What 72hrs in China taught me about the future (AI, EVs, more)

0:00You know, when you land in Beijing Airport, you are greeted with an incredibly clean, well-functioning, well-ordered airport. It's far above the quality of any of the US airports I've been to over the last few years. I would put it roughly at the level of Heathrow Terminal 5, so certainly better than Terminal 3 or Terminal 2, better certainly than Frankfurt Airport. Perhaps not as good as some of the great Asian and Middle Eastern airports that have emerged since then, but that's absolutely a sense that you get as you land. It didn't feel particularly police-y either. I mean, you do get a sense of CCTV cameras as you do, for example, when you're landing in a London airport and you obviously feel the police and security presence in many of these airports, particularly in the US.

0:49I was a bit surprised. I really expected much more of that. And there were, you know, lots of batteries of cameras you might find in public space in larger numbers than you would see in London. So you arrive at this airport. It's clean. It's really, really well-functioning. One thing that you notice really straight away is that there's essentially no obesity in China. And in three days and seeing hundreds of thousands of people, you get a sense that this is not an epidemic problem that they are dealing with. The scale is absolutely huge. I actually went to a conference as a sort of academic expert, and the conference center I was at ran to some 1.4 million square meters.

1:35To give you a sense of what that scale is, that's 100 times the size of Union Square in San Francisco. If you imagine Soho and Tribeca in New York, it's double that size. And this is a convention center that was put up in just a few years, about seven or eight years ago. It's really, really remarkable. So I want to put my comments into three categories. The first category is what I noticed about infrastructure and engineering. The second is what I saw in and around AI. And I talked to a number of AI founders and researchers while I was there. And also, the third set of observations really about electric vehicles and what's happening with the auto industry, again, having ridden in a number of cars, but also talking to a number of people involved in those areas.

2:29So I'll make my way through there. And as always, I'm happy to take questions about this or any other topic as we make our way through. So let's get started with point A, which was infrastructure and engineering. I would say you just really notice that there is a lot of infrastructure running around as you come off through the airport and you're on the highways. And it's such a dissonance compared to landing, say, at JFK and being on the Long Island Expressway, where things look crumbling, they look like they're falling apart, there's some very, very slow repairs going on. This is not the case in Beijing at all.

3:09There is just a sense that things are being built, they're being built to last and being built to be maintained. And that infrastructure story found its way across the high-speed rail. I took a high-speed rail from Beijing South Station through to Tianjin, running at just over 300 kilometers an hour, which is about 180 miles an hour. It was incredibly cheap. It was about£17, about$20 for that 30-minute journey. Beijing South Station is chaos. It's like so many people, so many people that you will get lost in there. But again, to give you some sense, it's not much different to Grand Central, to be honest, or to London's Waterloo.

3:55The trains themselves are super, super efficient, running on time, very, very quick, and really, really reasonably priced. So that's just part of that infrastructure story. But as you run through on the train journey, what you are seeing is all that concrete and steel that has been poured and laid down over the last 30 years to support these bigger and bigger cities. Beijing is, of course, about 20 million people. Tianjin, where I went, was a much smaller city. but still 12 million, so really nearly double the size of London. And you're seeing pristine, well-ordered roads and expressways. And in fact, as I drove back from Tianjin to Beijing, the roads again were really, really impeccable.

4:40Busy, the occasional jam of traffic, but running really, really smoothly. You know, thinking about other countries I visited, it certainly felt bigger in scale. It felt more invested than anything I've seen in the US and things that I've seen in many parts of Europe, even some of the great construction projects like the Orison Sound crossing between Sweden and Denmark. Compared to some of the things that have gone on in India, there isn't quite the same backdrop of chaotic India when you are in these in these highways and on the train lines. So I was really struck about with the degree to which the infrastructure has been taken really, really seriously.

5:27Of course, this is not a new story because we've read about it in the papers and many of you have no doubt visited China dozens of times in the last decade or so. Some of you have probably lived there as well. But I would say that you just, I certainly noticed that sense of scale and quality of the infrastructure. It was also distinctly Asian. So there were parts of it that reminded me slightly of bits of Singapore. Infrastructure is great. You've got to then do things on top of that infrastructure. And I had a chance to talk to some of the AI scientists and some of the some AI founders in China.

