Bill Drexel is an associate fellow for the technology and national security program at the Center for a New American Security.
Federal Newswire
How did you start studying artificial intelligence and China?
Bill Drexel
I grew up in Houston. Nothing particularly AI-centric about my upbringing. In college I started to look more into international affairs and I really focused on South Asia and only went to China for a comparative perspective. While I was there, I kind of stumbled into this Xinjiang dystopia and it was just one of the most chilling experiences of my life.
Since then I've really focused on how artificial intelligence can reshape our world in a lot of ways that could be really nefarious, and on how the Chinese Communist Party is driving that train in a very dark direction.
Federal Newswire
What were your takeaways?
Bill Drexel
I went to Xinjiang at a really opportune time. This was just a few months into the Uyghur repression story hitting the mainstream news. This was the beginning of October 2018. I've had friends [go] subsequently and [they] had a much more difficult time getting an unmediated view of the area.
I went as a tourist. I had been studying Chinese state surveillance of especially urban Protestants in urban China. Seeing all this news starting to come out about the Uyghurs and Xinjiang I thought, “I have the opportunity, I better go and just get a glimpse of it.” We stopped in Urumqi, which is the capital, and a couple other tourist route destinations.
The main place we stayed was Kashgar, which is where I learned the most. It's kind of the cultural center of the Uyghur homeland, and it's also seen some of the most dramatic surveillance state apparatus because of its cultural heritage–also because the whole city has been raised, rebuilt, and optimized for this sort of technological repression.
Federal Newswire
What observations did you make of the surveillance technology being used in Xinjiang?
Bill Drexel
Part of it is that the repression in Xinjiang goes back further than this technologically based surveillance boom. There has been a lot of surveillance since Xinjiang was absorbed into the PRC. That was another amazing thing about my research. I ended up interviewing a lot of people from Kashgar, who left in the early to late 2000’s, who spoke a lot about how they endured quite a lot of surveillance without high tech mechanisms. The technology that's now mature in the United States in large part helped China, [and] has just turbocharged pre-existing trends.
Before you had different ways of making citizens and family members report and keep tabs on each other. Now you've got a more automated system where they're able to link that more analog human system together. With facial, voice, and gait recognition, with all of these tools they're able to leverage control.
Federal Newswire
Why did the CCP decide to deploy these surveillance systems to Xinjiang as opposed to other areas?
Bill Drexel
I think it's a combination of a few factors. One is that the Uyghur are ethnically distinct from the majority Han Chinese, and they're also cordoned off in a far western branch of what is now China. Another element is that whereas Tibetans, for example, have the Dalai Lama who fled China and has acted as an international advocate for their cause, there hasn't been that same [symbol] for the Uyghurs.
Those factors combined with the religious makeup–being from an Islamic culture, the CCP was able to paint this as an Islamic Terrorism counter-effort.
It's true that there have been some terrorist events in Xinjiang that the CCP will point to as justifications for their crackdown. But if you look more closely at those events they're much more to do with inter-ethnic tension and the Chinese state trying to tame the region than they are about religion.
Federal Newswire
China’s current sprint to develop Artificial Intelligence isn’t the first time their hastened progress has resulted in bad outcomes.
Bill Drexel
There are quite a few examples that stretch back to the early days of the PRC right up to the present day. The biggest and most obvious is the Great Leap Forward where Mao Zedong, the leader of the Chinese state at that time, was obsessed with passing up the Western world in industrialization and steel production. [He] pushed the population to collectivize farms, melt down the farming equipment, and do this sprint to produce steel.
The upshot was the worst famine in human history, with an estimated 33 million individuals starving to death.
There are a few other examples worth noting in the present day. I would point especially to the Belt and Road initiative. There's been some great investigative reporting that points to how several of these mega-projects in impoverished countries are quite literally falling apart and foisting greater debt onto nations that are already struggling with their economic situation.
The one child policy is another one of these attempts to rapidly slam the brakes on population growth, and the results have been catastrophic for the country. You have a widespread gender imbalance. We’ll never know how many forced abortions and instances of infanticide resulted.
