An ALS treatment discovered with AI’s help starts human trials, TSMC cuts off a high-end Chinese chipmaker, and even more U.S. export controls could be coming

Sign up
Worth Knowing

A Treatment for ALS Discovered With the Help of AI Starts Human Trials: Verge Genomics, a California-based biotech startup, announced it had begun human clinical trials of a therapeutic discovered with the help of an AI system. Verge isn’t the first AI-powered healthcare company to bring a therapeutic to the trial stage, but the milestone is notable for a number of reasons. The first is that it is targeting a disease, ALS, that has proved notoriously difficult to treat. The second is that Verge’s use of AI allows it to pursue an “all-in-human” approach, as opposed to traditional research that relies on animal testing. According to the company, its “ConVERGE” system used 11.4 million data points gathered from the tissue and genetic information of approximately 1000 ALS patients and healthy controls in order to identify 22 potential target mechanisms for treating the disease — a process that took only three months. AI has not yet proved to be a panacea for healthcare, but if successful, Verge’s methods could help speed future discoveries. The four years it took the company to get to clinical trials is a significant improvement over traditional methods, both in terms of time and cost.
TSMC Reverses Plan to Produce High-End Chips for Chinese Firm: Despite initial reports that it would continue manufacturing flagship GPUs for the Chinese company Biren, TSMC reportedly suspended semiconductor production for the firm after concerns arose that its chips would violate the United States’ recently imposed restrictions on shipments to China. Since its founding in 2019, the Shanghai-based Biren has been able to quickly design chips that boast specs seemingly on par with its U.S. competitors like Nvidia’s A100 GPU. With PRC chipmakers likely unable to satisfactorily manufacture chips at those performance levels (as we covered in August, SMIC can reportedly manufacture chips on a comparable 7nm process, but it’s not clear whether it can do so at volume), Biren chose to outsource production to Taiwan-based TSMC. While the U.S. government’s new restrictions (see our coverage in the last edition) had been aimed at blocking all sales of A100-level (and up) chips to China, even when produced abroad, Bloomberg reported that Biren believed its chips would not be restricted by the new U.S. regulations. But one day later, Bloomberg reported that TSMC was suspending production of Biren’s chips while it determines whether they exceed the thresholds set by the new U.S. regulations. Making TSMC’s job harder, Biren has reduced the specifications listed on its website to bring its more advanced chip into compliance. It’s unclear whether these changes will convince TSMC to resume manufacturing. Biren, for its part, seems to be planning for the worst — the company is rumored to be cutting a third of its workforce.
Two Approaches to Address Concerns About Sourcing AI Training Data: How can AI developers source their training data ethically? Two new stories highlight possible approaches:
This will probably not be the last we hear of the issue — as AI systems like image generators and language models become increasingly popular, the problem of how training data is sourced is likely going to get even thornier.
Government Updates

The DOD’s National Defense Strategy — An AI Strategy Could Be Next: Last week, the Pentagon released a public version of its National Defense Strategy, its congressionally mandated quadrennial review. The NDS (which was transmitted to Congress earlier this year) hits many of the same notes as the National Security Strategy (released earlier last month), focusing heavily on strategic competition with China, which it identifies as “the pacing challenge” for the DOD. Like the NSS, the NDS emphasizes the need to “build enduring advantages” by marshaling the strengths of the U.S. defense industrial base. To that end, the strategy:
  • Criticizes the DOD’s current acquisition system and calls for one that rewards “rapid experimentation, acquisition, and fielding” so that the military can quickly adopt emerging technologies. As Defense One Executive Editor Kevin Baron noted, that plan will likely be “music to the defense industry’s ears,” but it’s also not the first time the Pentagon has called for an overhaul of its acquisition system.
  • Says the DOD will implement reforms to better integrate its data, software and AI efforts.
  • Lays out a plan to strengthen the defense industrial base by boosting the DOD’s support for research institutions, small businesses and other technology firms.
  • Calls for aggressive changes to build out the DOD’s workforce and fill expertise gaps, such as in cyber, data, and AI. In addition to changes meant to attract existing talent (such as streamlining hiring, improving incentives and offering flexible work environments), the NDS also says the DOD will work with colleges and universities to expand the talent pool.
Now that the NDS has been published, another important document could be close behind: a March Government Accountability Office report said DOD officials expected an updated DOD AI strategy would be released soon after the NDS.

