IGCC Postdoc Emily Kim Talks AI and Tech Governance

What first got you interested in the fellowship?

I did public policy for my PhD, but my background is in international security. My master’s dissertation was about nuclear nonproliferation, and my research back then first connected me to the work of IGCC.

After my master’s, my path had an abrupt change of direction. I’m originally from South Korea and began a post-college career working at a local science and technology company.  There, I realized that I was most interested in science and technology itself, rather than international security per se.

It just so happened that IGCC’s postdoctoral fellowship program allows me to reconnect my technology studies back to that international background.

And now you study the hottest tech topic of them all, AI, right?

That’s correct. I’m a social scientist, but I’ve always been interested in how advances in technology can impact society. Recent technological innovations promise to have a profound impact on all of society, from the high politics of national defense down to our everyday lives and how we interact with others.

I realized that AI was really going to be the critical technology of the 21st century back when I started my PhD in 2020. Even just five years ago, at a tech-focused school like Georgia Tech, not many people were interested in AI. Everyone knew about it, but nobody expected it to become so central to absolutely everything in such a short period of time.

What motivates you in your studies? Is there a big problem you’re trying to solve?

I want to contribute to resolving the pacing problem of AI. Technology develops very fast, but the policy and governance surrounding it develops much more slowly—that’s been the case for just about every new technology. But given the stakes involved with AI, we have to move a lot faster.

Everyone and their mother is talking about AI these days, and it’s mostly pretty dire. Is AI the harbinger of doom everyone thinks it is?

It’s more complicated than that. I was initially very interested in AI’s social impact in terms of automation and unemployment. It’s a hot topic, but that conversation can be overly pessimistic, because the truth is, we’re still figuring out how that impact will be felt.

When I began my PhD, most academic papers on this topic said that manual, blue-collar jobs would be affected the most by AI. But it only took two years for all of academia to reverse themselves on those conclusions. Now, with the rise of large language models and artificial general intelligence, we—the very researchers who first came out with those studies—are actually the most threatened by AI. It’s a lot easier to replace your average white-collar “knowledge” worker with software, versus a blue-collar manual worker where you need both software and hardware.

In other words, a lot can change in our understanding of AI in a very short period of time. Who knows what else we’ll be wrong about five years from now?

Five years is along time. But two years isn’t. What do you plan to do during the two years of the IGCC fellowship?

Write papers. When I started on my PhD, not many professors had done anything on artificial intelligence. The field was far too new, so I was integrated into the cyber faculty. I published on cyber while, on the side, developing AI-related research for my dissertation. I want to build on that work, refining and publishing each chapter of the dissertation, and work towards putting out a book on the topic.

Is it useful to be here in Washington, D.C.?

Of course. Compared to Georgia, there are a lot more chances to communicate with researchers in think tanks who share the same interests. There are also more opportunities to participate on the policy side. The physical proximity to policymaking communities provides so many opportunities to harness this research to make a real impact.

I assume the policy community’s understanding of AI has some limitations.

Yes. Not only does AI suffer from the fact that people who research it from a hard science and social science perspective aren’t speaking with each other, then you have the policymaking community on another island, and that creates real problems when the technology is essentially being propelled by an industry-led black box innovation process.

What do you mean by a “black box?”

[Laughs]. I call AI a black box because—unlike engineers in practically every other technology—AI developers can’t trace their steps back if something goes wrong. They can’t go back from point B to point A because the technology is just way too complex. AI algorithms evolve in this opaque manner influenced by machine learning, which means that even the engineers who work on them don’t fully understand how the actual algorithm works. And because of that, the AI companies don’t really understand how their final product will affect their consumer.

And if the engineers don’t understand, that means policymakers certainly can’t anticipate how society is going to be affected by this technology. And it’s impossible to regulate when you can’t anticipate what the system is going to look like.

The bottom line is that there is an enormous amount of uncertainty for society overall. But introducing regulators to these concepts like machine learning can get the process off to a good start. That’s why I think it’s so important to have interdisciplinary collaboration in this field.

Your studies took you from Seoul to Atlanta. How do the two cities compare?

Seoul is a very big, heavily populated place—there’s so much packed together in a relatively small amount of space. When I lived in the city, I could go to watch a movie, shop at a department store, or hop on public transportation and go elsewhere, all within just a 10–15-minute walk.

When I first got to Atlanta, I thought it’d be pretty much the same. Then I landed in Atlanta for the first time. On my taxi ride from the airport, I expected to see a lot of tall buildings and packed streets, only to see trees and what looked a lot like the countryside of South Korea. I thought to myself—it’s the eighth largest city in the U.S., but where is the city part?

Although my first impression wasn’t great, living there gave me the chance to meet and connect with many people. By the time I had to leave, I felt truly reluctant to go.

You’re moving to D.C. at the right time—September, where the weather is nicest. You won’t have to deal immediately with the rough summers we have.

I’ve gotten used to East Coast summers. They’re pretty brutal in Georgia [laughs].

What do you typically do for fun, when you’re not knee deep in research? Please don’t say reading journal articles.

[Laughs]. I’ve met a lot of academics who insist that reading academic journals in their free time is their idea of fun. That doesn’t really sound like my idea of fun.

I’m a huge fan of film, and love going to the movies. I also really enjoy baseball and sports more generally—more watching than playing.

I also do CrossFit—I’m not sure if you’ve heard of it.

I have. CrossFit is huge in D.C.—how about in Seoul?

CrossFit wasn’t very popular in Korea at first. But slowly, more and more gyms started to appear, including a very good gym right by my house. So I gave it a try and got hooked.

I’m very much an academic in that I always want to optimize, and CrossFit is one of the most efficient workouts you can get. You work up a sweat very quickly, and the program can be very easily adjusted. You can get a very well-rounded workout in just 20–25 minutes. I just hope I can find a gym in D.C.

I doubt that will be a problem.