There's a demographic aging crisis across the country that the CCP is struggling with. There's this long history of top-down acceleration efforts that backfire tremendously.
Federal Newswire
Can you explain your assertion that our current position with AI is analogous to where we were with nuclear weapons during the Cold War?
Bill Drexel
It seems to be an increasingly understood fact that artificial intelligence will be the pivotal technology in the decades ahead, including in geopolitics. There are definite parallels to the nuclear race in that regard. There are also some important differences to state upfront, one is that nuclear technology had two applications: nuclear energy and nuclear weapons.
Artificial intelligence is very multi-purpose, much more like electricity. So the ways in which it will revolutionize war, and will revolutionize economies and so on, is much more diverse. A kind of consensus is emerging among tech industry and geopolitical leaders that it will be pretty fundamental to competition in years ahead.
Vladimir Putin noted that whoever has the advantage in AI will rule the world. We have similar statements from Xi Jinping. There's a whole chorus of figures in open societies arguing as well that AI will be fundamental to power in the decades ahead.
Federal Newswire
How has the Chinese government's use of these technological tools evolved?
Bill Drexel
In open societies we generally have the advantage in what's called Generative AI, tools that can create things like art [and] free expression. Where the Chinese excel is in surveillance technology, image and voice recognition, these sorts of tools where the government is especially interested in cashing in on the revolution.
When China took its autocratic turn under Xi Jinping the tech went with it, and that's really when things started taking off in terms of a lot of these technologies being let loose throughout China and increasingly around the world.
Federal Newswire
How much of these tools in China resulted from America’s ingenuity?
Bill Drexel
In a lot of ways the technological repression ecosystem that's coming out of China, we laid the foundation for unwittingly with the best of intentions. Ultimately when you look at things like Microsoft's AI research lab in Beijing and the alumni of that program, it reads as a who's who of the big AI tech companies in China today that are fueling the repressions in Xinjiang and elsewhere.
Zooming out a little bit we had this idea that by opening up Chinese economies and engaging with free trade with China we might pull them in the direction of a freer, more open society. In recent years we've seen that conclusion kind of turned on its head. You could argue that trend is most dramatic in tech where it's not that we've pulled them into a freer, more open internet but quite the reverse. They've pulled the internet and these technologies further into autocratic repressive applications.
Federal Newswire
Is China likely to harm US-China relations or ordinary American citizens through something like TikTok? Is China flirting with a potential AI catastrophe?
Bill Drexel
I think both are likely, I’m maybe more worried selfishly about the latter, but I would say a catastrophe is any kind of discrete event that causes large loss of human life or economic potential or so on. An AI catastrophe is any kind of use of artificial intelligence that suddenly leads to a major issue that has horrible consequences.
A few possibilities that already have proof of concept in 2010. There was what's known as the flash crash where there were algorithms that were playing off each other in the stock market that wiped out more than a trillion dollars of value nearly instantaneously. That is a mini-catastrophe. It ended up being kind of fixable.
Maybe the most chilling is, there was an American pharmaceutical company that had an AI powered system to minimize for toxicity as they're developing drugs and [they] flipped the purpose of the machine just as a demo to see what would happen. Instead of minimizing toxicity, they set the system to maximize, and what they found is within six hours this system designed 40,000 potential biochemical weapons.
Some of those may not work, some of them may not actually really be able to materialize, but some of them are known chemical weapons. Some of them are in this in-between area where they really might just be novel biochemical weapons that we didn't know about before but that were discovered by the system with really shocking ease and in an incredibly short amount of time.
You can see from those sorts of illustrations how this really powerful technology can suddenly and easily, and sometimes accidentally, go really off the rails with devastating consequences.
Federal Newswire
What would an AI catastrophe entail?
Bill Drexel
The difficulty with it is that it applies to quite a lot of domains in quite different ways. As a general use technology you can build AI into systems to really get huge gains in efficiency in complexity and sophistication but also new vulnerabilities. I guess if you want a working definition of AI, it's a set of tools to allow machine systems to learn and often to learn things beyond what humans are capable of.
