Claude 3, OpenAI, and Gemini headline a packed month for generative AI, the European Parliament adopts the EU AI Act, and the U.S. House’s AI Task Force

Sign up
Worth Knowing

Claude 3, OpenAI’s Sora, Google Gemini, and Mistral — A Roundup: It’s been a month filled with maneuvers in the world of generative AI, with developers debuting new capabilities, leveling the playing field, teaming up with the competition, and apologizing for mistakes:
European Parliament Passes AI Act: Yesterday, the European Parliament gave final approval to the EU AI Act, the bloc’s comprehensive AI regulation. Initially proposed in 2021, the law survived many rounds of negotiation and a handful of scares on its way to the finish line. When it begins coming into force later this year, the AI Act will be among the world’s most significant AI-focused regulations, with its standards likely to have a significant impact on AI development abroad. The new regulation imposes bans on specific AI applications that compromise citizens’ rights (such as biometric categorization, untargeted facial recognition, and social scoring), sets stringent transparency requirements for general-purpose AI systems, establishes strict conditions for law enforcement use of real-time biometric identification, sets obligations for high-risk AI systems to ensure safety, transparency, and accountability, and mandates clear labeling of deepfakes. The law is expected to enter force in May or June (after some more rubber-stamping), with different provisions taking effect in stages. The AI Office tasked with guiding its implementation was established earlier this year and has begun hiring (though low salaries may make that difficult).
Scientists and Researchers Call Attention to AI Risks: Two separate letters signed by prominent researchers pushed for steps that could help mitigate the potential risks associated with AI development:
  • More than 300 AI researchers called on AI companies to enact policy changes that would allow for more independent research on their systems. In a public letter and an accompanying arXiv paper, the researchers pushed for AI developers to commit to “legal and technical safe harbor” to protect and encourage independent evaluation and red-teaming of their AI systems. While security experts can usually probe traditional software vulnerabilities thanks to established norms and legal safeguards, similar protections are absent for AI research. Researchers interested in exploring the risks and vulnerabilities of many of the most popular generative AI tools — like OpenAI’s ChatGPT and Google’s Gemini — risk running afoul of developers’ policies, resulting in account suspension or even legal action. While many AI developers perform extensive testing in-house, the lack of external oversight raises concerns about the impartiality of these evaluations. The letter signatories argue that greater protection for independent research is a key step to addressing the risks associated with generative AI.
  • In a March 8th letter, more than 100 scientists, including Nobel laureate Frances Arnold, committed to “a new proactive risk management approach” to protein design in light of evolving AI capabilities. Current AI tools have already shown tremendous promise for protein structure prediction and design. While these rapidly advancing capabilities could be a boon for drug discovery, observers have raised concerns that the same tools could be used by nefarious actors to design and synthesize bioweapons. The March 8 letter echoes some of these concerns and its signatories pledge to balance the benefits of scientific openness with “efforts to identify and mitigate meaningful risks,” among other commitments.

  • More: Can Chatbots Help You Build a Bioweapon?
Japan Opens TSMC Fab In 20 Months as U.S. Projects Creep Forward: Late last month, TSMC opened its first chip plant in Japan. The $8 billion plant was stood up barely two years after it was announced — an impressive effort, especially when compared with the company’s delayed projects in Arizona. TSMC’s U.S. and Japanese expansions are not entirely comparable; the Taiwan-based chipmaker’s CHIPS Act-supported, $40 billion Arizona project will make higher-end 3 and 4 nm chips, while its new Kumamoto-based plant will produce older 22 and 28 nm chips. But TSMC appears encouraged by the initial success of the Japanese project. Last month, it announced plans to build a second Kumamoto facility that will produce more advanced 6 and 7 nm chips, bringing total investment in the project to more than $20 billion. As TSMC aims to diversify its footprint beyond Taiwan, Japan could emerge as a favorite target for future expansion efforts. As Bloomberg notes, ample government support (the government plans to commit nearly $5 billion in support of the second TSMC plant), efficient construction, and round-the-clock work schedules have made Japan an attractive destination. While the United States has been willing to allocate funds, significant delays — observers have cited worker training, permitting, and slow-going subsidy negotiations as causes — could make it difficult to beat out Japan as the go-to destination for TSMC’s chip plants abroad.
Government Updates

U.S. House Launches AI Task Force: On February 20, House Speaker Mike Johnson and Democratic Leader Leader Hakeem Jeffries announced a bipartisan task force charged with proposing new standards and congressional action on AI. Composed of 24 members — 12 appointed by each party — the task force will be chaired by two of the House’s more tech-savvy members, Reps. Jay Obernolte and Ted Lieu. The task force will produce a comprehensive report that outlines “guiding principles, forward-looking recommendations and bipartisan policy proposals,” according to a press release. To date, most attention has been paid to the Senate’s AI regulation efforts — Sam Altman’s widely covered congressional testimony and Chuck Schumer’s AI Insight Forums were both Senate affairs — but no comprehensive legislation has emerged from either body. The task force’s report is expected by the end of the year.

