✨ DeepSeek, China's new AI challenge

America shouldn't fear competition. We should adapt and accelerate. Hear that, Washington?

When I give public talks about my conservative futurist, Up Wing vision, I of course talk about what’s been happening with artificial intelligence since OpenAI’s 2022 rollout of ChatGPT. And I almost always present my audience with this hypothetical:

Imagine what the past two years would have been like had the company at the heart of the emerging Age of AI been based in China rather than America. Imagine the total freak-out, perhaps one surpassing our reaction in 1957 when the Soviet Union launched Sputnik, the first artificial satellite. If you’re sick and tired of hearing about AI now, that’s nothing compared with the deafening level of public discourse that would have happened in my scenario. It would have been wall-to-wall lamentation about American decline, the Chinese techno-military threat, our diminished stature on the global stage, and the utter failure of the American innovation system — both Washington and Silicon Valley. The Late, Great, United States.

Well, I think we’re about to get a test of that scenario in the coming days. Let’s talk about DeepSeek, a Chinese AI startup founded by hedge fund manager Liang Wenfeng, who runs the High Flyer trading firm. This week, they released their R1 model, an AI system matching capabilities of leading chatbots like those from OpenAI, Anthropic, and Google. What's remarkable is they apparently achieved this using only 2,000 Nvidia chips (versus competitors' 16,000-plus) and about $6 million in computing costs. They says built their system by leveraging efficient training methods and open-source technologies, partly driven by US chip restrictions. From the Financial Times:

DeepSeek’s engineers know how to unlock the potential of these GPUs, even if they are not state of the art,” said one AI researcher close to the company.

Which leads us to all the issues raised by DeepSeek’s success. Among them: