Worth Knowing
Google Quashes DeepMind’s Bid for More Independence: Google executives reportedly have rebuffed efforts by DeepMind — its esteemed AI subsidiary — to gain greater independence. According to The Wall Street Journal, which broke the story, DeepMind leaders had been in negotiations with Google regarding the future of their relationship for several years, but Google cut off those talks last month, effectively ending DeepMind’s pursuit of an independent legal structure and more control over research decisions. Since being acquired by Google in 2014, London-based DeepMind has produced a number of groundbreaking AI systems — from its board game-mastering AlphaGo to its protein-folding AlphaFold (which we covered last year). Those advances, though, may not have been possible without the financial backing of a tech giant like Alphabet, which has floated the outfit through hundreds of millions in operating losses over the years.
Amazon’s One-Year Ban on Police Facial Recognition Extended Indefinitely: Last week, Amazon announced that its self-imposed ban on selling facial recognition software to police would continue indefinitely. The ban, originally announced after the murder of George Floyd and the protests against police violence that followed, was set to last one year, meaning it would have expired in June 2021. At the time, Amazon said it hoped the moratorium would give Congress enough time to enact regulations on police use of facial recognition systems. While a handful of bills have been introduced, the technology remains conspicuously unregulated at the federal level, leaving decisions on its limits up to the companies themselves and the handful of states, cities and municipalities that have enacted controls. While other tech giants — among them IBM and Microsoft — have made similar decisions, police agencies around the country still have access to facial recognition platforms through companies like NEC and Clearview AI.
- More: In Machines We Trust Podcast: Land of a Billion Faces | DC-area facial recognition system halted
- In a paper published in the journal Nature, researchers from MIT, the National University of Taiwan and TSMC detailed a breakthrough in monolayer transistors that could lead to 1 nanometer semiconductors. In the last edition of policy.ai, we covered IBM’s development of the world’s first 2nm semiconductor, which is still years away from reaching consumers. In the case of 1nm chips, that timeline is likely even longer. But the finding is an important sign that Moore’s Law — Intel co-founder Gordon Moore’s eponymous observation that the number of transistors on a chip doubles every two years — might still be relevant, even as semiconductors begin to push the physical limits of silicon.
- In another paper published in Nature the same day, researchers from Intel and QuTech (a Dutch collaborative venture) showed they had successfully eliminated a bottleneck that has plagued quantum computing development. The paper showed that Intel’s Horse Ridge II cryogenic control chip — which, unlike current quantum control chips, can operate inside a cryogenic refrigerator — maintained control at the same level of fidelity as room-temperature hardware without the bottleneck that comes from wiring individual qubits inside a refrigerator to control electronics outside.
- Google, meanwhile, announced its new Quantum AI campus in Santa Barbara, as well as its plans to have a “useful, error-corrected quantum computer” by 2029. That timeline is an ambitious one — Google says it will need a 1-million-qubit machine to perform its desired functions, while its current systems have fewer than 100 qubits. According to The Wall Street Journal, Google is planning to pour “several billion” dollars into the effort.
Government Updates
Endless Frontier Act Transformed into Larger U.S. Competitiveness Package: Since we covered its passage out of committee in the last edition of policy.ai, the Endless Frontier Act has undergone a significant transformation. In addition to a name change — it is now the U.S. Innovation and Competition Act — the bill has expanded to 1,420 pages after being merged with several related pieces of legislation. Originally designed to establish a new National Science Foundation Technology Directorate and authorize $100 billion to support innovations in key emerging technology areas, the bill now provides nearly $200 billion in funding — including $52 billion in emergency appropriations to support semiconductor manufacturing and research incentives — and includes a host of new provisions related to the U.S. research and development ecosystem, funding for semiconductor manufacturing support, and research security. However, Sen. Young, one of the bill’s original sponsors, and others have expressed frustration that significant portions of the $100 billion meant for the new NSF Directorate have been diverted to other programs and that language viewed as outside the bill’s intended scope has been added. Senator Schumer has initiated the process to try to end debate and hold a final vote, but negotiations are ongoing. For more on this and other Congressional news, see our latest legislative roundup.
May the Air Force Be With You: Last month, the United States Air Force’s autonomous UAV system — dubbed “Skyborg” — completed its first flight. The system was deployed aboard a Kratos UTAP-22 Mako — a low-cost drone — for its 130-minute flight, during which it performed a handful of basic maneuvers. Skyborg, which Air Force officials have compared to R2-D2, is meant to fulfill multiple roles: as an assistant to a human pilot and as an autonomous system capable of flying “semi attritable” aircraft — aircraft inexpensive enough that their loss in combat is tolerable — in missions too risky for human pilots. The successful flight is the first step toward fielding Skyborg-powered aircraft in regular operations, which the Air Force has said it plans to do by 2023. Additional tests are planned for the next several months, during which the Skyborg system will perform more demanding tasks, such as teaming with human pilots and flying multiple autonomous systems together at the same time.
Back Office Processes and Antiquated Tech: The CIA is finally set to retire its secure fax machines in favor of a special email network. The agency has used faxes to conduct classified procurement-related discussions with industry partners, but announced in 2019 that it would be switching over to the comparatively cutting-edge email. A related story on behind-the-scenes tech upgrades emerged from the Pentagon’s Joint Artificial Intelligence Center earlier this month: By applying AI, machine learning and automation to the DOD’s back office functions — such as personnel management, accounting and health management — JAIC officials say they have managed to shrink tasks that previously took 20 minutes down to a handful of seconds. JAIC’s project is still in its early stages, but Deputy Director Bryan Lane told Breaking Defense that the programs are being scaled up in both the Army and Navy and could be expanded across the DOD.
