The tech cold war continues: US announces bans on AI investments in China

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each weekday, we scour the web to aggregate the many stories related to artificial intelligence; we pass along the news, useful resources, tools or services, guides, technical analysis and exciting developments in open source.

In today’s Dispatch:

• DARPA, the US Defense Advanced Research Projects Agency, is launching a new two-year program called the ‘AI Cyber Challenge’ to spur the development of AI systems capable of rapidly identifying and fixing software vulnerabilities in critical government infrastructure code. Top performing teams will be eligible for $20 million in prize money and work with Google, Microsoft, Anthropic and OpenAI.

• The Japanese government has published new workforce development guidelines to help cultivate a new generation of skilled AI workers, emphasizing communication skills to properly leverage the technology while guarding against overreliance that could hinder personal growth.

• A new report finds that nearly half of all AI jobs in the US are concentrated in a few major cities, creating a stark geographic divide. The study's authors warn that federal policies and investments will be needed to distribute AI innovation and job growth more broadly across the country and prevent the formation of "tech deserts".

Plus: More US tech bans aimed at China, an incredible use case of AI in antibody development, AI stocks cooling off, trending tools and more!

US announces bans on AI investments in China

From CNN: Amid escalating technology tensions with Beijing, the Biden administration has unveiled new guidelines restricting US investments in advanced technology sectors in China. White House officials emphasized that the primary focus is on limiting Chinese access to cutting-edge American tech that may inadvertently benefit the Chinese military, making these efforts a matter of national security.

More details:

• The new regulations will impede investments by US private equity, venture capital firms, and joint ventures in China's artificial intelligence, quantum computing, and semiconductors sectors.

• The initiative seeks to block American knowledge and funds from aiding the Chinese military. According to officials, the objective is not to harm China's economy but to protect US national security interests.

• China’s Commerce Ministry conveyed that such a move disrupts the normal, fair functioning of economy and compromises the global trade order. Additionally, the Chinese Foreign Ministry emphasized their discontentment with the USA's continuous imposition of investment restrictions.

• US Treasury Secretary Janet Yellen conveyed the rules' objective is to delay China's technological and military progression.

• The regulations also mandate companies to report their investment activities in China, providing more transparency to the government.

Takeaways: This New York Times headline from July should give you an idea of how serious the tech sanctions on China are getting. The White House states their intent is “not to harm China’s economy” - but the sanctions in sum are so impactful (and the related technology so important) that some international experts have dropped the forecast of China ever overtaking the US as the top global economy.

Former South Carolina governor and Republican presidential nominee Nikki Haley called for even stronger sanctions: “To stop funding China’s military, we have to stop all US investment in China’s critical technology and military companies — period.”

AI is discovering highly-effective antibodies that humans couldn’t imagine

From WIRED (may be paywalled): Inside a repurposed old biscuit factory in South London, LabGenius is heavily leveraging AI in the development of new antibodies. A machine learning algorithm designs the antibodies to target specific diseases, then automated robotic systems grow them in the lab, test them, and feed the results back into the algorithm - all with little human supervision.

More details:

• Antibodies have been synthetically manufactured since the 1980s, but traditional human methods of designing them involve painstaking searches through millions of amino acid combinations, with a risk of missing optimal solutions.

• The model selects 700 initial options from a pool of 100,000 potential antibodies at a time, and then autonomously constructs, tests, and further refines them - enhancing the effectiveness of treatments with each iteration. This process takes six weeks.

Takeaways: The article doesn’t mention LabGenius’ progress thus far, but their website notes that they are using this AI platform initially in co-operative projects to find new antibody-based drugs for treating cancer and inflammatory diseases. The AI helps optimize these antibodies to have desired properties like stability, potency (ability to produce desired effect), and selectivity (ability to target specific cells). One area they are focused on is immunotherapeutics - antibody drugs that help the immune system fight cancer. Their AI can optimize "immune cell engagers" which get immune cells to attack cancer cells.

From our sponsors:

Water doesn't have to be boring!

​​Hint Water specializes in flavored water infused with fruit essences. Unlike sweetened beverages or artificially flavored drinks, Hint Water offers a refreshing and tasty alternative without the addition of sugar or diet sweeteners. By using purified water and natural fruit essences, Hint Water provides surprisingly accurate fruit flavors. We offer 8 delicious flavors to choose from. New customers get over 45% off your first order + FREE Shipping!

Try Hint today with code SWAPSTACK

Social media/YouTube:

• (Discussion) Humanity is on the brink of major scientific breakthroughs, but nobody seems to care [Reddit]

• A man left paralyzed by a motorcycle crash over a decade ago is walking again after a surgical procedure aided by artificial intelligence [Reddit]

• HeyGen CEO Joshua Xu shows how good AI Avatar’s are getting [X]

• Growing up with AI in China [X]

• Highlights from Nvidia’s SIGGRAPH keynote [YouTube]

Did you know?

Nvidia has unveiled their next-generation AI processor, the GH200 Grace Hopper Superchip. This new chip has triple the memory capacity of Nvidia’s current-gen H100, and is designed specifically to handle complex AI workloads like large language models and vector databases.

The high bandwidth and multi-chip scaling abilities of the GH200 aim to meet increasing demands for generative AI models in data centers. Systems with the GH200 are expected to be available in Q2 2024

Trending AI Tools & Services:

• Vimeo: just launched their ‘one-take video creation’ campaign with AI.

• Spinach: an AI scrum-master for devs

• Fynt AI: supercharge your financial department with automated GPT-powered financial insights, reconciliation, reporting and more.

• recast: Turn your want-to-read articles into rich audio summaries.