Ep. 69: China AI with Jeff Ding

Episode 69 of Tech Buzz China features our co-host Rui Ma in dialogue with Jeff Ding, a Rhodes Scholar and D.Phil Researcher at Oxford in the Future of Humanity Institute. He is also the creator of a free weekly newsletter called China AI. For his talk, Jeff focuses on artificial intelligence in China, specifically, some of the “unsexy” technical applications of AI across several industries.

This is the second in a series of experimental, non-scripted episodes that we will be releasing this summer. Today’s episode is a lightly edited version of a live webinar that Tech Buzz hosted on June 5. To hear these (and more) as they happen live, you can sign up for free at techbuzzchina.com/events.

As always, past transcripts are viewable at pandaily.com and techbuzzchina.com. If you enjoy our work, please do let us know by leaving us an iTunes review, and by tweeting at us @techbuzzchina. We also read your emails, at rui@techbuzzchina.com and ying@techbuzzchina.com. Thank you to our growing community for your always valuable feedback!

We are grateful for our talented producers, Caiwei Chen and Kaiser Kuo, as well as SupChina production associate Jason MacRonald. We hope you enjoy the episode.

Transcript

(Y: Ying-Ying Lu; R: Rui Ma ; J: Jeff Ding)

[0:00] Y: Hey everyone! Happy longest day of the year, at least if you’re in the Northern Hemisphere like we are. Earlier this month, on June 5, Tech Buzz hosted a really interesting webinar on the topic of China AI. My co-host Rui Ma spoke with Jeff Ding. Jeff is a researcher at the Future of Humanity Institute at the Oxford University, and he’s also the creator of a popular weekly newsletter called China AI, which you can find, and subscribe to for free, at chinai.substack.com. What follows is a lightly edited recording of Jeff’s presentation.

R: A quick reminder to visit us online at techbuzzchina.com, where you can sign up to be on our mailing list, subscribe to our biweekly paid Extra Buzz newsletter, or just generally keep tabs on our ventures beyond podcasting. For example, we are working on an e-book on the most talked about Chinese internet company this year — Bytedance — and just published our second annotated transcript of an interview with the CEO, Zhang Yiming. Take a read, it’s super interesting!

Y: As we mentioned in the last couple of episodes, we are taking a break from our regular programming for another couple of episodes, and in the meantime we’re experimenting with new formats like this one. Definitely let us know your feedback!

R: And if you miss our voices, don’t worry, we have been pretty visible being guests on other podcasts. I was just on the Use Case podcast for example with Ravish Bhatia, which generally explores the start up world in India but kindly made an exception for us because I think there is genuine interest from Indian entrepreneurs and listeners but also Chinese internet companies have just been relentlessly pursuing expansion there. Definitely check out the show and the episode!

[2:42] R: Hi everyone! We are TechBuzz China by Pandaily, powered by the Sinica Podcast Network by SupChina!

Y: We are a biweekly podcast focused on giving you a peek into what’s buzzing within the tech community in China.

R: We uncover and contextualize unique insights, perspectives and takeaways on headline tech news that don’t always make it into English language coverage. So you can be smarter about the world of China tech. TechBuzz China is a part of Pandaily.com, an English language site that tells you “everything about China’s innovation.” I’m one of your two co-hosts, Rui Ma.

Y: And I’m your other co-host, Ying Lu. We are a part of the Sinica Podcast Network, created by SupChina! In addition to Tech Buzz, you can also find Sinica which covers current affairs.

R: And we are also proud to be partnered with Financial Times’ Tech Scroll Asia. It’s a newsletter on Asia tech news from one of the best publications in the business. Go to ft.com/tech-scroll-asia to sign up today!
And as always, we are still looking for more reviews on Apple Podcasts! Send us a screenshot of your review — and yes, past ones count as well — and we will gift you a free three month subscription to our Extra Buzz newsletter. Just email us at rui at techbuzzchina.com. Thank you so much for those of you who’ve taken advantage of the offer.

