$6k Contest + AI Exec Order Bites: Immigration, Diffusion, and Cyber Threats
Essay Competition to Predict China's Future!
For today’s Friday Bites, I brought together three of my favorite policy analysts to share their key takeaways on the AI EO’s implications for immigration, diffusion, and cyber.
But before that — we’re excited to announce a new ChinaTalk Essay Competition!
We’ve teamed up with Manifold Markets to give away $6,000 in prizes — as well as an interview on the podcast — to essays making a bold prediction on the future of China.
Here’s how it works:
Starting with a prediction on a China-related theme, write an essay explaining your point of view and supporting evidence. Check out our website for details on submitting your essay, which is due by January 15, 2024.
I’ll will read every essay we receive, and then highlight the best ten in a series of posts on our site, as well as on Manifund.
On February 15, we’ll announce the winner — first prize will take home $3,500 and two runners-up will receive $1,000 each (pending laws in your home country).
The first five credible essays we receive will each receive $100.
The grand-prize winner will also be offered the opportunity to be interviewed on the ChinaTalk podcast to discuss their prediction!
Matthew Mittelsteadt—Biden Just Wants to Diffuse!
Monday’s executive order on the safe, secure, and trustworthy development and use of artificial intelligence contains a lot — from new AI reporting requirements to modest immigration reforms and even action to safeguard biotechnology. Dozens of pages long, naturally it’s a mixed bag with many highs and several lows. All things considered, though, what has me most excited about this order is its clear emphasis on federal government AI diffusion.
In a September essay in Noema Magazine, Jordan and I argued, “Facilitating AI’s diffusion will be key to national competitiveness for the coming decades,” and that “American policymakers should do their utmost to ensure that not just leading-edge labs, but also firms, schools and government bureaucracies themselves are able to make the most out of AI.”
How To Win The AI Race
The best way to create AI abundance and compete with China isn’t policy focused on R&D (though investment is certainly encouraged) or protectionist controls. Rather, we must pursue what we call a diffusion-centric AI strategy: focusing on rolling out this tech, building talent, and easing barriers to adoption.
If implemented well, the recent AI executive order has the potential to form the cornerstone of such an approach.
Originally, I expected the Biden administration’s attention to focus on taming AI, putting the brakes on its overapplication. While these elements certainly exist, the administration clearly sees AI as lightning they want to capture in a bottle. Throughout the order, we find countless provisions devoted to “AI capacity.” The administration is embarking on a whole-of-government push to educate, bring in talent, and assess IT with the aim of creating a modern, AI-driven government that (hopefully) is more efficient, approachable, and fair.
On AI.gov we already see the tip of this iceberg: the administration has already begun asking citizens to join what they call “the National AI talent surge,” an effort backed by new postings, prioritized hiring, and eased clearance requirements for both citizens and (uniquely!) noncitizens. All these efforts are encouraging and exactly what is needed.
Even so, the National Institute of Standards and Technology (NIST) will be implementing many provisions of the order — and it currently has only twenty employees staffed on its Responsible AI team.
From a recent Washington Post piece:
The need for capacity-building in key agencies like NIST is plainly needed to meet not only the weighty demands of this order, but also the uniquely technical demands of this AI moment.
While the administration’s ambitions are high, both implementation design and competing priorities may hinder this big AI capacity push. Countless provisions in the order are devoted to capacity-building, but countless more are devoted to standards, regulations, and ethical limits. Wednesday’s release of the OMB implementation guidance offers an early indication of the raft of requirements these systems must meet. No doubt more rules will only follow. Limits and standards are certainly needed, but so too is caution in installing stoplights. If procurement and use of AI is overregulated, the red tape could slow down and impede this rollout — just as it did with the federal government’s digitization efforts.
If we really want AI capacity, developers and staff will need the freedom of action to explore, innovate, create, and implement — which today’s excessive rules prevent. Thankfully, some early signs suggest an understanding of this balance. The administration explicitly states that “agencies are discouraged from imposing broad general bans or blocks on agency use of generative AI” and to provide employees generative AI services (with limits of course) for “purposes of experimentation.” There is little Luddism to be found here. While they have reasonable reservations, the administration clearly wants to make use of this tech. Only in the coming months will we know how these worthy goals will fair under the rules binding bureaucratic implementation.
Divyansh Kaushik—How Much Progress Did We Get on STEM Immigration?
In its ongoing quest for global talent, the United States, under President Biden’s directive, plans to revise its immigration policies to better cater to international scholars and workers. The potential beneficiaries of these new visa policies, spelled out in the recently released executive order, are as wide-ranging as they are profound.
The EO proposes changes in seven key areas:
Domestic revalidation for J-1 and F-1 visas;
Modernization of H-1B visa rules;
Updates to the J-1 Exchange Visitor Skills List;
Introducting a global AI talent-attraction program;
Issuing a request for information (RFI) to seek updates to Department of Labor’s (DoL) Schedule A;
Policy-manual updates for O-1A, EB-1, EB-2 and the International Entrepreneur Rule.
And using discretionary authorities to attract and support AI talent.
Together, they pave a new road for the foreign-born talent which forms the lifeline of America’s innovation-driven economy.
The new revalidation steps for the F-1 (student) and J-1 (exchange visitor) visas are a welcome simplification, reducing costs and allowing hundreds of thousands of students — especially those in STEM — to focus their time and energy on their research instead of visa-related procedural nightmares.
