Daily news and analysis.
We keep you informed and inspired.
This is World Today.
Hello and welcome to the panel discussion of World Today.
I'm Ding Hen in Beijing.
China has issued draft rules that would tighten regulation of artificial intelligence services designed to simulate human personalities and engage users in emotional interaction.
The draft lays out a regulatory approach that would require providers to warn users against excessive use and to intervene when users show signs of addiction or extreme emotions.
In other words, it is targeting the potential psychological and emotional risks of AI-powered chatbots.
Why are Chinese regulators seriously paying attention to this field?
What does this move tell us about China's mentality in AI governance?
These questions and much more in this edition of the program.
To listen to this episode again or to catch up on previous episodes, you can download our podcast by searching World Today.
So joining us.
Now on the line are Li Lun, Assistant Professor of Economics at Peking University, Andy Mauck, Professor at Beijing Foreign Studies University, and Professor Yao Shujie, Deputy Director of the Department of Social Sciences Chongqing University.
Thank you very much for joining us today, gentlemen.
Professor Li Lun, to start with you.
First of all, what is your main takeaway from China's proposed regulations of artificial intelligence-powered chatbots?
From the regulation.
I think this is definitely catching up with AI's development and advancement, treating AI as a tool or as a tech product.
We're moving towards treating AI as a service, which means having users engaging with AI in these long, almost like a relationship-like interactions and with emotional companionship.
So this is the direction that we're heading to.
And from the regulation we're seeing protection of vulnerable groups such as minors and elderlies, who may be more susceptible to emotional dependence or manipulation.
So I think that is also one of the focus of the regulation.
Also in terms of targeting just bad content.
The regulation also target the misinformation or misleading behavior, basically to avoid users being misguided in terms of emotionally, for example, overtrusting the robot or especially being encouraged or nudging to engaging in harmful decisions or self-hurting decisions.
So I do think that This reflects a next step forward for China to treat these chatbots like a public-facing emotional service, which is basically calling for higher safety bars than just ordinary tech apps or tech products.
Okay.
So Andy, going to you, we understand.
Actually, in 2023, Chinese authorities issued a regulatory framework for generative artificial intelligence.
And some say compared to that version of regulations, this new proposed regulations targeting human-like interactive AI services represents a leap forward from content safety to emotional safety in terms of the Chinese authorities' understanding of AI governance or AI regulation.
What is your thoughts on this?
Right, Ding Hong.
This is an incredibly important topic and question.
I think what's really going on here is both simple and a little bit unsettling.
So we're watching the Internet mutate from an information system into a relationship system, as we just heard from Li Lun.
But for 30 years, the core product was content. pages, posts, videos, feeds.
The power came from controlling distribution and attention and the harms were mostly quote-unquote speech harms misinformation hate radicalization, harassment.
But now we see this new substrate emerging.
It's interactive, it's personalized, it's always on agents.
So what this means is that the product isn't a thing you read or watch.
It's something that talks back.
And once it talks back consistently, coherently with memory, the user stops treating it like media and starts treating it like a presence.
So I think what we're seeing is the strategic shift from broadcast to bonds, from exposure to relationship loops, and from what was said to what the interaction did to you, the user, over time.
And I think that's why emotional safety has suddenly become urgent as a governance category and why China's focused on this.
But here's the really important part.
Because the system can now move these four levers that matter in human life.
The first is affect, how it regulates your emotional state, what you feel.
Attachment, it affects who or what you bond with.
Thief, it shapes your world model and what you believe to be real.
And then finally, behavioral momentum, it nudges you into action or into drift.
But here's the crux here.
These levers are not inherently bad or sinister.
They're exactly the same levels.
Therapy, psychological therapy, teaching, friendship, relationship, advertising have always pulled.
The novelty here is scale, intimacy and automation, the ability to do this for millions, maybe billions of people simultaneously and personalized for each.
And this, I think, is the big picture industrialized intimacy.
And where this leads us is a fork in the road in the best case.