6:01And I want to really reflect on on that, plus certain things that I've read and have gone on over the last few days. I think there's a very distinct path that the Chinese AI world is taking that's different to the US AI world. So in one sense, what I see happening in the U.S. is a really strong emphasis on building these very powerful, abstract, highly general large language models with a real commitment and a belief that the generalizability is there. It's the generalizability that will provide better quality applications that users will like much, much more. Now, of course, there are vertical things happening in the US, but they don't have the same degree of visibility and the same degree of perhaps of traction as a big general LLMs.

6:54What we're seeing in China from the big labs who I didn't get to meet, I didn't get to meet at DeepSeq or Baidu and so on. What we're seeing from those labs, of course, are competitive large language models, you know, the Quen series. And, of course, DeepSeq and the world waiting with bated breath for DeepSeq R2. I should say those of us who are elbows deep in AI are waiting with bated breath for DeepSeq R2. But one of the things that we learned this week was that DeepSeq R2 may well be a little bit delayed because they are feeling the semiconductor controls. They're not able to access the degree of compute that they need in order to build out this model and get this model to the quality that they need.

7:40So you certainly get that sense of pressure present on those models just because of the compute shortage. And I think that that crunch on compute or other capacity is really being felt there. But China is also, the ecosystem is cracking along with more vertical apps, taking advantage of the ability to get lots of users and from lots of users, lots of data. A great example is Apollo, which is the self-driving car consortium. I actually spent some nice time with the chairman of Apollo. It's running in Wuhan, and we've written about it. You should go and check it out in the Excellential View archives.

8:25It's essentially profitable on a unit cost basis, running these cars roughly meant to cost about$11 ,000 across that city. So you've got this example where there is quite an advanced and mature vertical application. But I also had a couple of great conversations in the education and health space. Now, the reason these are interesting, and I think they're really interesting for American audiences, is that we are familiar with the problem of cost disease in services in the U.S., where a whole load of things have got much, much cheaper where you apply technology to them, whether it's cell phones or computers or TVs.

9:03And there are a bunch of things, healthcare and education, which are getting more and more expensive. And one of the arguments from Silicon Valley is that it's because technology has not had a chance to disrupt those areas, bring higher quality at lower cost. And that once it is able to do that, we will start to see really, really substantial, substantial changes in the way those services get delivered. Well, what's happened in China is that there have been examples, some examples of that happening at scale. So I was able to speak to the chair of Yidu Tech, which is a healthcare company, Gong Ruijing, and their scale is quite something.

9:40They've analyzed five and a half billion authorized medical records covering nearly a billion patients. As of last year, nearly 28 million people had completed at least one interaction with their healthcare platform. And it was really interesting to see that scale. And that scale is enabled, of course, because there's a public health system in China, the public health insurance. I'm going to absolutely brutalize the pronunciation, I think is Huiminbao. through that, Unitech serves more than 34 million people in a dozen cities. So you get this point where actually you do see all of the benefits of a large population, which you can serve, and then start to get the data and the interactions from all of this.

10:30Another founder I spoke to, and I spoke to many, was a really impressive guy called Derek Lee. And Derek is the founder of something called Squirrel AI, which is an education platform. Now, we've got these ideas, right? The AI can adaptively deliver kids' education, keep them in that flow state where questions are hard, but not too hard. Progress can be very quickly tested and the difficulty of the lessons can be adapted appropriately to maximize learning rates. Derek was a winner of the International Maths Olympiad. He was an absolutely superb, superb high energy person. And, you know, as he started to talk to me about this app, you know, as a product person, I was genuinely pretty impressed with the way he thinks it through.

11:15So the way they think through delivering curricula through an AI teacher where there's no human teacher involved is to think through how do you deliver a degree of metacognitive capability that might vary from subject to subject. So the art of metacognition, if you're doing maths, is quite different if you're doing perhaps a social science. What that required then was a very, very granular breakout of a series of skills and subskills that then become pathways that are adaptive pathways that students get through. They've served 24 million students. They also give this service away for free to millions of students in the poorest areas of China.