R: So hey guys, today we will be focusing on AI in China, specifically, some of the very unsexy but true facts about the industry, and also a list of things that you didn’t know but should about the topic. It’s my pleasure to introduce our speaker of the session, Jeff Ding, who’s prepared a whole presentation on the subject. I’ve never actually met Jeff in person, just over Zoom, and that’s because most of the time he is in Oxford, where he is currently finishing up his PhD as a Rhodes Scholar, as well as working as China lead for the Centre for the Governance of AI at the Future of Humanity Institute

R: Jeff is one of my favorite people to follow on China AI, and I find that he is one of the most thoughtful thinkers on it. For our listeners on the podcast, you are not gonna be able to see his presentation, but we’ve made sure that you will be able to follow along just by listening! However, you can also find a recording of this session on YouTube as well by going to the Tech Buzz China YouTube channel. So without further ado, here we have Jeff Ding.

[5:24] J: Thank you, Rui. I’m a fan of the Tech Buzz China podcast. I like to put it on for a run. It’s a nice bite-sized 20, 30 minute thing and I like how they make it really light and digestible, but also there’s a lot of information. I was listening to the Bytedance one yesterday.

J: It’s exciting to be here. I’m flexible in terms of how we do this. I have a bunch of slides. The first five are going to be about what I call “unsexy” China AI and looking at the companies that aren’t often covered in the Western news media, or aren’t as flashy or visible consumer companies. And it also involves what’s happening in cities outside of the first tier cities– what’s happening in cities outside of Beijing, Guangzhou, Shenzhen, Shanghai.

J: Then, I have a bunch of other slides, where I give the standard “10 Thoughts” I have about China’s AI development to policymakers in DC.

J: Awesome. Let’s dive in.

J: This is slightly blurry. This is from a previous China AI translation from a media platform called Lei Feng (雷锋). The English translation is Leiphone. It’s one of the sites that I constantly follow, I subscribe to. It’s one of the Subscription accounts on the list of WeChat Public Accounts, and they regularly pump out information on China’s tech scene, smart industry. I loosely call them the “MIT Tech Review” of China.

J: They regularly pump out really high quality articles on China’s AI scene. They did a ranking of the best future AI companies, or the companies with the most growth potential. And you can see a list of these companies. It gives you a good sense of the diversity of the scene.

J: You have the strong tech giants, you see Huawei and the row of AI security and under the column of “Best Fortified Growth,” so Huawei, tech giants are up there. There are branches of tech giants on there.

J: You have the standard facial recognition, the big four facial recognition recognition companies. If you look at the “AI + Medicine” row with the “Product Growth” vertical, you have Yitu (衣图)’s precision medicine app. And then you have other well known startups that have been covered, like Squirrel AI. Momenta is a strong computer vision company in AI in cars and commercial growth. So some of these names are standards.

J: Some of the names that stuck out, like Cloudwalk (云从科技), I covered in the recent issue of China AI, that’s also one of the “Big Four” computer vision startups.

J: Some of the names are less standard. One that stood out to me was Ultrapower (神州泰岳). I’ve never heard of them before. And they are one of the few that appears twice, right? The best in “Future Growth” both in 2018 and 2019. 神州泰岳, I think is how you pronounce that. So why is this company interesting?

[7:55] J: Well, first of all, it’s been around for 20 years. It’s one of these long standing companies, it probably has a lot of good relationships with the government. It’s worked closely with China Mobile before in the past, I think they developed one of the first IM messaging apps.

J: But what’s more interesting about this company is it’s doing something that’s not going to make headlines. It’s not some cool face swapping app or deep fake thing, but it’s doing a lot of the back-end enterprise processing, which is going to be, I think one of the big areas where it’s going to make increases in productivity. So, stuff like building a natural language processing NLP banks, and LP factories for companies to better streamline how they manage all their operations, how to make smarter the processes behind software updates, very unsexy stuff, but just coordinating everything that happens in a company better and more efficiently.

J: That is a lot of what Ultrapower does. And I think these types of companies, other ones that I’ve covered include State Grid, which is a state owned company, one of the biggest in China. They actually are number one in terms of the patents that they filed in AI.

J: Now some of this is inflated, because there’s different incentives for state companies to file patents, but I would still say some of the more well-known startups probably have better technical capabilities, but something like State Grid, is it going to be at that kind of second order, who is adopting and then specializing some of these more general AI algorithms for specific applications, like State Grid in terms of flexible energy grid management.

J: The other aspect of “Unsexy AI” is– I think that making knives better is really, really interesting. I think it’s one of the key applications of AI in China. Especially if you look at the July 2017 plan. This was part of my deciphering “China’s AI Dream” report. If you look at the language carefully, a lot of people, including myself, we have trouble translating this… I forget what the exact term was, but it was something that I translated as “gross output.” And that’s a term that’s very much a manufacturing indicator. And there’s a heavy emphasis on manufacturing… economic forces, is the driving force behind this plan.