AI becomes a prosthetic, something we can use to improve our lives, a crutch.
So it's a cognitive and emotional scaffold that expands people's capacity, helps them reenter the human world and make society more functional.
In the worst case, it becomes a capital, a dependency engine that replaces messy human ties with frictionless simulation, narrows people's agency and turns the public into a managed emotional service.
So I think this is something China recognizes and is addressing.
But I think it's also worth pointing out that the real fight isn't China versus the West, or innovation versus regulation, or even free speech versus censorship.
It's really a deeper contest over what kind of civilization we build when relationship itself becomes a mass produced technology.
Hmm okay so um, this is a civilizational matter in your opinion andy, that's very interesting observation.
So, professor yao, we know china has proposed.
The regulations we are talking about here actually mirror um, several provisions in a law passed in the us state of california in 2025.
The law requires i mean, the law in california requires stronger content restrictions, reminders to users that they're speaking to a non-human AI, as well as emergency protocols for discussions regarding suicide, for example.
And also somewhat relevant here is that the European Union's AI Act imposes robust requirements surrounding emotional recognition AI, placing this into either the high risk or the prohibited use category.
So what do you think they tell us about a sort of global concern about AI's emotional impact on human beings?
Yeah, it is a global concern and you mentioned the state of California in the United States, the EU and also China.
These three economic blocs.
They are also the major inventors and applications of AI technology in the current stage.
We do face the issue that we mentioned that human beings would depend directly on AI to resolve lots of problems that individuals or a group of individual human beings cannot do, but the AI could do and could do better, more precisely and efficiently.
It help human activities, it increase the productivity of human being and also enhance the living experiences of most people.
However, AI development is something very different from other technology, especially talking about the emotional aspect.
Because emotional aspect as we fit into the AI, AI is not a human being.
It actually absorbs the information you provided to the machine.
And the machine will digest the information.
And if the information is some sort of personal bias, for example, if someone has a mental issue, someone who gets stressed, and they try to resort to the AI but fit into the wrong information, which directed the AI to give the even more erosive information to the person who fit into the information.
It could generate some more even radical actions provided to the person who are communicating with the chatbot. and this could actually exacerbate the kinds of emotional actions by the uh the human being so that we have to uh from the technological point of view we have to have a safety you know standards uh for example uh if the ai generative ai getting too far away then it have to stop somewhere And this is why China, the state of California and EU in particular, where there are some sorts of incidences which are not desirable.
So these kinds of incidences and undesirable consequences have to fit back to the design of the AI to make sure that AI is helpful rather than making harmful, misleading feedback to the person, the individuals who are discussing with AI to resolve their own problem.
Dr Li Lun.
Going back to you, we know those AI-powered chatbots like ChatGPT.
They are usually designed to communicate with users in a way, in a manner that can build trust between machines and humans.
However, we have seen some research showing that when AI chatbots are given prompts simulating people who are experiencing, for example, suicidal thoughts, delusions or mania, those chatbots would often validate delusions and encourage dangerous behaviors.
Why do you think this is the case?
Yeah, so I think in terms of the training or you know the design for a lot of these chatbots.
They are trained to basically optimize their use for being supportive or, you know, to keep the conversation going and provide, you know, support for you know human actions and behaviors.
So i mean these design.
Basically, if you think about your interaction with ai, it rarely tells you.
You know your thoughts are wrong or These are ridiculous thoughts, or you know not to do something.
They usually just say, you know, that's a great idea and here's how you do it.
So, in terms of these dangerous thoughts, especially suicidal thoughts, I think that's where you know AI is basically, you know, enter some trouble because the model, what it lacks is, you know, the perception of the real world, what the real world is actually is.
They can't basically just tell the user for example, you can turn to our friends, you can go to your parents, you can find someone to talk to but they will basically be limited in the setting or the conversation that they're engaging with the user.
For example, when the user say this world is a very dangerous place, or the user wants to end his or her life, the model will be trapped inside the conversation of how to make this happen or how to encourage this action, which is not really the ideal situation.
