On this episode of News World, how close are we to a crisis? With China and the Taiwan Straits. My guest today is doctor Waifong Zhong, Senior Advisor for the Office for Fiscal and Regulatory Analysis at the America First Policy Institute and an affiliate scholar at the Mercada Center at George Mason University. He leads the open source Policy Change Index PCI project, which uses artificial intelligence and machine learning to read large volumes of propaganda text to analyze
and predict government's actions based on their words. Using PCI, doctor Jeong is out with a new paper entitled Predicting Taiwan's Straight Crisis using Propaganda, a new open source method. Pay Fong, welcome and thank you for joining me again on news World.
Thank you, mister speaker for having me again.
Before we get into China and Taiwan. I have to ask you, because I'm fascinated about the whole process. The system you've now built scans how many items a day? On average?
We catch every news article that comes out of the Chinese propaganda machine, which by which I mean the People's Statey. Not only the people of Statey, the central newspaper, but also now many other provincial and ministerial newspapers from China, so we collect them every day.
The system reached Chinese. Yes, do you also monitor Taiwan?
We also just started monitoring new speed from Taiwan too. It takes two parties to take part in the conflict, which is a big part of why we are not expanding the data collection to Taiwan.
When you're looking at all this material, how fast does your system analyze it?
Well, we catch all these content as it comes out, and then we have in the back end a series of models that ingest these data into the algorithms, and then we do our calculations for different purposes. And for example, we monitor Chinese officials their rights and fall in the political fortune in Beijing. Oftentimes we catch ahead of time the fall of key officials when they got into trouble
with the authorities. Which do you thinking that was showed up in propaganda too or included a rise sometimes you know you see black forces coming up in the scene in Beijing. And the same is true for monitoring Chinese attitude towards Hong Kong, Taiwan, the US on a number of issues.
Given this kind of quantitative analysis, are you sort of looking for word patterns or what are you looking for?
The most important thing we look for is anomalies, things that deviate from the trend or the baseline. And this is very unique when it comes to analyzing authoritarian regimes propaganda because we are not really talking about news, the real fake news, which is propaganda coming out from regimes like the Russians and Chinese. And with these regimes, we should not think of their let's say, their newspaper or their radio broadcast as facts. We should think of them
as the regime's attitude. So what's really important is to measure to catch the moment these attitude changes, because oftentimes indicate the intentions, Beijing's intention, mascots, intentions, and intention.
From that standpoint, how should Americans understand the Chinese Communist parties view of Taiwan.
A good analogy to make it actually to look at the CCP's attitude toward the student protesters going back to nineteen eighty nine, the Tmn' Square protests, and that's an excellent example. I mean, it's a tragedy, but it's also a very good example to understand how propaganda operates. Because initially,
when the students were protesting in Beijing. They were advocating for a more open government, open society to some extent, beaging knowledge that they said, these people they came on with good intention, the protesters, demonstrators, and they all have
a hope for a better future for the country. But then when things got out of hand, it's actually necessary, tragically necessary for the regime to change how they describe these students, they would say, because if they're good people, you can't crack down, you can't send in troops to kill them, right, and so they had to become in some sense, in the propaganda sense bad people. We saw in nineteen eighty nine. Looking back is that the propaganda
dramatically changed how they describe these students. At the end, they became like morons in the newspaper, and that's needed before they could actually send in troops. The same logic holds here when it comes to the Chinese stitude toward Taiwan, because the ordinary Chinese people, they don't really hate the people in Taiwan, but war or conflicts connectic. Conflict requires negative sentiment, requires popular support for the regime to you know,
send missiles over. And so this is exactly what we're monitoring is how the regime if at any point trying to move the public attitude to the negative side toward Taiwan, which is something that happened very significantly in nineteen ninety five to nineteen ninety six for the last Taiwan stract crisis. I know, mister speaker, you had some personal connection to this episode, because I know you went to Taiwan the year after.
Is that nineteen ninety seven.
I think it was in ninety seven, But in ninety six I was on TV saying that President Clinton was right to send a second aircraft carrier into the streets and that we were sending a clear signal to Beijing that in fact, we would fight if they tried to invade Taiwan. When I visited, it was a fascinating experience. But as you look at the data sets you're getting
what can you tell? For example, the military leaders are making predictions, and it seems to me that at times they're predicted that oh this is Chinese commiss military leaders their predictions would imply some kind of combat in the not very distant future.
Yeah, absolutely right, mister speaker. But the same is true for American military leaders too. They're all sorts of estimates, bry predictions about when something like it would happen some set twenty twenty four.
