, we feel a lot more confident-,
HH: When we do it in the US.
ER: So, as you survey the regulatory environment in China, in the United States, do you think one is more favorable to companies than another? Do you think one country is priming their autonomous companies for success, more than another?
HH: Um, I think China, definitely, they are more, um-, uh, China is actually more reserved on this, but um-,
ER: More conservative.
HH: Conservative, yes. But there are exceptional cases, because, I think for Nansha-, basically, Nansha, it's a national new area, it's a free trade zone, and Nansha has the mission, and power, to try new ideas, to do new things, and that's-, I think that's the mission that President Xi gave to these, you know, free trade zones and new areas. So, it's good, and if we do it in the right way, and people can see the results, and people can see that actually, the actual fleet is running, in a very nice way, if we have three years of records with no accidents, you know, and with hundreds of millions of miles accumulated, then people will just trust us. And because of this mechanism that Nansha has, for innovations, and all this, we really have an edge, comparing to the other competitors in China. But the US is open ground, and I think, you know, maybe it's-, so, everybody has this-, so, if you try to actually launch your fleet in California, there's not too much restriction on this, but you have to make sure that within yourself you have confidence at doing this.
ER: So, is it fair to say that you see your biggest competitors, on the self-driving side, coming not from within China, not from other countries, like Japan, which have companies working on it, but from within the United States?
HH: Um. I think, for players from the United States to get in to China, they have to deal with several issues. First, is-, is the (inaudlbe). I think it's-, the map, it's kind of restricted, uh, because it's basically sensitive data, and you need a way to store it, but it is sensitive, so they probably need to find a local partner for that-,
HH: But-, but if you-, if you have a different map, or-, and a different datacenter-, um, so-, so, basically, you need to, kind of, fine tune your whole model, because you have-, so, that-, that-, so-,
ER: So there are challenges on both sides.
HH: So there are challenges. Yeah.
ER: Well, thank you very much, Harry. Ladies and gentlemen-,
HH: Thank you.
ER: Harry Hu, from poni.ai.
HH: Thank you.