AI development prospects in China: advantages and challenges coexist

The rapid development of artificial intelligence (AI) is transforming industries and reshaping the global landscape. Innovations driven by AI technologies are enhancing efficiency, and creating new opportunities across sectors. 

How will the rise of AI transform industries? What are the development and potential risks of AI? And what is the development prospect of China in the global arena of AI? Tune in for some valuable insights and interesting points!

Guest:
Denis Depoux, Global Managing Director of Roland Berger

Transcript

You Zhixin:
Welcome to China Trends. From chatbots to new business strategies, the advances of AI (artificial intelligence) are dazzling. As AI pushes forward a new round of industrial revolution, maybe it's time to rethink how would AI alter the society and us. Let's welcome our guest to share his view with us. 

Denis Depoux:
Hello, I am Denis Depoux. I am the Global Managing Director at European consultancy, Roland Berger.

I've been living and working out of Shanghai for the past 9 years. 

You Zhixin:
Nice to meet you. My first question is, I think everyone might get interested in how would you foresee the innovation or trend that will reshape the AI future?  

Denis Depoux:
Sure. I think 2023 was a kind of a pivotal year for artificial intelligence, with the advent of the large language models, such as ChatGPT by OpenAI. And that's a revolution because it created a natural language interface with the traditional AI. Traditional AI has been there for quite a long time already. But what was missing was the interface with us, with the real world. That's existing now. And since then, I think this GenAI, seemed has attracted massive investment only in 2023. It's 12.1 billion dollars of equity investment for startups in the GenAI field. So, that's considerable. That's 5 times of what was there before. The year before. Also, the tech giants, Baidu, OpenAI, Microsoft, Meta, Google, all invested massively to develop capabilities, also to develop the infrastructure, the cloud infrastructure which is needed to simply enable these larger language models, and also enable the computational capabilities that is needed. But that's only the beginning, right? There's still much progress to do on the general AI side, also on the large language model side to enable new use cases, new capabilities. 

I think a few directions. I think first is in further integration of the large language models with general AI to improve accessibility. A second direction is the shift from, the current models are rule-based deterministic. So, they are basically managing through a set of rules. The next step is to shift to probabilistic models that will integrate a lot more context that will enable more problem solving, more complex problem solving. And another evolution is semantic models that will help interpret language better and basically smooth the interface. That's a lot of efforts around the interface. AI is still at the surface of industrial processes, service processes. So, what comes next is the deeper integration use of AI for all sorts of problem solving.

So, we're only sort of touching the beginning. We see AI use cases for marketing purposes in consumer companies. As an example, our own company is using AI agents for research through our databases, for problem solving or issue. And that is to help our consultants accelerate and increase productivity. Or we're also using AI for administrative purposes. But that's only at the surface of things. So, what comes next is deep integration into industrial processes. 

Maybe a few examples in the pharmaceutical industry. One of the key things is drug discovery, is accelerating, drug discovery. And that's where the combination of GenAI and and traditional AI can yield faster drug discovery, because a lot more data can be processed, and a lot more simulation can be done. 

And I think there is one Hong Kong based startup Insilico that is now in clinical trial phase for drugs that's against autoimmune disease. Everything has been basically AI generated. So, that just shows that it's starting to penetrate processes. We see it a lot in research and development, actually. Basically using AI to increase efficiency, productivity, and speed of R&D, by using AI in complement, and maybe eventually replacing some of the human research. 

Other examples, I think it's accelerating autonomous drive, or it will accelerate autonomous drive. Also, automation in manufacturing, I think is another big field. The energy sector, which is my sector of origin. The energy sector is also something that will leverage AI a lot. For prediction, you know we cannot store power. So, at any moment in time, we need to match power supply with power demand. And that entails predicting power supplies, predicting weather, temperature, that's very complex. And it gets very very complex on any geographical area. That's where AI can help solve the problem, improve the prediction, and therefore enhance the efficiency of, again, existing processes. AI is only at the beginning today. 

You Zhixin:
As you mentioned, AI really brings so much possibility to different sectors. And in your view, which sector or which industry might experience the most profound transformation? Given the AI integration.

Denis Depoux:
Sure. I think anything that has to do with data. So, you would probably tell me, data is everywhere now. But today I’m seeing the service industry, obviously the financial service industry is using a lot of data, is generating a lot of data from its clients. That's where AI can make a difference. That's where also GenAI with the human language, natural language interface can actually help facilitate and automate things that were done by brokers, by basically service or customer service people. 

We see it a lot in anything marketing-related where obviously there's nothing new marketing has been using, has been driven by data already for a long time. But of course adding layer after layer of AI processing of these data, processing of broader set of data, of complex, unstructured data on a consumer market, on a product, on a channel, on the combination of all this, also simulating different pricing schemes, that's already happening. 