11:58You know, one of the observations Derek made was that there are lots of places where you'll be taught maths in rural China, far in the east, maybe in the Tibetan plateau or something, where you won't be taught by anyone who's been trained in math tuition. And now you can be with Squirrel AI. Some of the numbers that he shared with me of performance improvement were absolutely crazy and off the charts. So I'd recommend you search Squirrel AI and take a look at the way in which they're able to bump up test scores. But there was one thing that really struck me. So Derek comes from a family of teachers, his father and his grandmother.

12:38And he said that in China, there's a competition to find what are called super teachers. And the super teachers are teachers who in their subject are the best out of 3 ,000 teachers. And typically a town may have, you know, one super teacher in a subject. presumably the cities have more. But it's an incredibly, incredibly competitive space to get into your skills, your capabilities, your teaching record, your pedagogy. It has to be astounding to be a super teacher. So Derek tells me that they take some cohorts of students and some get taught by super teachers and some get taught by Squirrel AI where there is no human teacher involved.

13:22And you can see where this is going. But I don't have to tell you the result of this. I mean, the result was that the students who went through the Squirrel AI system had far better learning outcomes than those who were taught by the very, very best teachers in their field. So that was Derek Lee. I think the thing that struck me about Squirrel AI, again, once again, with my product hat on, was the amount of work he had to do to build these specific learning pathways, subject by subject, level by level, while keeping in mind this idea of a metacognitive framework that you would need to learn. And that speaks to the importance of that kind of data in these types of vertical apps.

14:07And it's a lesson, I suspect, that could be brought over when we think about how do we want to deliver AI-based education in other countries, in the UK and Europe, or in the US. Anyway, I was really, really taken in by his story. So again, I encourage you to go and check out Squirrel AI, or maybe I'll try and do something with him at some point in the future. So the third thing I wanted to talk about were electric vehicles and Chinese car companies. So as everybody who reads Exponential View has known for several years. We really thought that the investment in EVs, skipping internal combustion engine innovation, focusing on the EV platform was going to put Chinese car companies far ahead.

14:50They were going to be able to use Wright's law and the learning curve to exhibit tremendous cost leadership. And boy, have they done that. So this vicious domestic market at scale where they have been fighting like angry tigers for years has really, really disciplined them. Now, I live in North London. It really is Tesla central. And, you know, for years on the commute or the commute, when I'm dropping my girls into school, you would see the sort of alternate of a BMW i3, which is that gorgeous electric vehicle they no longer make. And then a Tesla, then an i3, then a Tesla. In recent months, you've started to see some BYDs on the road, not a large number of them, frankly, about as many as you see of Polestars, but there is a BYD showroom down in central London in Mayfairn, but off Barclay Square.

15:41So you're starting to see BYDs here, you're starting to see BYDs on the Uber ranks, but to be fair, you see many more Kia Iconics, and you see quite a lot of VW ID.4s and even ID.7s on the road here. And a previous guest on Friday with Azeem when it isn't just me was Greg Jackson, the founder and CEO of Octopus Energy, and he has just done, I think, the most interesting car deal that I have come across, which is a deal that involves the BYD Dolphin, which is their urban city runaround, BYD, this Chinese EV manufacturer. And what you can do with the BYD Dolphin deal through Octopus and Greg is you'll pay your leasing fee and provided you plug it in enough, you'll get paid or get free charging because it's going to form part of a virtual power plant.

16:37And in order to smooth out grid demand, they will borrow bits of your charged electricity overnight or during the day in order to basically smooth out demand and make the grid work more efficiently. The result is you'll get paid by not having to charge your car. And effectively, it could work out that you get that car for free. Because what you're contributing, you're contributing in a kind of economic democracy in a funny way. You're contributing to the overall infrastructure. And as a consequence, that contribution comes with a car that will ultimately come for free. I actually think it's much more interesting than the robo-taxi model that's been discussed with with Tesla.

