4 More Take-Aways from Red AI on Artificial Intelligence in China (2 of 2)

Facebooktwitterlinkedin

This is Part 2 of my 9 take-aways from Nina Xiang’s book Red AI. Part 1 is located here.

#4: China is Great in Electric Vehicles but the US is Leading in Autonomous

China has lots of great companies doing electric vehicles. And they are the largest market for them (and growing). And they are now going international.

But autonomous vehicles appear to be a problem. You see American companies pulling further ahead in this. The book talks a lot about how AI for autonomous requires tons of road testing for training. To train AI for driving, you really do need large, varied and real data sets. And the only way to do this is to put cars on the road and rack up the miles, which is happening in the USA but not China. It’s still the early days so we’ll see how this plays out.

The book doesn’t go into the EV/AV players that much (there are lots of them). But it does mention that Pony, Baidu, WeRide, and TuSimple (trucks) are all doing road testing in the USA

#3: China is Behind in Government Provided Data

China is widely assumed to have a data advantage in AI versus the West. And this is definitely true in consumer data. There are more Chinese consumers and they are producing a staggering amount of particularly useful data (especially given mobile payment). Although the consumer data does tend to be walled off in the digital giants.

But the data situation is not the same on the enterprise side. And it is definitely not the same when it comes to government collected and provided data. The book argues that 80% of data is actually held by government. And while, this is widely data shared in the West, this does not really happen in China. The Chinese government has few real incentives to make its data public. It’s a problem.

#2: Chinese Speech Recognition is Starting to Look like a Commodity

iFlyTek was founded in 1999 by Liu Qingfeng (a PhD from University of Science and Technology of China). It is one of the most famous of China’s voice recognition companies. It went public in Shenzhen and, in theory, has +40% of China market (who knows). Other players in China include Mobvoi andBaidu.

At first glance, these speech recognition companies seem similar to computer vision companies like Megvii and Sensetime. They are all AI-centric and are tackling frontier-type technology questions.

But not all AI is equally valuable. Speech recognition is looking more and more like a commodity service, closer to spell check than Skynet. Once the algorithm has mastered +5,000 words, it is trained and can just be copied. It is more of a static skill. And that is very different than how computer vision must continually analyze new and changing situations. So, unsurprisingly, we are seeing lots of new entrants jumping into speech recognition and producing high accuracy rates. That’s not awesome.

And it gets worse.

Because China’s digital giants (Alibaba, Tencent, Baidu, JD, Xiaomi, etc.) all view voice as another technology for interacting with their consumers. They view it as a support function for their core businesses – and they are willing to do speech recognition at a loss. You can see this in China’s smart speaker business, where Alibaba (Tmall Genie), JD (Dingdong), Baidu and Xiaomi are dumping smart speakers on the market below at cost or below (usually around $15). That is bad news for pure speech recognition companies.

In my opinion (not the book’s), speech recognition appears to be becoming a commodity that lots of companies can replicate – and that the major players are just subsidizing as a support function (not a revenue engine). We shall see. But I suspect the pure speech recognition companies are going to have to license broadly and cheaply – or go after industry verticals with specialization or regulatory requirements. Like legal, financial services, and healthcare. But these are early days still.

#1: HiSilicon is a Company to Watch for Chinese High-End Semiconductors

AI depends on specialized semiconductors, which are mostly made in the USA and Taiwan. And given the current US-China issues, it raises the question: Can China make high-end semiconductors at scale? We still don’t know.

After the US tech ban of Huawei, the entire digital China world saw itself at risk. If it can happen to Huawei, it can happen to anyone. And virtually everyone began focusing on ending their dependence on US tech, particularly semiconductors (i.e., Nvidia) and operating systems (i.e., Android). And while China can probably develop an operating system and a supporting developer ecosystem (which it is doing right now), developing homegrown semiconductors is a real question. It’s a pretty complicated supply chain. And if any company is going to pull this off it is Huawei’s subsidiary HiSilicon.

It turns out the semiconductor industry is really difficult to enter. It is a global market with big technology barriers, rapid advancement and big capital requirements. It takes a long time to build this type of expertise. And as the technology frontier moves rapidly, the leading edge players tend to get almost all of the global market. Additionally, the leaders create technology standards that become entrenched. So even if you manage to enter high-end chips, you are still playing in a game where Qualcomm, Nvidia and others are setting the standards. It is a global winner-take-all market with high barriers for entry.

The book lays out some of China’s very interesting history in semiconductors. The State has actually been very active in this for decades. Lots of initiatives and money spent. Yet China had largely failed to create a semiconductor industry beyond basic chips. Definitely not leading edge ones. Until recently.

HiSilicon is arguably the closest to breaking into the top tier. Huawei founder Ren Zhengfei has actually been talking about not being reliant on US semiconductors since the early 1990’s. He founded HiSilicon in 1991 to address this issue. And it was the first Chinese company to crack the top 10 for global semiconductors.

The book goes through some of the other players as well:

  • TMSC in Taiwan, which is 50% of global chip manufacturing. 
  • Cambricon, a somewhat mysterious project in Beijing.
  • ViMicro, an interesting State-backed venture that rose and fell pretty dramatically. Founded in 1999, the company grew with government subsidies and tax breaks before it went public in 2015. It subsequently crashed and burned.

My take-aways on this are:

  • Most of the major AI players are at risk, especially iFlytek and Megvii, which the US government have already mentioned as potential candidates for the entity list.
  • HiSilicon is the company to watch in terms of high-end semiconductors. They have solid marketshare and are probably one of the few companies that can match Nvidia and others on spending. Plus the emergence of AI may give them a chance to leapfrog the technology and set the new standards.

That’s my take.

Overall, I thought it was a really solid and helpful book. Lots of details at the company level and an entertaining read. Highly recommended. Plus you can get the ebook for only $0.99. It is available here.

Thanks for reading, jeff

———

Related articles:

From the Concept Library, concepts for this article are:

  • Artificial Intelligence 
  • China

From the Company Library, companies for this article are:

  • n/a

———–

I write, speak and consult about how to win (and not lose) in digital strategy and transformation.

I am the founder of TechMoat Consulting, a boutique consulting firm that helps retailers, brands, and technology companies exploit digital change to grow faster, innovate better and build digital moats. Get in touch here.

My book series Moats and Marathons is one-of-a-kind framework for building and measuring competitive advantages in digital businesses.

Note: This content (articles, podcasts, website info) is not investment advice. The information and opinions from me and any guests may be incorrect. The numbers and information may be wrong. The views expressed may no longer be relevant or accurate. Investing is risky. Do your own research.

twitterlinkedinyoutube
Facebooktwitterlinkedin

Comments are closed.