J: It’s very much rooted in this idea of trying to escape the middle income track. And one of the key things here is, a lot of times when you read Chinese writings on AI, they’re not necessarily comparing themselves to the U.S. as much as U.S. media is, where everything is wrapped up in a U.S.-China two-player game.

J: In this article from jiqizhixin.com (机器之心), which is pumping out a lot of good, long form articles on the tech scene from their sub platform called 机器智能. And you can get good English readouts from there from the platform Synced (syncedreview.com), they do condensed translations of their work.

J: They had this long article on computer vision and machine quality inspection on production lines. And see the quote from the CEO of a small startup that’s making this comparison not between China and the U.S. on these areas, but between Germany, South Korea, Switzerland in terms making stuff like this knife better.

[10:50] J: In this article, it says that in these production lines of just something as simple as making a knife, 20 to 30 percent of the human capital is spent on just inspecting the knife for defects as it goes through the production chain, and that’s where Shuzhilian (数之联) comes in. It’s basically saying, can we apply computer vision to improve our ability to get China to a higher level of efficiency in manufacturing?

J: That’s a very clear example of where you’re actually getting to productivity growth. I think that’s important. We talk about facial recognition, and it’s definitely important, it’s going to be spread across a lot of verticals. But where is the actual impact on productivity growth, which is the key driver of long term economic growth? I think this sort of stuff is also really important.

J: Last thing about “Unsexy AI” is, can we talk about things outside of the top four cities? They did a ranking of who are the top five cities in terms of demand for computing power? That’s a good proxy for who’s running the most training, the most AI algorithms, and who’s doing inference on the most AI algorithms? The two dark horses that they found were Hefei (合肥) and Hangzhou (杭州), two non-first tier cities. This draws from a report that I contributed to recently by Nesta (nesta.org.uk), where I looked at these two AI ecosystems.

J: The key part of both of these cities is they have a key anchor tenant tech company. For Hefei, that’s iFlytek. For Hangzhou, that’s obviously Alibaba. And then they also each have an elite university that glue the ecosystem together. For Hefei, that’s the University of Science and Technology of China (USTC), and for Hangzhou that is Zhejiang University. I think both of those are Top 10 universities, top tier universities.

J: And in each of these cities, there have been consistent local level, provincial level funds to support AI development. For Hefei, they’ve strategically targeted speech recognition because they’re just not going to draw the global talent that’s needed for a comprehensive AI ecosystem. For Hangzhou, they’ve been able to diversify in a lot of areas, because they have more extra-regional linkages, and they also are able to attract local talent. Just because it’s a better place to live in, higher living standards, et cetera.

J: What’s important here is, a lot of times we focus on who’s going to build the next Silicon Valley, which is important. And we focus on the “Silicon Valleys” of China, like Zhongguancun (中关村) in Beijing. What’s also important in a lot of these countries that are trying to develop AI capabilities across the board, and if you think that technological diffusion is important, it’s about: how do you connect the Silicon Valleys to the Detroits of the world. And how do you have clusters outside of the Silicon Valleys? So these are also developments that I think are interesting.

[13:33] J: This is the start of my standard presentation of “10 Key Points” that I talk about when I think about China and AI governance more broadly. Number one is: China is now and will remain an indispensable actor in AI governance. I use “indispensable” to reference Madeleine Albright’s term about the U.S.’s role in the world. Which is just not the idea that it consumes everything and that it has to be the policeman of every single sphere of governance, but without it, nothing can really happen. And I think that applies now to China as well.

J: This is from Mary Meeker’s most downloaded slide deck every year, where she looks at Internet trends. The top 20 worldwide internet leaders from five years ago, you see only two Chinese companies. And then, today, which I think is 2018 in the context of this slide, you have nine Chinese companies in the top 20 worldwide internet leaders whereas the rest are U.S. companies.

J: This is imperfect, but a good metric, because a lot of these companies also have AI labs, some of the best AI labs attached to them, doing fundamental research in this space. It’s one of the first areas where AI is applied relatively easily is through internet data, consumer internet data, social networks, recommendation algorithms.