President Biden Budget Calls for $3 Billion in AI Funding: President Biden announced his FY2025 budget request on Monday, a $7.3 trillion proposal with substantial funding proposed for AI and emerging technology. The budget request includes:
  • $1.8 billion for the Pentagon’s AI efforts. While the DOD’s request for AI funding remains the same as its FY2024 request, military officials emphasized that this was in response to caps imposed by last year’s Fiscal Responsibility Act and that AI development and funding remain a top priority.
  • $900 million for the Directorate for Technology, Innovation, and Partnerships, the National Science Foundation’s “crosscutting platform” for emerging and critical tech, established last year.
  • $1.4 billion for National Science Foundation STEM education and workforce development programs with an emphasis on diversity, equity, inclusion, and accessibility.
  • $65 million for the Commerce Department to support the new U.S. AI Safety Institute. The AISI, announced last November, is tasked with leading many of the federal government’s AI trust and safety efforts.
  • $30 million for the second year of the pilot program of the National AI Research Resource, which was stood up earlier this year pursuant to the president’s AI executive order.
  • $70 million for federal agencies to hire Chief AI Officers and bolster their deployment and use of AI, plus $32 million to support the AI Talent Surge.
  • $223 million for the Bureau of Industry and Security, the Commerce Department agency leading the federal government’s export control efforts. The request, if enacted, would represent a 17 percent increase over the FY2024 enacted level.
In Translation
CSET’s translations of significant foreign language documents on AI

PRC CAS Budget: Chinese Academy of Sciences 2023 Budget. This document is the 2023 budget for the Chinese Academy of Sciences, a state-run complex of scientific research institutes and think tanks. CAS strives to enhance China’s overall S&T capabilities and to improve the country’s self-sufficiency in key technologies.

PRC Research Ethics Guidelines: Ethics Guidelines for Brain-Computer Interface Research. These guidelines, issued by a Chinese Communist Party commission, offer non-binding principles for ethical conduct in brain-computer interface research. The guidelines categorize BCI technology by its level of invasiveness and distinguish therapeutic BCI designed to help people with neurological disabilities from augmentative BCI that enhances the abilities of healthy people. The guidelines urge “moderation” in the adoption of BCI technology, particularly in the case of augmentative BCI, which the guidelines warn could exacerbate social inequality if adopted without restraint.

If you have a foreign-language document related to security and emerging technologies that you’d like translated into English, CSET may be able to help! Click here for details.

Job Openings

We’re hiring! Please apply or share the role below with candidates in your network:
  • Research Fellow – Workforce: We are currently seeking candidates to support our Workforce line of research as a Research Fellow. In this role, you will be a key player in helping us accomplish three major research tasks: updating CSET’s assessment of the U.S. AI workforce, executing an analysis of the U.S. cybersecurity workforce, and assessing the U.S. educational pipeline for AI, cybersecurity, and STEM talent. As a report to our Senior Fellow who leads this research area, you will play a key role in shaping research priorities, overseeing research strategy and its execution, and leading a talented team of researchers. If you’re passionate about technology policy and possess strong leadership skills, apply now to make a meaningful impact on the future of America’s Workforce. Apply by Monday, March 18th
  • Data Research Analyst: CSET is seeking capable data storytellers, analyzers, and visualizers to be Data Research Analysts. This role serves as a bridge between the data and analysis teams. It draws on knowledge of research methods, subject-area expertise, and analytic skills. We encourage those with experience in data visualization, programming, and/or statistical analysis to apply. Apply by Monday, March 25th
Please bookmark our careers page to stay up to date on all active job postings. You can also subscribe to receive job announcements by updating your subscription preferences in the footer of this email.

What’s New at CSET

ANNUAL REPORT

In honor of CSET’s fifth birthday, our annual report — CSET at Five — is a look at CSET’s successes in 2023 and over the course of the past five years. It explores CSET’s different lines of research and cross-cutting projects, and spotlights some of its most impactful research products.

REPORTS

PUBLICATIONS AND PODCAST APPEARANCES

EMERGING TECHNOLOGY OBSERVATORY

EVENT RECAPS

IN THE NEWS

What We’re Reading

Report: Biennial Report to Congress on International Science & Technology Cooperation, The Subcommittee on International Science and Technology Coordination of the National Science and Technology Council (February 2024) 

Paper: Towards AI Accountability Infrastructure: Gaps and Opportunities in AI Audit Tooling, Victor Ojewale, Ryan Steed, Briana Vecchione, Abeba Birhane, and Inioluwa Deborah Raji (February 2024)

Report: The State of U.S. Science and Engineering 2024, Steven Deitz and Christina Freyman, National Science Board (March 2024)