Luckily that has passed. Now a lot of eyes are on twenty twenty seven, twenty twenty nine.
I think the Chinese witterer leaders did the saying, but I think a lot of these military assessments they focus more on the capabilities, like when China has the capability to overtake Taiwan, but that depends on how much interference we see from other parties such as the US, the rest of the Western.
World, and maybe Japan.
So all these are important variables to the capability portion. But I think from again the data, we see more of them.
Do you see the Chinese communists that are showing up in your data sets becoming more bella coost and more leaning towards a military operation, Not.
In a very recent month.
So we developed this AI algorithm to learn from what happened back in nineteen ninety six, arguably the most intensive crisis in recent times, and then we use what we learned from that episode to monitor subsequently former Speaker Pelosi's visit to Taiwan a couple of years ago, as well as the meeting between the Taiwanese president and former Speaker Kevin McCarthy. What appears to be the pattern now is
that there are strong reactions. I think when how he placed US officials visit Taiwan a meet with the Taiwanese leaders. But the magnitude has decreased.
It's almost like what was initially a crisis has now begun to be They grumble, but it's now sort of normal to have visits. I remember originally, boy, it was like a giant crisis. Even to land in San Francisco or Los Angeles was a huge deal. Analysis kind of background grumbling.
Right, if you compare from speaker to speaker, right from you, mister speaker to Speaker Pelosi, speaker of MC coffee, the magnitude of reactions from Beijing has declined, which actually shows a lot of value of demonstrating American strength in this situation. When we show strength, Beijing will think twice about how aggressive they want to be. And that's really the purpose. I image of your visit to Taiwan too.
Back in the days.
You've developed this large language model. Does it actually report back to you in Chinese or does it translated into English.
It reports back to me in Chinese, which is where my human intelligence comes in, is that I would interpret that in ingresh writing to our audience as our readers. But the machine is clellly operating in the Chinese language, not only Chinese language, actually the Chinese culture and Chinese way of news reporting in the system. Because that's alway ingesting to the model.
As an amateur, did I read about the idea that when you have an artificial intelligence model like yours, it then has to go through a pret being trained.
What does that mean?
Let's start with what's called large language models, which is basically models built on every word ever other on the Internet, possibly the big companies they can possibly find right which include Chinese language. But that gigantic lake up data, if you will, it's not particularly helpful when it comes to China, for example, of China versus Taiwan, because it's a very specific context.
So the training in.
This case means if you just ask a question to the AI in Chinese, it would respond to you, but it will be based on the vast number of Chinese language material on the Internet, which included Chinese language that were data spoken on the US Internet. Because they are Chinese immigrants here, but their Chinese language content is very different from Chinese language content coming out from the CCP.
In our case, when we do these projects, training means to steer the AI to focus more on content in the propaganda coming up on the propaganda system, So pay more attention to those and ignore whoever you know Chinese immigrants in San Francisco are talking about in blogs.
So that's not really rather than to the question at hand.
So, as I understand it, as you were developing this, you actually were training it using the people's daily articles during the nineteen ninety five Taiwan Straight Crisis, which led to military escalation by China, Taiwan and the US. I'm curious as you went through that assessment, did you learn anything. Did you see the nineteen ninety five crisis any differently after you got the report back from the machine.
I think what surprised me the most after seeing what the machine returns is how much the road the US plays in the escalation in the whole crisis. Fundamentally, Taiwan is an issue of reunifying with Taiwan is almost a requirement for the CCP region. So that in our assessment remains the number one most important factor, but the number two most important factor was the Chinese attitude toward the United States, and that changed drastically during the crisis too.
So it's not really an issue.
Taiwan is not really an issue just between the mainland China and Taiwan. It's how the Beijing preceved strength of the West, which goes back to the point of when we demonstrate strength, Beijing's reaction to these, even just bilateral visits between the US and Taiwan less and less significantly over the years in recent years, I think that should be counted as an achievement of our foreign policy rather than weakness in any sense.
So, in a sense, a little bit less Bella Coost, a little bit less on the edge of war than they were in ninety five ninety six.
Yeah, if we're just comparing from like a sets from speaker to the speaker in the last few episodes, it seems to be the case. But I also wonder it's hard to distinguish is to what extent the weakness now in Beijin comes from the weakness of the Chinese economy or the Chinese development stage, so to speak, which really has been having a really hard time in the last
few years too. So the difference between Colosi and mccaffee, for example, could be to some extent due to the weakness of the Chinese economy that's trending down since the COVID lockdown was lifted in much of China.