J: I think the argument is, we can debate over how far along China’s AI capabilities are, but I think the reason why I stress that China will remain an indispensable actor is even if the capabilities aren’t advanced, other countries will prop up Chinese AI capabilities to justify actions in the space. This is an interesting leaked memo about a centralized 5G network, and the reasonings behind that. Why there’s a push to centralize the U.S.’s 5G network is because they were fretting about China’s dominance of artificial intelligence. That is a scene-setting case.

J: But that falls on to my next point, which is that Western observers consistently overstate China’s AI capabilities. This draws from testimony last summer that I did for the U.S.-China Commission, and where my conclusion is basically, China’s not poised to overtake the U.S. as an AI superpower. I think that oftentimes, when we try to measure national AI capabilities, we just don’t have a good sense of what we’re trying to measure.

J: It obviously matters for different things. If you’re measuring AI capabilities for overall productivity growth, I would be measuring different things than if I were measuring national AI capabilities for who’s going to have the most advanced weapon system to be able to apply autonomous control, so I would use different indicators.

J: What I’m using here is very broad drivers. I try to separate things out along three cross sections. One is, what are your scientific and technological inputs and what are your outputs? Some people think outputs are a better measure because you’re actually seeing them translated into actual value. Some people say inputs are better because they say you can’t capture all the outputs perfectly. So you might as well see what’s being put in in terms of: How much talent are you investing in? How much R&D spending are you putting in? Rather than what are your patent and publication costs.

[16:39] J: A consistent finding around here is that China leads in the raw counts. If you control for quality, and how much these publications and patents are cited and forward-cited by other people in quality, the U.S. is still pretty far ahead. And we’re still uncertain about things like R&D spending.

J: Another thing you have to divide things by is you have to look at different areas of the value chain. A lot of Chinese government documents complain about China’s lack of foundational AI open source software. I actually translated a whole white paper that they did on this very subject where they said 66 percent of open source software is developed in the United States, and they view that as a key weakness.

J: Finally, I think there are some areas where China is ahead in some subdomains and in some aspects of the value chain. Facial recognition for surveillance is an obvious one. But also, you have to make clear distinctions in the space between Chinese NLP and English NLP. The “L” in NLP makes a difference in terms of which companies have better capabilities. It’s obvious that because Chinese companies have more of a demand from Chinese consumers, they would specialize in Chinese NLP.

J: Okay, number three.

J: China is not a monolithic actor, and this distinction is meaningful for key issues in AI governance. Here I just want to focus on the companies. There’s a lot of disagreement between bureaucratic departments too, that I mentioned in my report Deciphering China’s AI Dream. One of the big ones is MIIT, which is the Industry and Information Technology Ministry versus Ministry of Science and Technology, over who does technology planning. In the Nesta piece I talked about how a lot of the outlays on AI planning are coming at the provincial local government level. So those are other important distinctions to make.

J: But I want to focus on the companies. There’s this concept of the “national team” in AI and a lot of framing of the national team, which has been around this outdated notion of national champions, or governance, governments are directly subsidizing companies or directly own the companies. Whereas a lot of these new national team companies are largely private, multinational companies with international ambitions. They don’t just want to compete in the Chinese market.

J: But I think you have to make distinctions. So the first batch of national team companies, Batch 4 companies, was the BAT companies and iFlytek. And iFlytek is interesting in that of the first batch of national team companies, it was definitely the most national. They rely heavily on government subsidies. They were born out of USTC, the university I mentioned earlier, and that is a university with very strong connections to the state.

J: Why does this matter, which company is more national? Well, with the increase in techno nationalism, the increase in ensuring data security, and tying in the increasing linkage between technological issues and national security issues, I think there is a trend towards– at least on the part of the Chinese government– wanting to indigenize innovation, maintain critical information infrastructure, within the scope of domestically invested companies. And you see some of those dynamics play out as well in terms of which companies are invited to set technical standards in the space. I think there’s a reason why Sensetime very quickly moved one of their headquarters to Beijing. They’re originally a Hong Kong company, but I think there’s a lot of incentives for Sensetime, one of the big facial recognition startups, to be the Chinese company.

[19:57] J: You see these distinctions among, even the four big computer vision startups now. So the recent issue of China focused on CloudWalk, which is one of the four, you have CloudWalk, Yitu, which we mentioned before, Sensetime, and Megvii. This article made the argument that of all these four, Cloudwalk is the most national. They analyzed this based off of, where they are getting their funding from, are they getting their funding from state funds? Or are they getting them from overseas VCs? And Cloudwalk is getting a lot of their funding from state funds.