Export Controls Targeting AI and Quantum Computing Could Be on the Way: The Biden administration is considering additional export controls to target China’s access to AI software and quantum computing technology, according to a report from Bloomberg. The restrictions, if enacted, would add to the sweeping controls imposed last month that targeted China’s access to and ability to develop advanced node semiconductors (see our coverage here). The original Bloomberg report is scant on details about what specific products any new controls would target, and, as Protocol’s Joe Williams and Max A. Cherney noted earlier this week, quantum computing’s still-nascent development makes identifying potential chokepoints difficult. But it seems the Biden administration isn’t willing to wait around — during an event last week, Commerce Department Undersecretary for Industry and Security Alan Estevez said of the potential for new controls, “If I was a betting person, I would put down money.” The New York Times pinpointed the next meeting of the EU-U.S. Trade and Technology Council, set for next month, as a key moment for the new controls, as the United States is likely to need allied buy-in if its efforts are going to be successful.

Project Convergence — the Army’s Emerging Tech Showcase — Is Underway: Project Convergence 2022 — the third iteration of the Army-hosted joint exercise meant to test the military’s newest technologies — is set to wrap up later this month, having kicked off in September. This year’s exercise has been the biggest yet, involving all of the services as well as partners from the UK and Australia. The Army says approximately 300 technologies will be tested at this year’s event (up from 110 during last year’s exercise), including unmanned aerial systems, autonomous fighting vehicles, and advanced sensors. Because Project Convergence is focused on longer-term capabilities, the Army added a “technology gateway” experiment that gives industry partners a chance to get involved early in the process, get a sense of capabilities and gaps, and build capacity for future collaboration. Approximately 50 early-stage technologies were tested at this year’s “gateway” event. Officials from the Army, Air Force and Marine Corps said the exercise has also shown the services are “aligned” in their implementation of JADC2, the Pentagon’s plan to connect and coordinate its services’ sensors in a single network. Criticism of JADC2’s rollout has intensified in recent months, with officials as high up as Deputy Secretary of Defense Kathleen Hicks raising questions about the strategy. It remains to be seen whether the success of Project Convergence will help assuage those concerns.  

In Translation
CSET’s translations of significant foreign language documents on AI

PRC Think Tank Report: China’s New Generation AI Technology Industry Region Competitiveness Evaluation Index. This report, published by a think tank affiliated with China’s Nankai University, compares the strength of the AI industry in various Chinese provinces, cities, and urban agglomerations across an elaborately weighted collection of different metrics.

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.

CSET Job Openings

We’re hiring! Please apply or share the roles below with candidates in your network:
  • Fellow – Emerging Technology Supply Chains: We are currently seeking candidates to lead and coordinate our Emerging Technology Supply Chains Line of Research, either as a Research Fellow or Senior Fellow (depending on experience). This fellow will shape priorities, lay out an overall research strategy, oversee execution of the research and production of reports, and help hire and manage supporting researchers. Applications due by December 12.
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

REPORTSPUBLICATIONSEMERGING TECHNOLOGY OBSERVATORYEVENT RECAPSIN THE NEWS
What We’re Reading

Commentary: How Food Delivery Workers Shaped Chinese Algorithm Regulations, Matt Sheehan and Sharon Du, Carnegie Endowment for International Peace (November 2022)

Working Paper: Artificial Intelligence and Economic Growth, Philippe Aghion, Benjamin F. Jones and Charles I. Jones, NBER (October 2017)

Upcoming Events


What else is going on? Suggest stories, documents to translate & upcoming events here.