top of page

This Week in AI: the United States and China vie for dominance

Writer's picture: Lewis D. SorokinLewis D. Sorokin

Updated: 4 days ago

It’s been a big week for pushing forward the Artificial General Intelligence (AGI) race between world powers. Here are the highlights:


The Stargate Project: America's answer to acceleration abroad.

In its first few days, the second Trump administration, along with leaders from private industry, announced a $500 billion initiative called the Stargate Project.


This partnership between OpenAI, Oracle, Softbank, MGX, Microsoft, Arm, and Nvidia will focus on private-sector investment in artificial intelligence development in the United States.

This project likely takes some influence from a November 2024 report by the US-China Economic and Security Review Commission which recommended that "Congress establish and fund a Manhattan Project-like program dedicated to racing to and acquiring an Artificial General Intelligence (AGI) capability," which heavily emphasized the importance of collaboration between government and private sector in ensuring that America stays ahead of the competition.


That report defines AGI as “systems that are as good as or better than human capabilities across all cognitive domains and would surpass the sharpest human minds at every task”; other definitions take a less absolutist approach and draw the line at surpassing 80% of humans in 80% of cognitive domains, or other numbers. All that is for sure is that the Turing Test was broken long ago.


DeepSeek: new open source AI king, or CCP propaganda engine?

DeepSeek is an AI company founded by Liang Wenfeng, the co-founder of a successful Chinese quantitative hedge fund which is the sole source of DeepSeek's funding.

In benchmarks, DeepSeek's R1 model performs nearly as well as OpenAI's o1 model and is 50x less expensive to run. Considering that DeepSeek's models are open source and can be run locally through Ollama, this is extremely promising.


At the same time, DeepSeek adheres closely with Chinese Communist Party policies and ideologies, even when running locally and not on Chinese cloud infrastructure. This is because the Cyberspace Administration of China (CAC) requires Chinese AI models to adhere to "core socialist values," provide "politically correct answers" that align with CCP policy, and to censor sensitive topics such as the Uyghur genocide and Tiananmen Square. Although the model is open source, the complex architecture of these models would make it no small feat to unravel the web of CCP alignment.


Now that DeepSeek has shown that open source reasoning models can perform at high levels for low cost, it is only a matter of time before the rest of the industry catches up and builds models that balance these factors without the political requirements of a regime known for stifling the flow of disfavored information.


Some closing thoughts on containment.

Bill Gates’ “favorite book on AI” is “The Coming Wave” by Mustafa Suleyman, in which Suleyman establishes what he calls "the containment problem" as the essential challenge of our age. He defines contained technology as "technology whose modes of failure are known, managed, and mitigated, a situation where the means to shape and govern technology escalate in parallel with its capabilities." The book argues that containment is crucial for creating space to experiment with new technologies and explore their potential to support human flourishing while protecting us from their worst possible outcomes.


As the race to develop the best technology continues, we must hope that world leaders believe in containment. Both the United States and China are pushing forward distinct visions for the future, and each has the technical chops and financial backing to make it so.

 

bottom of page