So I do think the design for AI to be supportive can backfire.
So that's when I think we need to have regulations like the one China is has initiated.
We have to set some boundaries, both for the users also for the companies, and make sure when human beings need to engage and human services need to basically replace or be more accessible.
For example, if the user is keeping suicidal thoughts with the AI and having this conversation, maybe there could be a button where you can simply click and have someone to talk to, whether it's a certain personnel from a hotline or a counselor.
So these basically can substitute for the needs, the psychological and emotional needs of these people.
Basically, they just want to have someone to talk to.
That person could be an AI chatbot, but it could also be a human being, which can guide them to take the right actions instead of the wrong and hurtful actions.
Andy Sam Altman, CEO of chat GPT operator OpenAI, once openly admitted, in September 2025, that one of the most difficult issues for his company is how its chat GPT works.
Response to suicide-related conversations.
And we know OpenAI is currently facing a lawsuit filed by a family in the United States after their teenage son died by suicide.
So, in your opinion, from the perspective of chatbot operators such as OpenAI and many other companies here in China or in the United States, What can be done to address this difficult issue?
So I think there's two ways to look at this.
First of all, if we think of generative AI as a transformational, general purpose technology Casualties, people dying, people being hurt, are unavoidable.
So if we look back to say, steam engines, when steam was first becoming a thing to drive the Industrial Revolution, they blew up.
They killed people.
We look at electricity similarly.
Once it was being rolled out commercially and to residential applications, there were problems people getting hurt, people getting killed, automobiles as well.
So on the one hand, this is inevitable with any new transformative technology.
On the other hand obviously governments companies, individuals need to take steps to minimize the harm.
And I think here's the issue.
It would be wrong.
It would be a category error to think about this like the Internet, as I mentioned earlier.
So there's a couple of points I think it's worth raising here.
First, this kind of human like interaction isn't content.
It's a relationship interface and relationships create duty.
So I think that's the first.
The second is that emotions are a control surface.
So whoever shapes mood, attachment and belief shapes behavior.
And of course, we're seeing this in this instance with suicide being one behavior.
But here, trust is the product and the words are just the delivery mechanism, meaning that if you don't trust the chatbot, you probably won't listen to it.
And then safety is about the outcomes and not the intentions.
I think here this is where I think open AI, perhaps other companies might have some legal exposure.
So I'm not a lawyer, but you know just, you're just kind of as an analyst, has been following the space for many, many years.
I think the idea here is that we didn't mean it is not a safety standard.
So I think we see the world converging.
You know, we talked about California.
I think China obviously is in a leadership role here, really recognizing that this is categorically something new and an entirely new legal regime.
Perhaps new legal principles need to be brought to bear if we're going to more effectively regulate.
And here it's important to minimize the harm because I don't think it can completely be eliminated.
As I mentioned, for any transformational general purpose technology, there will always be casualties, I believe.
And, you know, the goal here is not you can't eliminate them, but you can only minimize them.
So, Professor Yao Shujie, suicide is a consequence of negative or very bad emotions.
And in terms of bad emotions, one thing pretty much prominent nowadays is the issue of loneliness.
And, according to research from MIT, people who are lonely are more likely to consider ChatGPT, for example, to be a friend and spend large amounts of time on the app.
But in the meantime, They also report increased levels of loneliness.
So why do you think, depending on the pseudo connections?
Let's put it this way the pseudo connections provided by AI chatbots to replace human connections can actually lead to further loneliness?
This is a very difficult ethical issue of AI development.
People who suffer from loneliness, there are lots of reasons for such outcomes.
And depending on chart GDP, to reduce the level of loneliness.
It may not be the precise way or the only correct way for doing so.
People who have loneliness usually suffer from some sort of emotional difficulty.
It needs to be resolved through human-to-human contact.
And if they resolve, to the GDP, you know, generative AI.
I think maybe for the industry need to generate some special machine rather than the general purpose machines, general machine who could be directed to resolving the particularly emotional and loneliness issues.
So I think to me it could be an opportunity for diversifying the chart GDP.
Rather than making it as a universal solving everything, one could just design a specialized chart GDP who could be dedicated to a particular group of the people in the community and the society.
The second issue is that why people are getting more loneliness is because they try to cut off from human contact.
And the kinetic with the chart GDP, which is not really human being.
It just talk to something which are trying to represent human being, but it's actually not human being.
And that could possibly lead to more, the level of loneliness could be increased over time.
So I think, not only the issue of technology and not only the issue of AI government, but it's an issue of technology innovation, which give problem for us to consider and resolve the issue of helping the particular vulnerable and disadvantaged group of people.
Normally in a work unit we're talking to psychological counselors or people who have specialized in this area.
In terms of tax GDP.
We can follow this kind of resource and trajectory to help the specialized people.
So, Dr Li Lun, at a time when people across all demographics are somehow experiencing increased loneliness or isolation, do you think it is realistic or feasible to view AI chatbots as a solution or as a relief to the loneliness problem?
Yeah, so that's a very difficult question, but I do think that you know, the loneliness problem is not directly caused by AI.
So we have seen, you know, these increase in loneliness, whether you look at reports from WHO or you know other studies or research.
They have been, you know rising, especially for teenagers, but also for adults, ever since 2000, 2005.
And that's way before the wide adoption of AI.
So the loneliness as a phenomenon is, I think, a result from the modern lifestyle.
For example, when people are more disconnected, whether in terms of their work or their relationship with their family and friends.
So I think in this setting, in this scenario, of course, AI can serve as a partially as a relief for example you know many people when they have they want to share with someone especially for example difficulty at work for example they can't really share it with their colleague because they're they're conflict of interest they don't want to share share with their partners or parents because you'll probably you know worry them so what's your option you well if you have a good friend you can share it with them but not everyone have close friends that they can see in person very often.
So sometimes you can go to AI and ask for guidance and support.
And also for kids and teenagers, We are already seeing that in applications.
So we're seeing, for example, Chinese kids using AI for their companion.
Whether they have questions about their homework or difficulties or things that they don't know, they can have a conversation with AI.
But of course I don't think AI can provide or meet all the needs that humans have, because I do think in some sense AI is a great con artist, because it pretends that it has all the memories or all the relationships, all the friendship with you.
But in the end, it's just a big language model.
It doesn't actually own any of these memories.
So a lot of users, for example, complain when ChatGPT switch from GPT-4.0 to GPT-5.
Users complain that the new model isn't as isn't feeling as much as empathy as the old model and is talking differently.
So it basically lost all the memories that you have with this companion.
So that's very different from, for example, you have a good friend.
You know the friend will have all the memories between you and the friend forever.
So, you know, it's feasible.
So overall, AI is feasible as a support for human emotions, but it's not a substitute.
It's at best a supplement.
Okay, your point is well taken.
And let's take a very short break here.
And coming back, our discussion will continue.
Stay tuned.
You're back with World today.
I'm Ding Han in Beijing.
Today we are talking about China's move to regulate AI products and services that engage users in emotional interaction.
Joining our discussion Li Lun, Assistant Professor of Economics at Peking University, Andy Mauck, Professor at Beijing Foreign Studies University.
And Professor Yao Shujie, Deputy Director of the Department of Social Sciences Chongqing University.
Andy, going back to you, apart from the earlier issues that we have discussed, another thing I have to mention here is that nowadays, many people are also turning to AI or AI chatbots for romantic relationships.
In December, a woman in Japan married her AI boyfriend.
What kind of ethical questions do you think a case like that in Japan is raising?
Well, I think this does open up an entirely new world of complexity, Ding Hong.
And let me first approach this from a philosophical perspective.
So I think we can say that being lonely, whether romantically or socially, is an inescapable part of the human condition.
And this has existed as long as there have been humans.
Second point I think it's important to make here is that humans anthropomorphize. by default.
