Jacob Feldgoise and Hanna Dohmen are data research analysts at the Georgetown Center for Security and Emerging Technology.

Federal Newswire

What does the Georgetown Center for Security and Emerging Technology (CSET) do?

Hanna Dohmen

We’re a research institution affiliated with the Georgetown University School of Foreign Service. We are…tasked with providing decision makers and policymakers with nonpartisan data-driven analysis of the security and national security implications of emerging technologies. 

We really take pride in the fact that our analysis is data-driven, and I'll let Jacob talk a little bit more about the data-driven aspect since he is part of the data team here at CSET. 

We have over 50 dedicated staff to examine emerging technology, national security issues, and we also have a dedicated data science team and translation part of CSET as well. We're both part of the “compete” line of research, which deals with US national competitiveness and US-China tech competition, both on the “protect” side. [These are] things like export controls and investment screening, thinking about US government grants, industrial policy like the Chips Act, and those kinds of things. 

We have a team dedicated to biotechnology, cyber, and AI. One team is focused more specifically on different regions and competing in different regions of the world.  The AI assessment team is focused more specifically on AI standards and testing and safety, and the applications team is more focused on the kind of applications of emerging technology in a military context.

Federal Newswire

What makes something an emerging technology?

Hanna Dohmen

…To some extent, [that’s] an open question for a lot of parts of the US government. I don't think there's really an agreed upon definition. 

When we think about emerging technologies, we think about technologies that are in development still and are still largely unrealized. Of course, when we talk about emerging technologies there's also an aspect, take semiconductors for example; to some extent, semiconductors are a foundational technology. We have commodity chips and chips that we use in everyday applications. 

But then we think about the development of chips and the direction that AI chips are moving in, and the industry as a whole. To some extent, it's still an emerging area, but it's still an open question and it changes over time, of course. AI especially is such a fast-moving space right now. Even within AI, there are so many different areas that are developing quickly and are still emerging. 

Federal Newswire

What data drives parts of this analysis?

Jacob Feldgoise

I always find data to be the most exciting direction for policy analysis, because it helps provide us with new information to inform the critical decisions of policymakers. 

That's something I think that the founders of CSET recognized, that there was an opportunity here to put together an organization that tries to advise on critical and emerging technologies but using new data. To do that, we spend a lot of time buying and combining different datasets together to form more complete pictures of how emerging technologies are developing. 

For example, one of the more interesting datasets that we've relied on for years and continuously improved is something called the map of science. It's a massive data set of research literature, and you can do all kinds of interesting things with it. 

You can look and see who's publishing in the technology areas that we think are important, where are those people located, and what research institutions are producing the most innovative research. 

All of these details make a lot of difference and help policymakers get a more complete understanding of how emerging technologies are developing, [and] what steps they should take to ensure that the US remains competitive in the long term. 

To enable all this work, we have a data team with a number of full-time data scientists and a machine learning engineer. In some cases we're applying emerging technologies to solve policy problems about emerging technologies. [It] gets a little bit meta but it's very exciting work. 

I think one of the things that we consistently hear is that this form of analysis is compelling to policymakers. They want to see this data, because it helps them feel more confident in the decisions that they're making, when they can substantiate it with rigorous data analysis. 

Federal Newswire

What is the nature of U.S. technology competition with China?

Hanna Dohmen

The trade conflict and the pandemic certainly played a big role in shifting the relationship. The Trump administration's policy towards China, continued through the Biden administration, has turned out to be quite a bipartisan issue. 

There were also some other changes during the Trump administration that we haven't mentioned.  On the export control side, [we placed] one of China's biggest chipmakers–the Semiconductor Manufacturing International Corporation–on a trade blacklist. Also the focus on 5G and Huawei. 

Then of course during the Biden administration, one of the biggest turning points or indicators of this change was the speech by Jake Sullivan in 2022. [He said] that we're going to approach competition with China… shift [through] export controls. 

Also the focus on restricting US outbound investment into Chinese companies or startups. 

Jacob Feldgoise

A lot of the work that the past two administrations has been to get allies on board. Bringing the rest of the allies along with us has been a struggle, but also a moderately successful effort to date.

Federal Newswire

Has technology changed or is our concern about China due to geopolitical clashes?

Hanna Dohmen

It is a little bit of both, and we've seen significant developments in emerging technologies like AI over the last couple of years. Of course, ChatGPT was just released a year ago and thinking about how much that will change education, healthcare, and other parts of society. 

At the same time, we know that China has been developing those same technologies. Now it just adds more pressure to think about how we want to compete with China in those technologies as they are developing so quickly. 

The tough times that we went through during the pandemic highlighted a lot of geopolitical considerations about supply chains, particularly in the semiconductor industry. When we think about where chips are manufactured, especially advanced chips, which are necessary to power those emerging technologies, we are almost entirely dependent on one company that is based in Taiwan, TSMC. 

That's added a lot of pressure and also incentive for governments in the US, the EU, and elsewhere to bring some of that manufacturing to the states and to increase investment in R&D to advance developments in semiconductors. 

Jacob Feldgoise

Policymakers, particularly in Congress, have been willing to legislate additional funding in some cases and also additional authorities to deal with some of these risks. 

Federal Newswire

What’s changed in terms of how we approach export controls?

Hanna Dohmen

There is something just inherently about export controls that makes it kind of a difficult tool. 