You can almost say “what does a catastrophe look like in any particular domain” and then you apply artificial intelligence to it and it can often make that domain advantageous, and the risks a bit more acute just because it can turbocharge that domain. In the same way we might worry about critical infrastructure hacks taking out our water, electricity, or critical systems, you could imagine an AI turbocharged cyber security hack towards critical systems doing that in a more devastating or more dynamic way.
You turn to nuclear reactors and obviously it's not the AI that's exploding but in the same way we've used computer viruses to derail nuclear reactors in Iran, you could imagine that a next generation of computer viruses integrates AI and could lead to similar effects.
It's not a substance. It's not one system. It's not touching one particular sort of catastrophe. But it's an accelerant. It's an accelerant in a lot of domains and…an accelerant in a lot of possible disasters.
Federal Newswire
What policies should we put in place to help with these potential problems?
Bill Drexel
Particularly in AI, the United States government has restricted the export of the most advanced AI specific chips to China, which really helps to slow down their advancement in these technologies. We read the writing on the wall and we know if they have unfettered access to these cutting edge capabilities the results are going to be nasty.
I applaud this Administration's efforts to curb that risk, and in general the move away from equipping China with advanced technologies, I think, is a positive move.
It's a very delicate art. Our economies are still very interwoven. I think at the end of the day our adversary is not the Chinese people, it's the Chinese Communist Party. We also need to balance this with trying to find ways to empower the people and also to technologically empower them to get around the great firewall to avoid these sorts of new autocratic technologies that the state is building out. But I think it's going to be painful one way or another both for our bottom lines and for a lot of the economic ecosystem that exists in China today.
Federal Newswire
Taking a broader perspective, what are our neighbors, friends, and allies in the Indo-Pacific region saying on these topics?
Bill Drexel
It's different. I think in aggregate it's still headed in the right direction maybe just not as quickly as I would like. We see that public perceptions of China have been plummeting in recent years among Europeans and other Asian nations. The governments have not always caught up to that change in perception and you can see places, especially Germany for example, still growing in their economic dependency on China.
In light of Ukraine you would think that there might be this awakening of conscience about the dangers of autocratic regimes, and it doesn't seem to be materializing as much as we might like. But I think at the end of the day, China's posturing on the global stage is doing quite well at alienating the rest of the world. I'd say I think there's still a long way to go. One of the major problems is, the way we are taught history really underemphasizes how big a problem the CCP has been continually. And when people think of the early 20th century they think about Hitler and Stalin–evil paragons, which is fair–but they usually leave out Mao and he should absolutely be part of that triad.
Similarly, if you go to Tiananmen Square, we think of the 1989 massacre of students calling for democracy. Any American that goes [to China] may be shocked that you don't find a lot of that memory among Chinese around you. But what you do find are long lines of people with flowers, sometimes crying, waiting to leave flowers at the mausoleum of Mao Zedong. Who steered China into the greatest famine in human history. So there's a long way to go in terms of correcting the perceptions of China in the US even, but certainly among our allies and friends.
Federal Newswire
What are your recommendations for leaders at the federal, state, corporate, and academic levels to handle the threat of artificial intelligence in China, but especially Xi Jinping?
Bill Drexel
I think we're in the middle of a society-wide discussion on where artificial intelligence and these emerging technologies will go, what their risks are, and how we should approach governing and building them out. That's a conversation we need to have.
If I have one message for everyone in that conversation it's to recognize that we have underestimated the Chinese Communist Party in tech and geopolitics for too long and to our peril.
The same is true with AI, so as we're trying to envision a world in which these technologies are safe, dependable, and really promote human flourishing, yes, we absolutely need to get that right in our own societies. But we also need to recognize that what may be the largest threat to that future is the Chinese Communist Party, and figuring out how to get that right should really be at the top of the agenda.
Federal Newswire
Where can the audience go to follow your thoughts?
Bill Drexel
You can find me on Twitter, @bill_drexel. You can also find my scholar page at the Center for a New American Security, that's www.cnas.org. I look forward to engaging with anyone and hearing everyone's thoughts.