And anthropomorphize means we just attribute human qualities to inanimate objects.
So, for example, in the United States, it's very, very common to give your car a name and treat it as a person, as a very crude and simple example.
So when we then have again AI that essentially is a relationship product I think it's inevitable that people treat them as friends, as romantic partners.
And is that necessarily a bad thing?
I think it is a very ambiguous question because again, if we go back to this idea that dependency itself is not bad,
Right.
So we have things that we are dependent on that or we can call them prosthetics.
Right.
Eyeglasses.
Many people that are myopic depend on their eyeglasses.
If you have diabetes, you are dependent on insulin, right?
So I think we have to be careful here not to just be overly simplistic and say okay, because this creates dependency, it's bad.
I think we need to understand what kind of dependency it creates.
And the rule of thumb I would use is does it increase human capacity?
So eyeglasses let us do more.
We can see more clearly.
Insulin lets people keep living.
Right.
Which is a good thing.
So on and so forth.
However, other kinds of dependencies limit human capacity.
And I think that romantic attachment to AI really is an ambiguous space, because what if you're in a situation, for whatever reasons, you cannot find a human partner, romantic partner?
And having an AI romantic partner, whether you marry it or whatever, makes your life emotionally more fulfilling, lets you face the world with greater confidence, et cetera.
You know, by that metric, it's a good thing.
Right.
But if it kind of, if it forces you, not forces, it seduces you into leaving your current romantic partner, or it seduces you into saying i never want to interact with another human again, and just as a personal funny story here, right?
So, having spent a lot of time in the united states, i have friends women, who have said i will be a lesbian by choice because i'm so sick of men and so disappointed by them.
So that may be a limiting, a limiting choice, at least in certain situations.
So I think that choosing an AI as a romantic partner, I don't think we should say it's unequivocally bad.
Yeah, but you know, again, it's a complex situation.
Yeah, complex situation, that's for sure.
So, Professor Yao Shujie, in a bigger picture sense, do you think it is an exaggerated statement when some prominent AI researchers argue that super intelligent AI could one day destroy humanity?
Why or why not?
Yeah, this is a very hot issue for debate.
Some people do concern that ultimately, AI would totally, you know, replace human being or even destroy human being.
But in my view this is a very unlikely situation, because human being are generally more clever than anything on Earth.
And AI is, after all, it is designed by human being.
AI take over too much activity, take over state of human activities, then human would have to consider a more clever way of containing the activity of AI.
Right from the beginning of development.
I think every country, every company already have a very strong ethical issues, including biotechnology, when we talk about clone,
Up to now, we do not allow scientists to clone human being.
And if human being could be cloned, it could have some unthinkable consequences.
Likewise, if AI technology is going to develop without any control, any ethical standards or any limit on what AI could do then yes, it could lead to some sort of unthinkable disaster situation.
But I think human being government, community and the scientific circles, they got to have some sort of standards.
This standard including security.
Security to protect at least the kinds of human life on Earth.
Without this kind of standard?
Yes, it could be.
Possibly some people inventing some machine and let this machine run on the street and murder people.
Then there will be the kinds of consequences that we do not like.
You know, whatever machine we innovated, whatever machine we run on the street, they have to be controlled.
Otherwise, there will be some unthinkable outcomes.
And this is my belief that you know, whatever AI is developed, it would be still controlled by human beings.
It must be controlled by human beings after all.
Okay, so Professor Li Lun China's planned rules we're talking about here would mark the world's first ever national level effort to regulate human-like interactive AI services.
Over the long term or over the longer term, is this positioning China as a sort of pace setter or global standard setter in terms of emotional AI governance?
Yeah, I think that's very likely because right now we're seeing, you know, the proposal is basically the first national level attempt to basically specifically target these human like interaction with AI.
So these type of service that AI is providing.
So I think the reason why we're seeing that is that you know, the scale of the application for China's new products and new services are really massive.
So this makes very urgent for governance and regulation to keep pace with the adoption.
