Fear not the Chinese bots

The ‘Because China’ argument against pausing and reflecting on AI research is the opposite of persuasive.

“The first question people in the US should ask is, if the US slows down, do we believe China will slow down as well? I don’t believe for a moment that the institutions we’re competing with in China will slow down simply because we decided we’d like to move more slowly. This should be looked at much in the way that the competition with Russia was looked at [during the Cold War].”
—a Microsoft executive, explaining to a Vox reporter why the US can’t afford to slow the pace of AI research.    

Because China! Predictably, that has become a standard reply to people who ask why the US can’t take a firm hand in regulating artificial intelligence—and to people who go further and ask why the US can’t impose a moratorium on some forms of AI research while we figure out what regulations are in order.

There are two problems with the “Because China” argument, and together they undermine it pretty thoroughly.

The first was highlighted by Ezra Klein in his New York Times column this week: China is itself taking what is in some ways a slow and cautious approach to the development of AI. Klein reports that AI rules released by China last week “are much more restrictive than anything the United States or Europe is imagining, which makes me very skeptical of arguments that we are in a race with China to develop advanced artificial intelligence.”

China’s stringent approach isn’t surprising. Chinese governments have long put a premium on social stability, and AI has the potential to massively destabilize society. And the current Chinese government is, by recent Chinese standards, especially authoritarian, and so especially averse to turbulence. Beijing is no doubt concerned about the rapid evolution of a technology that could help small groups or even individuals make big waves—foment or mobilize dissent, overwhelm online censors, hack into vital infrastructure, and… well, it’s in the nature of this technology that you can’t anticipate all possible uses of it, a fact that probably hasn’t escaped Xi Jinping’s attention.  

Of course, the same authoritarian insecurities that lead Xi to constrain AI development presumably lead him to intensify it within certain channels, notably government research labs. After all, AI can be a powerful tool of social control—just look at the facial-recognition AI that pervades China’s domestic surveillance network. So the restrictive regulations Klein writes about may be intended not to slow the progress of AI research per se, but rather to keep the research centrally controlled.

Still, in practice, centralized control tends to slow progress. Just look at the other side of the coin—the way decentralized innovation is now powering AI evolution in America. And I don’t just mean that OpenAI’s famous large language model, GPT, benefited from cross fertilization among research labs in big companies, startups, and universities. I also mean that OpenAI is now letting entrepreneurs build products on top of its LLM, and one result is a whole new species known as “agents”—AI tools with names like AutoGPT, AgentGPT, and (perhaps ominously) GodMode.

These agents basically transform ChatGPT from a mere talker into a doer—permitting it to do, for example, complex and impressive market research with minimal prompting and no supervision but also raising the prospect that it could do complex and impressive hacking with minimal prompting and no supervision and thus raising the prospect that it could wreak much havoc.         

Which leads to the second problem with trying to use “Because China” as a killer argument against pausing and reflecting on AI research. Namely: “Because China” is also an argument for pausing and reflecting. After all, what happens if America doesn’t pause and reflect, and as a result America dissolves into chaos while China doesn’t? That’s no way to win a cold war!

And it isn’t just that this is one among many ways America could “lose” to China. When you try to flesh out the fears of Chinese dominance that seem so common these days, this China-stays-intact-while-America-doesn’t scenario starts to look like one of the more plausible ways for America to lose.

After all, what are the alternative ways? Getting invaded by China and conquered? Yes, that would count as a loss, but various things make it an extremely low probability outcome.

Then how about this as a way of losing: China expands its global network of allies until it is powerful enough to choke the life out of the American economy, or out of American liberal democracy? Again: Definitely bad, but not, in my view, very plausible, especially in light of the fact that China has no obvious reason to smother the American economy (which keeps food on the table for a lot of Chinese people) or to smother American liberal democracy (which the Chinese leadership wouldn’t particularly care about if America would quit demanding that China emulate America’s form of governance).

In contrast to these unlikely ways for America to lose to China, I find it wholly plausible that American society could spin out of control while Chinese society remains stable. And one reason that’s plausible is that America is now turning itself into the canary in the AI coal mine, while China stands back a few paces and monitors the canary’s health.

And, actually, the canary is already starting to look not very healthy. As a recent New York Times op-ed pointed out, the engagement-maximizing algorithms that have helped make social media a hotbed of polarization are driven by a primitive form of AI. Now imagine that, on these same social media platforms, you add some partisan applications of AI tools that are now becoming available. Such as: armies of very smart and human-seeming social media bots deployed by people on Team Red or Team Blue to spread misinformation (possibly via AI-created deep fakes) or just to dunk on and otherwise insult members of the other team en masse, amplifying tribal animosity by an order of magnitude or so. 

And don’t forget that America’s political polarization, though intensified by social media, wasn’t created by social media. There are underlying structural causes, reflected in, among other things, the grievances that fueled Donald Trump’s ascent to the White House—such as the dearth of good working class jobs and, more broadly, the growing economic and cultural gap between affluent Americans and ordinary Americans. Now we want to add to that recipe for social conflict the wave of fresh unemployment that breakneck AI progress is bound to create?

The various problems that AI brings may well prove soluble. Historically, technologies that take people’s jobs have often created more jobs in the long run. And the armies of bad bots unleashed by AI may eventually be neutralized by good bots—or by more vigilant and skeptical HI (human intelligence). But it seems pretty obvious that the more slowly all this unfolds, the better the chances that the solutions can start to kick in before the problems reach tectonic proportions.

I don’t personally buy into the idea, so prevalent in the US now, that an epic clash with China is our destiny and much of our public policy should be organized around the goal of coming out on top. But so long as that idea prevails, one effective argument for the cautious and judicious governance of AI will be: Because China. If America plunges into chaos, and China doesn’t, America will have lost.

And, come to think of it, plunging into chaos will be a bad thing for America regardless of what happens to China.      

Note: I recently moderated—and intermittently participated in—a debate about whether AI is a huge threat to humankind. The first hour of this Nonzero podcast is available to everyone, and the full two-hour version is available to paid NZN subscribers.

Image: By Clark McGillis using Stable Diffusion