CSET Job Openings
Please share with qualified candidates in your network or consider applying:
- Policy Communications Analyst will help advance our written and visual products by advising researchers on report clarity, narrative flow, cogency, presentation effectiveness and parsimony. Applications due by June 4
- 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 July 1
In Translation
CSET’s translations of significant foreign language documents on AI
CSET’s translations of significant foreign language documents on AI
PRC Ministry of S&T Guidelines: Application Guidelines for the 2018 First Annual Batch of Projects for the “Technology and Equipment for Public Security Risk Prevention and Control and Emergency Response” Key Special Project (Judicial Thematic Tasks). This announcement details the PRC government’s ambitious plan to use AI and other advanced information technologies to improve the Chinese justice system. These government-issued guidelines encourage Chinese corporations and researchers to apply facial recognition and other automated electronic search and discovery technologies to evaluate judges, identify relevant legal precedents, predict recidivism, and so on.
PRC Ministry of Education Article: 35 Key “Stranglehold” Technologies. This article briefly describes 35 technologies that China must import because it is unable to produce them domestically in sufficient quality or quantity. It expresses concern that key Chinese industries would be severely hampered if China’s supply of these technologies were to be cut off. This is a PRC Ministry of Education summary of a series of 35 articles—each profiling a different “stranglehold technology”—that a Chinese government-run newspaper published in 2018.
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
REPORTS
- China’s Artificial Intelligence Industry Alliance: Understanding China’s AI Strategy Through Industry Alliances by Ngor Luong
- Truth, Lies, and Automation: How Language Models Could Change Disinformation by Ben Buchanan, Andrew Lohn, Micah Musser and Katerina Sedova
- China’s Foreign Technology Wish List by Ryan Fedasiuk, Emily Weinstein and Anna Puglisi
- OECD: Public consultation on the OECD Framework for Classifying AI Systems. Earlier this year, the OECD.AI Network of Experts (which is co-chaired by CSET Interim Director Dewey Murdick and includes CSET’s Acting Director of Data Science Catherine Aiken) developed the OECD Framework for Classifying AI Systems. The OECD has just launched a public consultation to solicit feedback on the framework. Take a survey testing out the framework on an AI system of your choice and submit comments on the report by June 30 by following the link above.
- ChinaFile Conversation: How Should the U.S. Respond to China’s Military-Civil Fusion Strategy? featuring responses from Emily Weinstein and Lorand Laskai
- CSET: Legislative Roundup: An S&T-Packed Spring Legislative Agenda by Daniel Hague
CSET maintains a crowd forecasting platform. Sign up as a forecaster, and take a look at some of the predictions so far:
- (New) What will the ‘Big 5′ (Alphabet, Amazon, Apple, Facebook, and Microsoft) tech companies’ average reputation ranking be in the 2022 Axios Harris poll?
- (New) What percentage of U.S. AI publications will have a Chinese co-author in 2022?
- Axios: Axios Future’s Bryan Walsh covered the release of the new CSET report, Truth, Lies, and Automation, by Ben Buchanan, Andrew Lohn, Micah Musser and Katerina Sedova.
- Wired: Axios wasn’t alone — Wired’s Will Knight took an in-depth look at the report and spoke to Buchanan about its implications.
- Politico: More launch-day attention from Politico, which noted the report in its Morning Tech newsletter.
- Vice: Vice covered the report and discussed the potential impact of language models on the spread of conspiracy theories like QAnon.
- Nextgov: The report also received a writeup from Nextgov.
- War on the Rocks: An article about U.S. semiconductor policy priorities cited the September brief The Chipmakers: US Strengths and Priorities for the High-End Semiconductor Workforce by Will Hunt and Remco Zwetsloot.
- ChinAI: Jeffrey Ding’s ChinAI newsletter recommended Ngor Luong and Zachary Arnold’s new brief China’s Artificial Intelligence Industry Alliance: Understanding China’s AI Strategy Through Industry Alliances as a featured “Should-read” publication earlier this week.
- The Wire China: Didi Kirsten Tatlow reached out to CSET’s Daniel Chou for a piece about News Break, a news aggregation app, to discuss its privacy implications.
- Analytics India Magazine: An article in Analytics India Magazine summarized the March data brief Mapping India’s AI Potential by Husanjot Chahal, Sara Abdulla, Jonathan Murdick and Ilya Rahkovsky.
- GCN: For an article about cybersecurity in the federal government, GCN spoke with CSET Senior Fellow Andrew Lohn on the impact that AI has had on cybersecurity.
- Federal News Network: FNN’s Tom Temin interviewed Research Fellow Tim Hwang about the research agenda of U.S. corporate AI labs, the subject of a recent CSET paper, for the Federal Drive podcast and drive-time broadcast.
What We’re Reading
Article: People Are Using an Ancient Method of Writing Arabic to Combat AI Censors, Hakim Bishara, Hyperallergic (May 2021)
Article: A “Horizon Strategy” Framework for Science and Technology Policy, Christopher Ford, Charles Clancy and James M. Blackburn, The MITRE Corporation (May 2021)
Report: Artificial Intelligence and Autonomy in Russia, Jeffrey Edmonds, Samuel Bendett, Anya Fink, Mary Chesnut, Dmitry Gorenburg, Michael Kofman, Kasey Stricklin and Julian Waller, CNA (May 2021)
Upcoming Events
- June 22: CSET Foretell, Foretell in Conversation: A New Approach to Geopolitical Forecasting with Professor Michael Horowitz, moderated by Michael Page
What else is going on? Suggest stories, documents to translate & upcoming events here.