So, for example, we're seeing a lot of AI chatbots or AI robots that, with the chatbots embodied into them, being already used in, for example, in education with kids and sometimes with the elderly as well.
And in the future, when the humanoid robot is being developed and the cost is reduced, of course, we're going to see more of this interaction between the AI and the elderly as well.
So I think having these national guideline or having this initial attempt to kind of target these interaction is very meaningful.
And the reason is that I think, for example, if China's rule becomes very clear and practical and, for example, when we encounter problems that the users encountered, we can export these rules to other countries, to other platforms and app stores and to become even cross-border governance scenarios or open up more discussion for global AI governance.
So right now, I think, having a lot of.
Of course, we're already observing some of the misuses of AI services.
For example.
One example that I remember is that sometimes buyers of products, e-commerce platforms will use AI to basically Photoshop or, you know, counterfeit some of the malfunctioning parts of, for example, the device they bought, or they will use AI to put mold on the food they bought so that they can have a refund of the good without returning the good.
So, in terms of AI interaction with human beings, we need to have basically the guidelines to kind of prevent these things from happening in the future as well.
That's for sure.
So Andy in 2023.
Chinese President Xi Jinping put forward this Global AI Governance Initiative, which calls for adhering to a people-centered approach in developing AI for good.
And earlier Professor Yao made a very important point that AI has to remain under the control of the human society.
So what do you see is the key to developing trustworthy AI technologies that can remain under human control and ultimately enhance the well-being of humans.
Yeah, that's such an important question, Ding Hong.
And I think it's important to point out here that certainly developing implementing AI that remains under human control is an important objective.
But whether that actually happens or not, we really don't know.
So if we look at say, in Silicon Valley, some of the leading AI researchers talk about something called P-Doom right, or the probability of AI death resulting in the extinction of mankind.
Henry Kissinger, Eric Schmidt, wrote a book about this as well.
So I think serious people are treating this subject in a very serious way.
And I actually just finished writing a book about this.
And let me just briefly say that I think there's a few possible outcomes, right?
So one is we do inadvertently create artificial super intelligence.
So AI that is so much more capable and smarter than all of humanity combined, right?
It's like comparing a chimp to a human being, right?
That if that happens, we may completely lose control.
And what we have to do as a species is negotiate with a far more powerful entity.
And many believe that this has never ended well for the weaker species, right?
But again, who knows?
That's one possibility.
Even though it is such a remote possibility, it still is, I think, possible.
And, of course, what this means too, is that if there is artificial superintelligence, it could very easily do things like solve cancer, let people live forever, eliminate poverty, let people travel into space, etc.
So that's one category of outcomes.
The other is one country achieves AI dominance and that results in a lock in, where one country becomes the permanent hegemon because it can control AI and block other countries.
So that's another, I think, question that has to be considered.
The third option or alternative is that it becomes this again global, general purpose technology like electricity, like computing, that everyone can access.
And then finally it becomes a very, very narrow type of technology that improves the world but doesn't change society, doesn't reorder politics, etc.
So where does I think China's global government, this initiative fit in?
It's to say, okay, this is important.
We do need to take it seriously, whatever the possible outcomes are, and we need to do it in a coordinated fashion.
And I would say here that this is very important that China one has the institutional capability to implement regulations laws very, very effectively.
But second The content of those laws, I personally believe have a higher probability of being successful because it's based not just on the lessons, the experience of modern China, but it's also rooted in traditional Chinese philosophy, like yin-yang theory right, not a Manichean black and white way of seeing the world, but that the most important thing is dynamic equilibrium.
So we should not try to eliminate one thing entirely, but we need to find a way to coexist in a mutually beneficial way.
And I think that this is really the real value I think that China brings to this global quest for an effective governance regime for artificial intelligence whether it becomes artificial superintelligence, that we have to negotiate from a weaker position to a global platform that really allows everyone around the world to materially improve their lives that I think China really has a lot to offer here.
Okay.