17:17I think it's only because I'm a bit squeamish about my car and like who gets to go in it. And I've just seen how people treat taxis. But I think the BYD Optimus deal is really, really interesting. So of course, you know, if you've been reading Exponential View, you know that we really thought something very interesting was happening in the Chinese EV industry. And that the only person who understood it in the US, of course, was Elon Musk. I mean, he deeply, deeply understood those learning curves back over more than 12 years ago. And he unleashed the secret Tesla business plan, which he rather brilliantly put out in public.

17:53He said, you know, my plan is to build an expensive car. So we learn how to build a less expensive car, which we can sell to people who aren't as rich. And then from that, we'll be able to learn more and create scale and build an even cheaper car and sell it to more people. And that's what you're seeing out in China. And it's really noticeable on the streets because Beijing was$36,$38, degrees centigrade. It was really easy to breathe. I mean, it was a bit muggy, right, frankly, it's 37 degrees centigrade. But I was able to go for a run, a 5k, sort of relatively early in the morning when it was only 34 centigrade.

18:25And foolishly, I stayed in the sun because I wanted to adjust my circadian rhythms. But the presence of all those electric vehicles is really noticeable in the air quality. For those of you who want to know how it doesn't feel like, I strongly suggest you head over to London to the south end of the Finchley Road by the Swiss Cottage Tube Station by the Air Quality Unit Monitoring Station at around 7.45 or 8 in the morning and spend about 10 minutes there. And that is the opposite of what it's like to run in Beijing at quarter to eight in the morning, as I found. And that's because of the electric vehicles.

18:57Now, I managed to ride in a number of cars because I was using Didi, which is the Uber, the little brother, the Uber out there. And, you know, I took a couple of Western cars. So I took a couple of Mercedes. Actually, they were Maybachs, which is a sort of posher version of the Mercedes. And I had a bunch of BYD vehicles and a couple of others. And the one that I took on my three-hour journey, if you go to my Instagram handle, which is at Azeem, A-Z-E-E-M, you'll see a picture of it in one of my stories. It was mind-blowing. It was an amazing car. I guess the closest thing to a US This car would be Lincoln Aviator, maybe, if the Aviator is a seven-seater, the slightly smaller one.

19:41And, you know, that is the kind of prestige, you know, drive people from hotels to airports car in the U.S. And that has not a patch on this car that I was in. And in fact, I couldn't even see the brand. It had Harman Kadron speakers. But it had a tan leather interior. It had a fridge, an amazing console for the driver. It was super comfortable, massage seats and all the rest. and really, really remarkable. Drive quality sense of luxury, definitely above a Lincoln, definitely above a modern BMW 7 Series, probably not quite where the S-Class is or maybe not quite where the Range Rover is, but really, really close to that.

20:25And the prices of these cars are incredibly keen. I mean, it's going to be a tidal wave. It's going to be a tidal wave, tariffs or not. I mean, I think European car makers have had a moment of respite because Elon Musk and Tesla have rather fallen out of fashion in Europe right now with sales down 27 % of Teslas, even though the market's up 28%. But BYD has been one of the beneficiaries alongside the other markets like VW. But the cost leadership these firms exhibit is really second to none. So expect to see many more Chinese cars on the roads. I think there are some sensibilities that need to get right around interface and so on, but it's an incredible, incredible drive.

21:08So that's the end of my three observations. One key takeaway that I would give and recommend is that, you know, nothing beats actually going there and seeing it. I mean, a lot of what I've said today may have just reinforced what we've written about in the past. and I had been slow post-COVID fitting in a trip. And I'm glad that I was able to, even though, as I said, it's a really, really short one. And it obviously merits much more investigation. But I would say that a visit out there, as is true for trips to the Bay Area, which I think are absolutely critical if you've not been there in the past year or two, to just get a sense of the velocity of AI and the sense of how much AI is in the air there, just teach you a great deal.

21:56In that sense, it's always good to travel. And I'm glad that I did. Thank you so much for joining me today. We'll be back next week. And I'm now going to treat myself to a nice glass of sparkling water and a lie down to get over my jet lag. Thank you.