The process that we're in the middle of right now, Jiji panclearly has faced with. Even before Trump and the Terrified, the Chinese economy was in significant trouble, both because the one child policy has now led to a dramatic decline in the workforce, because they had overbuilt entire towns, not just overbuilt buildings, but entire towns. They had over extended
themselves in terms of internal lending. That could either lead him to be more aggressive so people would pay attention to and rally around nationalism, or he could lead them to be more cautious because he would feel like he doesn't have a stable and solid a base. What's your handsh of how those two way out well.
I think in the end, an actual kinetic conflict or even an invasion would have to be an outcome of two things. One is the Beijing's strength internally, because that has to deal with capabilities. But the other intention how desperate Shei Jimpin is in overtaking Taiwan.
Every time I think.
About this, I go back to the notion of legacy. Political legacy a lot in the Chinese context. If you compare Cjimpin with his predecessors, including those you met mister speaker in Beijing, Jian Jamin and his supportinates at the time.
So if you compare for something from Janjumin to Hu Jintao to announce C. Jimpin, you could name a few things as great as a achievements by Janatamine for example, right leading China into the wto taking back Hong Kong and Macau holding the Beijing Olympics to some extent that went to Puljintail as well. But if you think about
cj he really didn't have much achievements so far. It's running the economy down to the ditch and many other things down to the ditch too, And so one argument to be made is that he is desperate in looking for achievements, particularly when his next five year term comes up,
which is twenty twenty seven. I think that has a lot to do with other assessment about why the timeline has to be twenty twenty seven, because by twenty twenty seven he will be seeking another run, right, And even though China does not have democracy, popular support still matters to him, and the lack of it matters even more. And so one could make the argument that taking back Taiwan in the timely manner would matter a lot to his political legacy, because so far he has none. So
that's I think a major factor. But of course he's not going to be able to do that if the Chinese military power and economic power continue to decline, which is why I think it's very important to continue to monitor the situation that weighing these two factors over the next few years to see which one comes out dominant.
Do you see in that sense as being almost like an American politician in that he wants a way out which gives him dignity and makes him important, but he may not want to risk the destruction of the regime in order to get there.
Absolutely, I think in the end, as we always say, right, foreign policy fundamentally is domestic policy, and domestic policy fundamentally is about the legacy or legitimacy. Right in a normal democratic country, legacy takes a form of having more votes right in the next election and more support for the party if you do well.
But in China, I think the bar is much lower.
But if the legacy is low enough, it might tricker a lot of opposite even from within the leads.
Right.
We have seen that actually during the COVID lockdown period when people couldn't stand it, you know, we see university students coming out to protest too, and that makes people think again to look back to find a similarity to for example, student protesters in km and Square. So all these are very unpleasant images for the CCP.
One of the things that you came up with that I was surprised by is that your analysis of mass scale of information shows a real decline in reporting about the outside world by the People's Daily that is just less and less important than the story. What do you think is going on?
We actually just publish a journal article about this journal of Contemporary China, where we actually track how the CCP mentions all the different countries including Taiwan, the US and everywhere else, versus any content that's purely domestic.
So the CCP's.
Vision, I think, in the last almost twenty years now has increasingly narrowed, so their eyes are more on domestic issues less on the outside world. Initially, I think that's because the trading with a lot of outside world, engaging with a lot of outside world becomes less critical, Like, for example, they used to emphasize Europe a lot, but
not right now. I think CCP doesn't think Europe is relevant because Europe has been in some sort of decline, right But in later years, even the attention to the US has declined, and I think that it's not really an indication of the America in decline, is in indication
that Hijinpin himself is more and more narrow minded. And I think that's a very dangerous situation to be in because when you don't even report to the outside world, not to say the reporting is accurate, but when you report even to a lesser extent, it makes.
All the elites in the society.
And everyone else in China more nearrow minded, because that's the only source already right to the outside world, and even that is not covering a lot.
It seems to me like they're beginning to move more and more inwardly oriented and less or less aggressively trying to become a worldwide power because there just have too many domestic problems. I mean, does that seem accurate to you to.
A large extent.
That's not the same to say China's actual activities in terms of engaging with the outside world are necessarily in decline.
We have seen the continued effort to boost.
Up the Baton Road initiatives for example, right China is still trying to forge all these collaborations with the Baton Road countries in Asia, in Africa, in South America. But the way they approach that this projects has also changed. In the past they emphasized more on economic development, but now they emphasize more integrating them into the Chinese values because they come up with this value system that they basically try to turn all these other countries to the
same way of thinking as the CCP has engaging. So that's a narrow minded vision, not necessarily a narrow minded activity.