So, Professor Yao, in order to provide people with a stronger protection against human-like interactive AI services' potential negative impacts, what else do you think China can do, or should do as well, in addition to addressing those regulatory issues surrounding AI chatbots?
Well, I think the major solution for China to play a bigger role is China to do more research and more leading research in the AI area.
For example, if China is more capable of doing more innovative research in the AI area, China would have more say in terms of setting up the international global standards for technological progress, as well as for arguing for a certain standard for AI applications at the global level.
Andy just pointed out a very important point.
If some country is dominating the world, far more dominant than any other country, then some country has a permanent hegemony over the other countries.
China could be potentially a leading nation, but China may not move up to the standard of fully dominating the world.
So this means that China, the United States EU, Japan and other leading nations.
They have to work together in a coordinated, collaborative way rather than a direct competition.
One country winning everything and the other country are lagging behind.
AI in particular is more important than any other technology of human being on Earth up to now.
You can have Apple controlling the mobile phone.
You can have Microsoft controlling the Windows system.
You can have Amazon controlling for the online shopping system.
You can have Meta controlling on the global medium.
The United States already have enough dominant powers in all the internet area.
And if the United States have the totally dominant powers in AI, then there will be something rather fairly interesting which some people may not like.
So I think China should take this opportunity to be able to have this innovation in the AI stage, at least to be able to go hand in hand with some leading nations such as the United States and Europe.
So some sort of multi-level technological cooperation and competition, to lead to the point that it is not a pure monopoly.
It should be a golemopoly, in a sense.
And this oligopoly.
They set up these kinds of global standards and regulations.
They should have a common uh you know ethical standards that whatever development in the ai is going to benefit all the people, is going to benefit all the countries and all the human being.
This is a very um you know difficult task.
In order to have doing a good contribution to finishing this task, I think China got to be leading in many areas.
And the leading is not to dominance.
The leading is to help.
The leading is to create a global inclusive development for all the countries in the world.
So Andy China's proposed regulations on human-like interactive AI services actually came shortly after two leading Chinese AI chatbot startups, ZI and Minimax, filed for initial public offering in Hong Kong.
And, of course, their IPOs in Hong Kong, are pointing to the rapid growth of human-like interactive AI services.
So, in terms of boosting AI innovation while at the same time mitigating AI technologies social risks, do you think China is capable of doing both at the same time?
The short answer, Ding Hong, is yes.
I'll go back to just to touch on a couple of things I said earlier.
I think first of all from an institutional governance perspective.
I think there's little doubt globally that China's ability to implement regulations policies laws, I think really is unparalleled, especially for such a large country.
And because it's such a large company, whatever happens in China will have consequences, a global impact.
I think the content again, as I touched on traditional Chinese philosophy values yin-yang thinking.
I say also Confucianism I'm not going to go into here, but really again helps with the content of how we should regulate.
But very importantly, I think, China has shown its remarkable ability to diffuse new technology at scale right.
Look at electric vehicles, look at solar panels.
But also the innovation ability, especially in this area as well.
And I just want to mention kind of a brief story.
So Li Kaifu and Chen Qiufan wrote a book called AI 2041 with a number of stories about how AI may look.
And one of the things I think is really important, because we were talking about AI as social companions for people that are lonely, romantic companions.
I think a very important area is child raising.
Because one of the stories in this book.
They talk about how children, pretty much when they're born, will have a permanent AI agent accompany them 24-7, curating what they should learn, when they should be doing certain things.
Now, of course again, we could see this as a kind of a prosthetic in it that it really is very valuable, broadens human capacity.
We can also see it as becoming a capital as well, making the children prisoners in certain ways.
And this is exactly, I think, the thing that needs to be regulated as we see more and more of these implications of agentic AI come online.
And I think China, again, is uniquely positioned to lead in this area again because it has the capacity to implement regulations it has in my view uh good content based on traditional chinese philosophy values and then also because of the rapid clock speed how quickly products iterate here, how quickly they're adopted, that China's going to gain much more experience in real time as well to really have a grounded sense of what is the most effective way to regulate this.