J: Also, where do you list? Megvii was exploring listing on Hong Kong, which is kind of in the middle ground, whereas there’s rumors that Cloudwalk wants to list on this new STAR Exchange, which is a science and technology specific market off of the Shanghai Exchange that’s trying to incentivize these companies to provide to domestic investors.

J: I mentioned technical standards. I like this example because it shows that even Chinese companies that are supposedly more aligned with the state will also fight with each other. Technical standards have a lot of cross-cutting cleavages, and Chinese companies will often ally with their strategic partners, international companies to vote against standards pushed by other Chinese companies, which was the case with this example of Lenovo and Huawei, where Lenovo voted against a Huawei proposal.

J: Number four, China is not an impermeable national container. This distinction is meaningful for key issues in AI governance. I like to give this example of Microsoft Research Asia. Microsoft Research Asia is one of the key cultivation training grounds of China’s AI scene. You have 500 plus, key employees of different big tech companies, AI startups that spend time in Microsoft Research Asia. There’s a famous resonance paper. It’s one of the key innovations, one of the fundamental advances in AI. It was used in, I believe, AlphaGo. It’s a way to layer a bunch of neural networks together.

J: It’s interesting to see where the four co-authors end up. One of the co-authors Kaiming He, he’s now one of the lead developers at Facebook AI. This is a very technical, techno global example, global flow of talent. A lot of talented Chinese researchers who work at Microsoft Research Asia eventually help Microsoft or help Facebook or come and work for Snapchat. But also, three of the others went and started their own computer vision startups. Momenta, Megvii/Face++. We’ve mentioned some of these companies before, hopefully some of the threads are tying together here.

J: I also mentioned this point before, so I won’t belabor it: provincial and local governments will play an outsized role in implementing AI. And this is the case with overall innovation policy. If you look at actual outlays of R&D spending, the majority is coming from local and provincial governments, not from the central government. So the input implementation is all happening at the local level, and this proportion is only rising.

J: This goes back to the “unsexy AI” point, of what can we look at? Can we look at local ecosystems outside of the top tier cities, and digging deeper into the Hangzhou and Hefei case, you had China Speech Valley which was designated a national level AI industrial area all the way back in 2012. So this predates the 2017 plan by five years. This investment in AI did not start in July 2017 with the national plan; our attention to it started back then.

[23:05] J: Hangzhou AI town is another interesting one. One of the issues in this space is you never know what’s actually happening on the ground, and whether funds that are announced are actually being spent. What’s cool about Hangzhou AI town is they actually have a website, which I drew on for the Nesta report that it’s mentioning, where they actually list hey, here’s all the subsidies we gave. And here’s all the funding and the specific projects and which companies and how much funding we gave. It’s divided into office expenses, research expenses, and then a key component is cloud computing expenses, right? That’s important for training and running algorithms.

J: And that third vertical of cloud expenses was really interesting because the provider of all those cloud services all except for one project was Alibaba. So it’s very clearly just a way for Alibaba to anchor the system, but I think they added one for Tencent Cloud Computing, it wasn’t all Alibaba Cloud Services.

J: This stuff is where we have to get to, we have to get even more detailed than this. How do you actually get to the level of Hangzhou AI town and see what’s actually being spent, and what those specific projects are on the ground?

J: The next three points are on different drivers of AI development. One of these is hardware. To give you a sense of why hardware is strategic: I’ve mentioned training, the difference between training and inference multiple times, and different companies will supply different chips for training and inference. So Sensetime, for example, the World’s Most Valuable AI unicorn, their training comes from Nvidia, a U.S. company that designs GPUs. And then the inference, the end device implementation, comes from Qualcomm, another U.S. company, which is why you’re seeing compute targeted as a strategic asset in some of these discussions now.

J: But even the U.S. is not completely independent in hardware. I specified that Nvidia is a chip design company. They have a partnership with TSMC to actually manufacture and fabricate the chips, and that’s based in Taiwan. So even that GlobalFoundries, that number two on that chart, which is described as a U.S. company, that’s I think, controlled by a UAE state investment fund. So you start to see some of these complicated cross-cutting cleavages and linkages in the space.

