DeepMind Shares Thousands of Modeled Proteins — and the AI Behind It: Last week, DeepMind, the UK-based AI company owned by Alphabet Inc., released a database of more than 350,000 protein structures predicted by its protein-folding AI, AlphaFold, and announced plans to model more than 100 million structures over the coming months. The news came a week after DeepMind released AlphaFold v2.0’s source code and published a paper in the journal Nature explaining its methods. As we covered last December, AlphaFold made waves when it successfully modeled protein structures with an accuracy equivalent to state-of-the-art methods in a fraction of the time — AlphaFold mapped proteins that took humans months, years, or even decades to model. Observers say DeepMind’s decision to release its work — including predictions for the structures of nearly every protein in the human body — could help accelerate biochemical research and drug development.
- More: DeepMind’s protein-folding AI has solved a 50-year-old grand challenge of biology | Google Unit DeepMind Tried—and Failed—to Win AI Autonomy From Parent | Generally capable agents emerge from open-ended play
- More: These Bendy Plastic Chips Fit in Unusual Places | Delays with European Approval of Nvidia-Arm Deal Could Break It
- Taiwan-based TSMC — the world’s biggest chipmaker — received final approval from Taiwanese regulators to build a new fab capable of producing 2 nanometer semiconductors. Those chips would be significantly more advanced than today’s state-of-the-art 5 nm chips, but with fab construction set to begin next year, it will likely be 2024 or 2025 before the chips go into production. While that plant will be based in Taiwan, TSMC has reportedly been exploring building a new fab in Germany, which would be the company’s first in Europe. With the European auto industry suffering the effects of the chip shortage, European leaders have been attempting to attract semiconductor manufacturers and shore up supply chains, but TSMC representatives cautioned that the discussions on a German plant are still “in very early stages.”
- After falling behind rival fabs TSMC and Samsung, U.S. chipmaker Intel laid out a plan to reclaim its spot as the world’s top chipmaker by 2025. The company announced that it would debut a new transistor architecture called RibbonFET in 2024, which it says will offer major performance improvements over its current FinFET transistors, first introduced in 2011. CEO Pat Gelsinger also said the company would move away from the nanometer-based nomenclature that has been used to distinguish semiconductor generations — its 10 nm Alder Lake chips will be dubbed “Intel 7,” and its 7 nm processors have been renamed “Intel 4,” for example. TSMC and Samsung had enjoyed a growing lead in the “nanometer race,” but observers rightly pointed out that the naming schemes haven’t described actual gate lengths since 1997: Intel’s 10 nm chips were competitive with TSMC’s 7 nm chips, so this change simply harmonizes the nomenclature of the world’s three top chipmakers. And not to be outdone by TSMC, Intel has also been in discussions to build fabs in Europe. Following meetings between Gelsinger and French President Macron and Italian Prime Minister Draghi, Intel expressed interest in building $20 billion in new fabs in Europe, provided it can secure significant subsidies and a suitable site. Reports indicate that Germany, the Netherlands, France and Belgium are the chief candidates for a potential site, but Intel is seeking as much as $10 billion in subsidies before it commits to building the fabs.
NSCAI Summit Features AI Remarks From D.C. Big-Wigs: Earlier this month, the National Security Commission for Artificial Intelligence hosted a summit in Washington, D.C., that featured remarks from a who’s who of U.S. policymakers and foreign representatives. The summit served as a capstone to the commission’s work and publication of its final report. Speakers included Secretaries Austin, Blinken and Raimondo, Senators Schumer, Ernst, Warner and Young, OSTP Director Eric Lander, and National Security Advisor Jake Sullivan, as well as a number of foreign dignitaries, including Executive Vice President of the European Commission Margrethe Vestager and New Zealand Prime Minister Jacinda Ardern. Topics that featured prominently included: strategic competition with China, the importance of partnerships between like-minded democracies, the need to maintain and support democratic values, and the failure of the United States and its allies to lead the way on international standard setting for emerging technologies, such as 5G and AI. During his remarks, Secretary Austin announced plans to direct $1.5 billion over the next five years to the Joint Artificial Intelligence Center (though funding will ultimately be up to Congress — see the story below) and confirmed that the JAIC has been elevated to report to Deputy Secretary Kathleen Hicks, per the NSCAI’s recommendation. Sullivan, meanwhile, painted a grim picture of the future of U.S.-China technological competition absent significant federal investment and international cooperation, and to that end endorsed the U.S. Innovation and Competition Act passed last month by the Senate (for more on that bill, see our recent coverage).