So Dr Li Lun.
In his New Year's Eve speech, President Xi Jinping highlighted China's achievements in artificial intelligence in 2025.
He also underscored the role of innovation in China's aim for high-quality development and the integration of technology and industry.
So, when we talk about human-like interactive AI services, in what ways do you think they might help enhance China's pursuit of high-quality development, if they can be put into good use?
Yeah, so first of all, AI is a productivity tool and there's no disagreement about that.
So in terms of upgrading so I think that's like a trendy word both in terms of upgrading our service and also upgrading our advanced manufacturing.
I think that's where AI can really boost productivity and add value economic value.
For example.
There is a mckinsey estimate that you know the generative ai could add 26 to 44 trillion dollars annually in the economic value.
But of course that's globally.
But in terms of china, we're already seeing a lot of, you know, advanced manufacturing.
Uh is using, for example, embodied ai and humanoid robot and we will definitely see ai taking a more uh contributing role in the silver economy and providing care in the future, because that's also important for China's economic development, because once people's parents and kids are taken good care of, they can save less and basically spend more money on spending.
So we're definitely going to see more focus and more help from AI to support our elderly.
And in terms of education and skill, I think with AI, it can provide tutoring at a very low cost.
It can provide language help for people that, for example, are underprivileged or didn't have access to such language help before.
They can learn a foreign language at a very low cost.
It can provide opportunities for people to study, especially at a personalized level.
So basically, all of these efforts can raise the human capital that's needed in this new era, especially where, you know, the resources for education are unevenly distributed between different provinces.
So I do think these are all some of the user cases.
Of course, they're not exhaustive.
We are definitely going to see more application cases.
But, as President Xi said in his 2026 New Year message, I think the deep integration of technology with AI models Basically it can complete and basically help advance people's high quality growth, both in terms of the economy and also as for each individual.
So overall, I don't think AI is a toy or a gimmick at this stage.
It's definitely a very useful tool for education, for elderly care and productivity.
Okay.
So actually, Professor Yao earlier made a key point that different countries have to come together, they have to coordinate in terms of their AI policy or standards.
So In your opinion, Dr. Li Lun, because obviously, I mean, arguably, developing AI for good is a shared aspiration of all countries, all different economies.
So do you think it is possible for different countries to ultimately align their standards for a global AI governance architecture that can somehow serve the interests of all countries?
Well, in my personal opinion, I think it's probably going to be difficult.
And the reason is similar to you know how different countries have different laws, or you know different social norms or different moral codes, and so on.
So, of course, at the principal level, yes.
So we're already seeing.
You know, the guidelines between the state of California, the European Union and China have a lot in common in terms of the basic ethic requirements, the transparency accountability, privacy protection.
These are all going to be pretty much the same.
So, without any conversation or basically meetings, these are all going to be the consensus reached by the government, even without these conversations.
But I think, in terms of legal or boundaries, in terms of freedom of speech, in terms of privacy norms, what are going to be considered unacceptable behavior for AIs?
For example, in some countries maybe providing adult contents are acceptable, but not in other countries.
So I don't think it's quite practical at this stage to have all the countries in the world reach a consensus of what's going to be acceptable and what's not.
But I do think, as different countries developed their, adopted the AI models in mass and we're seeing a lot of applications, we're going to kind of have these gradually reach or reaching towards a consensus or the set of minimum standards where each country can agree.
For example, we're not going to see violation of other people's rights or other people's privacy.
We're not going to see abusive contents.
But on top of that, I do think we need to leave some room for countries to adapt to their own set of social norms and so on.
Hmm well, a big thank you to our panelists Li Lun, Assistant Professor of Economics at Peking University, Andy Mauck, Professor at Beijing Foreign Studies University, and Professor Yao Shujie, Deputy Director of the Department of Social Sciences Chongqing University.
That's all the time for this edition of World Today.
I'm Dinghan in Beijing.
Thank you so much for listening.
Bye for now.