On one hand, the Commerce Department is tasked with the economic success and growth of the United States. But the Bureau of Industry and Security [BIS] is tasked with restricting exports of key technologies. So you're dealing with balancing the national security considerations of exports with the economic considerations for US companies in the US economy, and that makes it quite difficult. 

The other thing that I want to mention is that export controls are really a delaying tactic, and I think that is often forgotten. We cannot expect that export controls will entirely be able to prevent advanced chips getting into the hands of the companies or countries that we are trying to restrict exports to. 

We've seen October 7th controls from 2022 and now the October 17th controls in 2023. The objective of these controls in particular, as articulated by the Commerce Department, is to avoid or restrict exports of advanced chips and semiconductor manufacturing equipment that is necessary to manufacture advanced chips. All of that is supposed to help slow China's military modernization, and slow China's ability to train frontier AI models that have the potential to develop weapons of mass destruction or advance conventional weapons.

Jacob Feldgoise

It's important to think about where we started this. This toolset was not designed for the types of problems that it's now been applied to. 

What I mean by that is that export controls since the 1990’s have traditionally been focused on nonproliferation objectives: preventing the export of items that are particularly needed for weapons of mass destruction and nuclear weapons. Now they're being applied to a whole new set of problems that are focused on dual-use goods that can either be used for military or commercial applications. In many cases, the goods that are now being added to these control lists are not modified for military use. 

That the same graphics processing unit can be used to train a machine learning algorithm for the Chinese military or for a completely commercial company, that presents a really tricky challenge, and it also has necessitated a variety of reforms and updates to the way that export controls work. 

We've seen creative expansions of the regulations over the past 10 years or so, including the application of a foreign direct product rule to Huawei and 5G. These controls that were unveiled a year ago [were] new controls on the activities of US persons, which created extraterritorial controls in a whole new way. 

An under-discussed element of the latest update has to do with this whack-a-mole approach BIS is trying, something new which involves targeting companies based on their location, headquarters, and also their ownership structure. 

That's a system that has traditionally been reserved for sanctions, but it's now being applied to export controls in a new way. I think BIS and the US government are adapting, and they're beginning to recognize that this whack-a-mole approach has a number of serious limitations, and they're trying something new. 

For one, I'm looking forward to watching this and seeing how this new mechanism changes and grows in the coming years and how it's interpreted by the lawyers at companies which are ultimately on the front lines of this work.

Federal Newswire

How do we make and enforce standards while still staying competitive?

Jacob Feldgoise

There are different types of standards, and there are ethical standards that are right. 

I'd say ethical standards, those are a little bit less my expertise. I think they're incredibly important. There's obviously been a lot of international movement over the past couple months, or even really just the past month, trying to set ethical standards on emerging technologies, particularly AI. 

We saw this large summit in the UK called the Bletchley Declaration. The US government is trying to figure all of this stuff out as well. I think there are a lot of interesting experiments that are in the works in different jurisdictions in the US. 

The EU has its own AI act which ingrains some of these ethical concerns. China is also attempting its own approach. So I think that this is such a nascent space, people are trying to figure out how to define terms still. Until you figure that out, it becomes hard to do anything else. 

There have been a number of concerns about technical standards setting. Basically the technical parameters for how technologies should work and what this looks like in practice is countries getting together at an international level or rather in many cases companies getting together at an international level and saying, “Hey, we all think it'll be better for interoperability [if] our devices communicate with each other [and] if we all have the same definition of what is 5G. Otherwise if you have your 5G phone and I have my 5G phone but we're in different countries, they're not going to be able to talk to each other.” Those kinds of discussions.

The ethical element is still there, but it's a little bit more muted. I think in those conversations there actually is a little bit more room to bring in Chinese companies and the Chinese government…into the conversations. It's certainly something that the US government should monitor to make sure there isn't foul play. 

To the extent that we're satisfied with how Chinese organizations are interacting in those domains, then I really do think that's an opportunity for collaboration. I think we need to think about how we can continue to engage with China in productive ways and not just treat this entirely as zero sum competition.

Federal Newswire

What are the challenges once these standards are set?

Hanna Dohmen

China over the last two-ish years tried to take a proactive approach to regulating AI. First we saw regulation on algorithms, then deepfakes, and then just generally more synthetically generated content, as well [as] including large language models or content created through large language models. 

Of course, the approach that China takes to regulating these technologies is in part driven by a desire to control content, which is different than the intention of controlling AI here in the United States or allied countries. 

For that reason, it is really important that the United States works with allies that also have a human rights-oriented approach to AI regulation, and not to allow China to take advantage of an early-mover advantage in standard setting.

There is also some room for learning from what China is doing. Not all of it is intended to censor content and whatnot. 

Federal Newswire

Where can we go to follow the Center and your work?

Hanna Dohmen

You can follow CSET on Twitter and Linkedin as well as threads @csetgeorgetown. You can also subscribe to CSET’s newsletters. My Linkedin is @HannahDohmen as well as my Twitter. 

Jacob Feldgoise

I'm on Twitter. You can also find my profile on the CSET website, which is www.cset.georgetown.edu. All of our other researchers' publications are presented.

Hanna Dohmen

Yes, we also just published a report called “spring science,” which looks at US government grant activity in AI, and we find that just at a high level, companies play a much bigger role in US government grants in the AI space than companies do in other areas. The Department of Defense appears to prioritize that funding in particular. But check out the CSET website for that publication.