Further modernization of H-1B (specialty occupation) visa rules could allow easier transition into new jobs or ventures and discourage multiple applications per applicant, thereby preventing exhaustion of yearly quotas.
The J-1 Skills List refers to a list of fields of specialized knowledge deemed necessary for the development of the J-1 visa holder’s home country; after the completion of their J-1 visa program, these visa holders are required to return to their home country for at least two years. The EO’s directed revisions to the J-1 Skills List promise to bridge the gap between the current needs of the AI industry in the US and the skillset of J-1 professionals — a majority of whom are from China and India. This shift better reflects the realities of the global labor market, directing skilled labor where it is most needed.
Section 5.1(c)(ii) of the EO directs the State Department to establish “a program to identify and attract top talent in AI and other critical and emerging technologies at universities, research institutions, and the private sector overseas.” The directive is a stark reminder that talent — particularly AI talent — is indeed global.
Schedule A occupations refer to those deemed by the DOL where there are not “sufficient US workers who are able, willing, qualified, and available.” Schedule A Group I was created in 1965 and has remained unchanged since 1991 — so the DOL’s RFI for a Schedule A update aims to better match the labor shortages in specific industries or regions to foreign workers who are qualified to fill these gaps.
[Jordan: Right now it’s just for nurses and physical therapists. If you can’t do bedside care but just happen to be an immigrant of “exceptional ability in the sciences or arts” who can clear the bar below, you get stuck in a years-long line for the DOL to “test the labor market and certify there are no qualified U.S. workers available for the job.”]
An update based on this RFI would reduce an otherwise lengthy Green Card process for immigrants with specified skills. Foreign workers with occupations that are on Schedule A do not have to go through the PERM (Program Electronic Review Management) labor certification process, which otherwise takes, on average, 300 days to complete. An updated Schedule A could cut PERM applications filed significantly down from current high volumes — over 86,000 already filed by the end of FY23 Q3. While the EO calls only for an RFI seeking information on the Schedule A List, this is a critical first step to an eventual update that is badly needed.
The updates to visas like O-1A (visa for those with “an extraordinary ability in the sciences, education, business, or athletics”), EB-1 (employment-based first preference), EB-2 (employment-based second preference), and the International Entrepreneur Rule (which allows immigrants to work on startups in the US) offer increased clarity on the criteria for “exceptional” and “extraordinary” ability. These clarifications demonstrate the US government’s recognition of emerging technologies’ importance on a global stage.
Finally, the possibility of using discretionary authorities to extend visa-interview waivers, recapture Green Cards lost during the COVID-19 pandemic, and use public-benefit parole to attract researchers stuck in lengthy visa procedures (and those fleeing authoritarian regimes) may significantly increase the number of AI experts headed to the US.
Overall, these changes promise to enhance the flow of foreign talent to the US, supporting areas of critical national interest like education, technology, and healthcare. The battle for talent is real, and America shouldn’t shy away from harnessing global talent. And to do so, it must embrace a transformative immigration structure — one that is forward-thinking, accommodating, and constantly on the pulse of the rapidly evolving global skills landscape.
[Jordan: I’m nervous that the point of highest leverage came before the EO dropped and that all the “recommendations to consider” language will be slow-rolled by a bureaucracy that isn’t resourced (or energized?) to make these process changes. And, if we don’t get action in the next few months, as we get closer to the election you’ll also probably see Dems worried about headlines of Biden letting in more immigrants in the leadup to the general. But, one can hope!]
Dakota Cary—How to Respond to China’s Cyber AI Threat
The Biden administration’s move to bolster AI and cyber research will make the technology a focal point for the research community and instigate its use by federal agencies. Specifically, the EO requires the Department of Defense (DOD) and the Department of Homeland Security (DHS) to provide the president a report of LLM’s ability to find, diagnose, and fix vulnerabilities within their respective systems. DOD already uses some non-LLM systems to perform these tasks, so this requirement will push agencies and industries toward greater defensive automation. Elsewhere, the Biden administration highlights the new AI Cyber Challenge (AIxCC) from DARPA, which puts money into teams competing to automate key parts of cyber defense.
A few think tanks — though mostly the Center for Security and Emerging Technology (CSET) — have published reports in the last three years on the potential impact of AI on cybersecurity and cyberattacks. While these reports have informed policymakers, they have not caused the cybersecurity research community to pivot toward the development of the technology. DARPA has continually invested in the application of machine learning frameworks to cybersecurity topics, as has the National Science Foundation.
For its part, China has been investing in applying AI to cybersecurity since DARPA wrapped up its Cyber Grand Challenge in 2016. Inspired by the US’s competition — which did not actually use any AI or machine learning frameworks — China has hosted its own competitions since then. The Ministry of State Security and Ministry of Education hosted the first three, beginning in 2017. Since late 2018, however, the PLA has become the main host of China’s “Robot Hacking Games,” as the competitions are called. Without access to entrants’ technology, it’s hard to say how many are using AI or machine learning in their automated hacking systems — but at least one university tied to the PLA is working on the technology.
Although the research community has speculated — sometimes wildly — about AI’s impact on both offensive and defensive hacking, it is clear that such systems can improve our ability to discover software vulnerabilities, and then to patch or exploit those vulnerabilities. The recent AI EO by the Biden administration — in addition to moves by DARPA and other agencies — recognizes this fact. Because the technology is so nascent, and classified, it’s hard to determine who has an advantage. Still, it is worth appreciating that both countries have prioritized the technology.