NDAA Draft Passes Senate Committee — House Version in the Works: Last week, the Senate Armed Services Committee approved its version of the National Defense Authorization Act for fiscal year 2022. While the bill’s text has not yet been released, a SASC summary indicates it includes several provisions related to AI and emerging technologies. According to the summary, the NDAA would implement “a number of recommendations from the National Security Commission on Artificial Intelligence, including accelerating processes to apply artificial intelligence capabilities to military systems, processes, and operations.” (For more on the NSCAI’s March report, see our coverage here.) The summary also includes plans to establish a commission on “Planning, Programming, Budgeting, and Execution” reform, which observers say could lead to an overhaul of the acquisitions process to make it better-suited for emerging technologies. Among the bill’s other provisions are a number of research funding increases and a plan to establish a national microelectronics research network, as mandated by the CHIPS for America Act included in last year’s NDAA. The House Armed Services Committee is considering its version of the NDAA this week, but it will not be finalized until after the full committee markup in September.
Executive Branch Appointment News — Nominations & Withdrawals: President Biden announced his intent to nominate Laurie E. Locasio to be director of the National Institute of Standards and Technology, while Michael Brown withdrew his nomination to become the Defense Department’s undersecretary for acquisition. Locasio worked at NIST for more than three decades before joining the University of Maryland system in 2017. If confirmed, Locasio would oversee NIST during a significant expansion — the Biden administration proposed a 45 percent funding increase for the agency as part of its 2022 budget request. While Locasio’s appointment process has thus far been relatively drama-free, Michael Brown’s was anything but. His request to have his nomination withdrawn following a re-surfaced complaint from a former civil servant in the DOD — where Brown headed the Defense Innovation Unit — was lambasted as a “national security disaster” by Slate’s longtime “War Stories” columnist, Fred Kaplan. Brown remains the head of DIU, but his appointment had been seen as a “potential sea change” for DOD, and the withdrawn nomination leaves the Pentagon’s acquisition operation in flux. A replacement has not yet been named.
CSET Job Openings
Please share with qualified candidates in your network or consider applying:
- Research Fellow – AI TEV&V will focus on the safety and risk of deployed AI systems by researching real-world AI incidents and use these identified incidents with other analyses of AI systems to inform policy recommendations regarding AI safety, test, evaluation, verification and validation (TEV&V) processes, standards setting and management, and the appropriate employment and operation of AI systems by businesses and the US Government (including the military). Applications due by August 1
CSET’s translations of significant foreign language documents on AI
CSET’s translations of significant foreign language documents on AI
PRC Cyber Plan: Open Solicitation of Opinions on the Three-Year Action Plan for the High-Quality Development of the Cybersecurity Industry (2021-2023) (Draft for Solicitation of Opinions). This July 2021 document is a draft PRC government plan for the near-term development of China’s cybersecurity industry. It names a host of specific cybersecurity technologies that the Chinese government is encouraging PRC companies to pursue. China allowed the public to comment on this draft plan, but only for a four-day period.
PRC Ministry of S&T Press Conference: Press Conference on the “Launch and Implementation of the National Key R&D Program”: Summary Transcript. This transcript of a 2016 PRC Ministry of Science and Technology press conference announces the creation of China’s overarching “National Key R&D Program.” Notably, the National Key R&D Program absorbs the former 973 and 863 Programs, two major PRC technology development plans that had been in place for decades.
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.