J: Talent is another key driver. There’s a three-part takeaway here, and I think it holds true although I need to update the data, but the three parts are: part one, while China’s researchers are climbing the ranks of international competition, and conferences; B, they are still not publishing the best of the best fundamental AI research; C, though this could change given China’s huge base of talent.

J: This is something I did with MacroPolo, and using a nice database called CES rankings where I think it’s better than citation counts, because it’s just looking at publication counts and top conference venues. So there’s no citation rigging that can happen. And it’s just, are your academics publishing in the top three conferences? Here’s the trend from 2013 to 2017, you see a clear increase in some of these Chinese universities.

[25:59] J: The b sub-point to this is the best of the best fundamental AI research is still at U.S. institutions, largely. And this is looking at specifically NIPS, which has been renamed NeurIPS. That is the top level forum for a lot of the fundamental research. You can see that a lot of the list is dominated by U.S. companies. The first Chinese entity that pops up is Tsinghua University at 22, I believe.

J: But the c-sub point is that this could change given China’s huge base of talent. So if you look at those NeurIPS papers that we’ve just looked at, and you dig deeper into one of the actual authors, because that data from the previous slide about affiliations, was just scraped based off of just the affiliation of the paper authors. But then, if you look at where these authors went for undergrad, that gives you a rough proxy of where they’re from– or at the very least, if they went to undergrad in China, there may be more of a likelihood that they could eventually go back to China– or they have choice to work in different places, two pretty vibrant AI ecosystems. So China does have this larger base.

J: Where his paper number 19 on “Ternary Gradients to Reduce Communication in Distributed Deep Learning” was coded in the dataset before as “people who represent U.S. AI talent.” But again, this is a technical global cross-cutting landscape. And this represents a lot of researchers who China views as part of their huge base.

J: Cold War arms race analogies do not translate to U.S.-China competition AI. I’ve talked about this a lot in recent China AI newsletter issues, so I won’t give my spiel now. But my Year Two interview was basically all talking about how, like I think this meme of– I call it like a glorified dick measuring contest in terms of who gets technological dominance in AI.

J: Basically, I think we have to talk about more things besides who has more AI between the U.S. and China. There are other countries in the world that exist, and China obviously views this as not just a two-player game. I talked about how economic benefits are the primary driving force. That’s my opinion, I think general purpose technologies, like AI historically, have the most fundamental impacts have come through the economic realm. And then eventually, they may have shaped military affairs, and then eventually, they also shaped social governance. But there’s just so many different applications with a general purpose technology that the economic incentives are too great. And I think, ff you look at that as the central driving force, that helps you unlock a lot of different things.

R: Jeff, to wrap up, a question I have is: what are some topics you think aren’t covered enough in the English language speaking side, that you wish were covered more? Or narratives that frustrate you? I think we both have many of those.

[28:35] J: Yeah, well another way to apply this is to look at where all the dominant narratives are herding people to go. One dominant trajectory right now is to view everything that happens with respect to China’s development through the lens of: who’s going to gain technological dominance, the U.S. or China?

J: There was a great Bloomberg article this morning about how China’s whole new infrastructure push– which is a really important concept— new infrastructure, there’s some real money behind it. But the headline is, “China’s new effort to steal the crown jewel of technology from the U.S.,” whereas there’s a lot wrapped up in the new infrastructure push. It’s not about competing with the U.S. There’s a lot about like, is this a stimulus amid COVID? Is this an effort to protect critical information infrastructure?

J: So look for things that go outside of where the dominant trajectories are going, and I think that’s a way to have this idea of information arbitrage that I’ve been trying to push.

R: Yes, well, thanks to efforts like yours. We do a lot of that too, at Tech Buzz. If you look at our transcripts, we are primarily citing Chinese analysis.

R: Thanks again, Jeff!

Y: Well, that was the end of our session with Jeff! What did you think about what he had to say? Send us your feedback!

R: Thanks for listening and don’t forget to write us that review for your free Extra Buzz subscription. Have any questions? Email us! We really enjoyed putting this together, and we are always open to any comments or suggestions. You can find us on twitter @thepandaily, @techbuzzchina, and my personal Twitter account is RUIMA.R

Y: And my Twitter is now spelled YINGLU2020. Tech Buzz China by Pandaily is powered by the Sinica Podcast Network on SupChina. Pandaily.com is an English language site that tells you “everything about China’s innovation.” Our producers are Caiwei Chen and Kaiser Kuo. Thank you for listening!