So, there are a lot of use cases already around the consumer industry, around the service industry. That's penetrating gradually. So far, it's still at the surface. You might have a lot of data to micro-segment market and come up with, maybe possibilities, to have products that are more adequate to smaller segments, and therefore will sell better. But then you need to integrate this through the whole supply chain process. 

So, how do you design these processes? How do you manufacture these processes? May be on demand. How do you bring this products to the client everywhere? So, it's only the beginning, but the consumer industries, the service industries are the first. I think R&D in general. Everything that is relying on R&D and engineering, which takes a lot of simulation, which takes a lot of integration of data, will benefit most from this industry.

Really. It's everywhere, but some industries will simply take more time. Because the integration of AI use cases in the processes will take more time. There's no surprise. It's been like this with digitalization. If you think of it, everybody sort of was thinking, okay, it's gonna revolutionize. Yes, it does. Actually, we know now. But it took, sometimes a decade or even more, so that every process is digitized. And we can reap the whole benefit of this. 

You Zhixin:
Now, AI mostly focuses on the automated tasks, especially in the initial stage, dealt with very simple tasks. But now we can see it's doing more and more complex tasks. The future world will involve human working with AI. I wonder how can we balance the advances of AI with measures to mitigate the potential risks, like ethical risks or security ones?

Denis Depoux:
Sure. I think it's exactly the dilemma. Because on the one hand, this is the next productivity revolution. And this is what the world needs, right? Grows of the economy just doesn't fall from the sky, unfortunately. And we realize it everywhere in the world these days. So, what we need is the next industrial revolution. So, we went from horses to steam engines, from steam engines to electric engines. All this generated a lot of productivity. AI is probably the, or one of the technologies that will bring the next productivity shift big time. But of course, it comes with quite some problems. Because it will reduce employment, same as the previous industrial revolution. 

What one can also expect that in the same time, it generates employment, if we manage to upskill people. So that they can be the the new users of AI, that they can be the new coders of these AI models, that they can be the new engineers of all these cloud infrastructure that is needed to support, that they can be the engineers that will invent the new chips and manufactures a new chip. 

As usual, a technology revolution and a productivity shift is destroying on the one hand, and creating something new on the other hand, how do you resolve that dilemma? No AI involved. How do you resolve this dilemma? It's policymaking. It's regulation. So, I think, and we can see already, probably not fast enough, not broad enough, but we can see that AI regulation or nascent AI regulations everywhere in the world. Several objectives, data security, privacy, national security. Because all that data that goes everywhere doesn't feel like too comfortable in some cases. That's one aspect. 

Another aspect is sovereignty. So, making sure that this can create a competitive advantage for a nation. That's one source of competitive advantage, like cheap energy is a competitive advantage, like climate performance probably is already and will be increasingly a competitive advantage for nations. So, the AI infrastructure, the whole thing, from the knowledge to use cases, through talents, through everything, exactly. Everything is a competitive advantage that takes policymaking and regulation. 

You Zhixin:
Yeah, you made very great points. AI competition is getting intensifying these days. And I notice not only in the business company, community, not only tech companies, but also different governments. They also want to join the race, to seize the opportunity. What do you think of the competitive landscape in the AI global arena?

Denis Depoux:
I think governments quickly understood the importance of AI and that was, well, before GenAI even appeared, which of course emphasized it. So, I think quite a number of government, I think 75 governments now have adopted so-called AI strategies with one form or another. I think China did that already in 2017. The first country was actually Canada, China in 2017 as well. China was soon to come in that same year. France, my country, adopted the sort of national AI policy in 2018. 

And again, since then, a lot of countries have adopted this. Because they see the competition. Because they see the importance of public funding. Because they see all the opportunity, also of either not losing dominance, or technology dominance, or technology leadership. That's obvious for the U.S., China, they see some countries, or group of countries, like the EU is seeing the opportunity to sort of catching up on something new, while maybe having lost grounds on some other technology leadership. A lot of other nations like South Korea, Singapore, just to name a few, are also playing a big role and positioning with a lot of public funding on R&D, on building the infrastructure as well. 

A lot of infrastructures are needed. You need computational power. You need, basically, data centers. In order to host this infrastructure, these data centers need energy. You need the energy and preferably green energy. So, you can immediately see that this is not something that can be only handled by private companies, or even one company, or even the whole private sector. It is a national strategy. And that's why governments step in. 

A lot of middle-income countries are also trying to step in because they see the opportunity and risk. The opportunity is to position again on something new. And of course they will not match the know-how and investment, and the results of the U.S. or China or the EU, but they may take a certain position. 

They will also mitigate the risk of increasing inequality. So, you have a number of countries that already today have a hard time catching up on the economy, even in some cases lagging behind. But actually, that inequality would even increase if digitalization does not pervade their economy. And within digitalization, if AI cannot pervade the economy, no digitalization, no data; no data, no AI; no AI, probably a big gap in the future with nations that will have it.

