AI News Roundup – U.S. announces further Chinese AI chip restrictions, Google’s AI weather prediction model, Trump names AI czar, and more

To help you stay on top of the latest news, our AI practice group has compiled a roundup of the developments we are following.

• • The U.S. Department of Commerce has introduced further restrictions on China’s semiconductor industry in an effort to block Chinese companies from accessing advanced AI chips, according to Reuters. The new measures announced by the Bureau of Industry and Security this past week target 140 companies, including chip equipment manufacturers like Naura Technology Group and several Chinese chip toolmakers, and impose export controls on advanced memory chips and chipmaking tools. U.S. Commerce Secretary Gina Raimondo stated that the goal is to prevent China from advancing its domestic semiconductor manufacturing capabilities, which could support military modernization projects. The restrictions also expand the foreign direct product rule, potentially limiting chipmaking equipment exports from U.S. allied countries like Israel, Malaysia, Singapore, South Korea and Taiwan, while exempting Japan and the Netherlands. In response, China banned exports to the U.S. of key materials used in high-tech manufacturing, including gallium, germanium and antimony. The conflict reflects the Biden administration’s ongoing efforts to hamper China’s semiconductor and AI technological development and comes just weeks before the expected swearing-in of President-elect Donald Trump, who is anticipated to maintain a tough stance on China

• • Google’s DeepMind AI subsidiary has developed a new AI model that can predict weather much faster and more accurately than existing methods, according to The New York Times. The model, called GenCast, can generate 15-day weather forecasts that outperform the world’s best predictions 97.2% of the time, according to a study published in Nature this past week, breaking through historical limitations on long-range forecasting. Unlike traditional methods that rely on room-sized supercomputers, GenCast uses machine learning models trained on 40 years of weather data to recognize atmospheric patterns and can generate probabilistic forecasts in minutes instead of hours. DeepMind claims that GenCast is particularly promising for tracking deadly storms like hurricanes, though experts note it will likely complement rather than replace existing forecasting techniques. DeepMind plans to make the AI agent and underlying code publicly available through Google’s Earth Engine and Big Query, allowing other scientists to test and build upon the technology.

• • U.S. President-elect Donald Trump has named David Sacks, a venture capitalist, as his administration’s AI  and cryptocurrency czar, according to the Financial Times. Sacks, an early PayPal employee and a close ally of Elon Musk, is a prominent Silicon Valley supporter of Trump who hosted a fundraiser for the presidential candidate in June. The appointment reflects Trump’s priorities to develop a laxer regulatory approach to AI, with Sacks also expected to work on creating a clearer legal framework for the cryptocurrency industry. His selection is seen as another step in bridging the historically Democratic-leaning Silicon Valley tech community with the Republican Party, following Musk’s prominent involvement in the incoming administration.

• • The rapid adoption of AI in China has propelled ByteDance’s Doubao AI chatbot to become the second-most widely used AI system in the world behind OpenAI’s ChatGPT, according to Nikkei Asia. Launched in June 2023, Doubao offers users diverse conversational experiences through various virtual avatars and has attracted 51.3 million users in China, compared to ChatGPT’s 258.16 million globally. While China is still considered technologically behind the U.S. in AI development, the country leads in practical AI applications, with nearly 60% AI adoption rates compared to around 30% in other advanced economies. Several Chinese tech companies, including ByteDance and the conglomerate Alibaba are integrating AI into other applications, such as Alibaba’s Zhixiaobao, which can connect to China’s ubiquitous Alipay payment service and help users with tasks like ordering food or hailing taxis. The country’s generative AI market is projected to grow to $3.81 billion by 2028, with intense competition among the 5,734 AI companies existing in China as of June 2023, though profitability remains a challenge and overseas expansion will require navigating different regulatory environments, especially amid rising tensions with the U.S.

• • OpenAI is partnering with Anduril Industries, a military contractor, to develop new AI technologies for the U.S. Department of Defense, according to The Washington Post. The partnership aims to enhance Anduril’s drone-related technologies, specifically focusing on systems designed to detect and neutralize unmanned aerial threats. This collaboration is part of a broader trend of tech companies increasingly engaging with military projects, with OpenAI recently modifying its previous policy that had barred military applications. The companies justify the partnership by emphasizing the technological competition with China and the need to maintain America’s national security edge. While OpenAI maintains that its technology cannot be used to develop weapons or harm people, the deal represents a significant shift in Silicon Valley’s approach to military contracts, driven by pragmatic considerations and a changing perception of defense technology among tech entrepreneurs.