It's always struck me that they're so good at running into totalitarian system and they've developed all of these cameras and facial recognition and massive ability to track a billion people that if I were a local dictator, I might like to have the Chinese come and show me how to make sure that I'm totally secure, and in that sense, they may have something to offer that we don't have.
Well, actually, the Chinese have offered that it's a speaker to a lot of developing countries.
The Chinese technologies had.
Over the years power to many surveillance systems in African countries, and some of these episodes have come up in the news reporting on how they helped these non democratic regions to monitor their own people. Which is why I think in the US is really a justified concern on a lot of technologies we have here because they could potentially and some did, provide the capabilities for beaging to monitor us.
And we'll talk about Hawwei. We're talking about TikTok.
TikTok's a perfect example Huawei, which is one of the major failures I had in the first Trump administration. I kept trying to warn them that Huawei is very dangerous and very powerful and currently very successful worldwide. It provides access for the Chinese communist that is I think very sobering.
Absolutely, and I think there are two layers of the dangers here.
TikTok is a speaker, it's one the proper ganda through content, so to create a pro China content right to try to influence online. But HAWA is a vulnerability where information itself could be tapped into by the CCP, and we actually have that with TikTok too, because the way TikTok collects information is down to your fingertip, like how you
scroll on your screen and how you select content. And in my opinion, it's actually very damaging in the longer term because when apps are able to monitor our behavior, it would be able to learn our preferences to how we think about things, or how American voters think about different niches, and knowing that it's incredibly valuable because that's a huge intelligence.
It has a huge value on intelligence.
I buy a lot of books on Amazon, and it now in two it's what I might be interested in based on my behavior. So the giantic system that Amazon has is sort of studying me and figuring out what to offer me in a way that is on the one hand, is very practical and very helpful. On the other end, and it's a little scary.
It's actually very scary. Even Amazon could potentially learn your schedule, mister speaker, like when you would scroll on the app to shop they would figure out your sleep patterns, you know when you couldn't sleep in the middle of the night and you're trying to buy some gadget on Amazon. Now imagine all that that every company in the world is doing that. But imagine a world where every company
submits that information to the CCP. That's their vision of governance, the global governance, and we should try very hard to prevent that from happening.
So do you think you'll extend this new methodology to looking at other countries?
We have.
Last year we put out an algorithm that monitors came young to North Korea, So we have gathered the North Korean content and we made the similar algorithm to detect whether kingdom or is thinking differently.
Given how closed the dictatorship is. Is that very revealing doing it that way? I mean, do you pick up queues and indications?
Yeah?
I wasn't sure when I started the project, but it performs surprisingly well. For example, we picked up signs of significant nuclear weapon tests shortly before in some ways that you somehow just changed the tone anten on their front page.
So that model we did was as off last year.
So this year we're going to random the model live in the sense that we'll be like getting the content from opinion every day and updating the algorithm every day.
So that's one of our next projects too.
Ye when you're doing some fascinating stuff, and I hope maybe an ear we can come back and you'll read us into all the different things you're learning, because you've found a way to really have a big impact.
I think thank you, mister Seeker for saying that I'd love to come back. Also, let me clarify the AI is doing most of the work now, it's no longer me, which is a very reassuring development.
I am personally happy to still believe that the human had a role in this thing. Makes me feel better. Waite Fun, I want to thank you for joining me and helping us better understand both your system, which is amazing, and how to analyze and think about China's plans for the Taiwan Straits. Our listeners can follow your work by visiting the America First Policy Institute's website at America First Policy dot com. And it's really been very enlightening and I really appreciate you visiting us again.
Thanks again for having me Missless speak here.
Thank you to my guest doctor Waifong Zhong. You can read his new paper predicting Taiwan's straight crisis using propaganda, a new open source method on our show page at newtsworld dot com. Newtworld is produced by Ganglish three sixty and iHeartMedia. Our executive producer is Guernsey Sloman. Our researcher is Rachel Peterson. The artwork for the show was created by Steve Penley. Special thanks the team at Gangwish three sixty.
If you've been enjoying Newtsworld, I hope you'll go to Apple Podcast and both rate us with five stars and give us a review so all this can learn what it's all about. Right now, listeners of Newtsworld can sign up for my three free weekly columns at gnglishwe sixty dot com slash newsletter. I'm new Gnglish. This is Newtsworld.