You Zhixin:
You named a lot of companies which invest a lot in the AI sector. Among them, there are some from China. Chinese giant companies invested in AI and also the governments invested a lot of infrastructures. What do you think of the Chinese position in the AI global arena?

Denis Depoux:
I think China is data land, right? Obviously, there's no single economy that is more digitized than the Chinese economy. So, there's a wealth of data available. That's true on the consumer side and we all live it every day. But that's also true on the industrial side. That's true on the government side. And when I say government, I'm not talking about sensitive data, I'm talking about traffic light data, I'm talking about health care data in cities, I'm talking about food supply. So, basic needs or traffic data, that's government data which is available, which is already used by quite a number of tech companies. It's consumer data. It's industrial data. As China is now one of the most robotized countries in the world. All these robots, all these industrial automation, is generating data. And I think national policies in China work well. And we've seen it in the new energy vehicle sector. We've seen it in the decarbonization. We've seen it before with 5G. So, I think there's a long history of waves of government policy that promote a certain industry and that's successful. 

In this case, AI is a combination of a lot of things, its knowledge, its R&D, its science, its skills, its talents to also fuel that science. But it's also the infrastructure. So, it's also the cloud infrastructure, the data centers, as I said before. It's a green energy that goes into it. Something that China has to catch up with quite rapidly with very ambitious objective.

You Zhixin:
You've listed the strength of China's AI development and what about the obstacle that you think that China might need to overcome? 

Denis Depoux:
I think the attempt probably temporary obstacles, again, access to technology and development of domestic technology to overcome this. But that will probably not take so much time. However, I would name climate performance as a key. I would not say a hurdle, but already in 23, AI uses 3% of power generated in China and the estimate, it's not from me, it's from the energy agency, is that it would probably go to 5 to 6% within the next couple of years. And that energy today is, let's say, still heavily carbonated. It needs to be decarbonized as much as the usage will also increase. And I think that's an overall objective for China, not specific to AI. But that could be in the global race for AI that could be an impediment. Because climate performance is a competitive advantage for all technologies. 

And you know a lot of companies now demand of their suppliers, that's whatever they produce, is producing a sustainable way. And of course, today the pressure is already high, but there's nothing too mandatory. But as time goes, it will become mandatory. So, China needs to be in that race also on the climate performance. 

You Zhixin:
The Global AI Governance Initiative proposed by China underscores the importance of equality, just as you mentioned, mutual respect and mutual benefit. How do you perceive this initiative？

Denis Depoux:
It's very important that there is global governance on AI. Now, that's a dream because we cannot figure out global governance on a number of other topics that in some cases are even more pressing. So let's be realistic. It's great that there is a global AI initiative in China, and there are a few others, I think the European Union, which already pioneered the data privacy with the GDPR (General Data Protection Regulation), as also adopted the EU AI act. I think earlier this year, and that will come into implementation in 2026.

It's good that there are multiple initiatives. Some with regional reach, like the EU. But globally meaningful. Because these regulations apply to any companies that deal with Europe just like the GDPR. 

I think China is following the same route to also set ambitious goals for the regulation. It's important that these initiatives converge and cross fertilize. Maybe I'm not optimistic enough. Maybe there could be a global AI governance. But meanwhile, if at least these initiatives cross fertilize, I think that's all good, because at the end of the day, if everybody has more or less the same rules with specificity that are related to domestic market differences, I think we can live with that. And that's probably already a decent objective. 

You Zhixin:
Do you think there is possibility that different regions and countries roll out different regulation are referring to AI that the difference between them may create some problems or create some challenges? 

Denis Depoux:
For sure. We see it already in cyber security and data privacy. I think there are several dimensions. I think everybody can agree on the fight against crime or fraud, and we see AI and GenAI being used for fraud with this fake videos pretending to be some somebody else and trying to claim money one way or the other. I mean this is already reality. 

So it's good that there is governance and that governance goes beyond borders, because all these people usually are in one region and trying to fool people in another region. So, I think everybody can agree on the fight against terror, or the fight against crime, the fight against fraud, the fight against disinformation, which is also a concern everywhere in the world.    

Then of course, nations or group of nations also have sovereignty issues. So, they want to make sure that their data stays where it is, or at least if it goes somewhere else, it is under full control and not everything can go somewhere else. China has this, the U.S. has this, Europe has this, many nations have this, and it's of course creating boundaries. But these boundaries are probably necessary or at least reflecting the world as we live in it. Good or the bad news I don't know, it is that companies try to bridge. Companies are multinationals. Chinese companies in this sector are multinationals, American, European companies, all the companies are multinationals. They don't like boundaries, right? So, I think that this is another dilemma between regulation which is needed for security reasons, for national security reasons, for individual protection reasons, and of course, encouraging as much exchanges, trade or investment, cross investments as possible. 

You Zhixin:
Thank you for sharing your thoughts with us and hope to listen to your other deep thoughts in the future. 

And thank you for watching this episode of China trends. See you next time.