What’s New at CSET
- National Power After AI by Matthew Daniels and Ben Chang
- The Huawei Moment by Alex Rubin, Alan Omar Loera Martinez, Jake Dow and Anna Puglisi
- Ending Innovation Tourism: Rethinking the U.S. Military’s Approach to Emerging Technology Adoption by Melissa Flagg and Jack Corrigan
- China’s CyberAI Talent Pipeline by Dakota Cary
- Defense One: What China’s Vast New Cybersecurity Center Tells Us About Beijing’s Ambitions by Dakota Cary
- The Hill: China’s new software policy weaponizes cybersecurity research by Dakota Cary
- CSET: Data Snapshot: Measuring AI RC Growth by Autumn Toney
- House Intelligence Committee: Research Analyst Will Hunt testified before the Strategic Technologies and Advanced Research (STAR) Subcommittee of the House Permanent Select Committee on Intelligence last week. Read his testimony and watch the full subcommittee hearing.
CSET maintains a crowd forecasting platform. Sign up as a forecaster, and take a look at some of the predictions so far:
- (New) Will the United States have the world’s fastest supercomputer in June 2022?
- (New) Will the Chinese military or other maritime security forces fire upon another country’s civil or military vessel in the South China Sea between September 1, 2021 and February 28, 2022, inclusive?
- (New) When will 1 billion people in India receive at least one dose of a COVID-19 vaccine?
- On July 22, the CSET Webinar China’s Long-Term Investments in AI Growth: Bringing the Public and Private Sectors Together featured a conversation between Research Analyst Ngor Luong, Chinese STEM Translation Lead Ben Murphy and Director of Strategy Helen Toner on China’s industrial policy tools.
- Marketplace: American Public Media’s Nancy Marshall-Genzer reached out to Andrew Lohn to discuss cybersecurity in the wake of the Microsoft Exchange hack revelation.
- Wired: Lohn also spoke to Wired’s Will Knight for an article about the Pentagon’s cybersecurity exercises, which cited his recent brief, Poison in the Well.
- Wired: CSET’s director of strategy, Helen Toner, also appeared in a Wired article, this one a Tom Simonite piece about Congress’s efforts to regulate AI.
- Bloomberg: Research Analyst Emily Weinstein discussed Congressional efforts to enhance U.S. research security with Bloomberg’s Daniel Flatley and Janet Lorin.
- National Defense Magazine: For a piece about the recent CSET report Mapping India’s AI Potential, National Defense Magazine reached out to Research Analyst Husanjot Chahal.
- Nextgov: Earlier this week, Mila Jasper of Nextgov covered the new brief by Melissa Flagg and Jack Corrigan, Ending Innovation Tourism.
- Science Magazine: Dennis Normile cited two CSET reports — Assessing the Scope of U.S. Visa Restrictions on Chinese Students by Remco Zwetsloot, Emily Weinstein and Ryan Fedasiuk and The China Scholarship Council: An Overview by Fedasiuk — in an article about the visa problems experienced by Chinese students in the United States.
What We’re Reading (and Listening To)
Report: Defense Acquisition in Russia and China, Mark Ashby, Caolionn O’Connell, Edward Geist, Jair Aguirre, Christian Curriden and Jon Fujiwara, RAND Corporation (2021)
Podcast: Labs over Fabs: Why the US and EU Should Invest in the Future of Semiconductors, Jordan Schneider, Chris Miller and JP Kleinhans, China Talk Podcast (July 2021)
Report: A Proposal for Identifying and Managing Bias in Artificial Intelligence, Reva Schwartz, Leann Down, Adam Jonas and Elham Tabassi, NIST (June 2021)
Article: Reading Race: AI Recognises Patient’s Racial Identity in Medical Images, Imon Banerjee, et al., arXiv (July 2021)
The Future Society: Survey on AI systems used in COVID-19 pandemic response. Were you involved in developing an AI system used in COVID-19 pandemic response? Please support research being conducted by The Future Society, the GPAI AI & Pandemic Response Subgroup, and the OECD by completing this survey by August 2nd.
- August 25: UNIDIR, The 2021 Innovations Dialogue: Deepfakes, Trust And International Security, featuring Katerina Sedova