Keep pulling the thread on Clement Delong.
China is currently the strongest contributor to open-source AI models.
Most startups and academic institutions in the United States using open-source AI are utilizing models developed in China.
There is a potential investment bubble in the specific domain of large language models distributed via APIs, rather than in the AI field as a whole.
The business model for large language models distributed via APIs shows significant revenue growth but has uncertain margins, long-term sustainability, and competitive moats.
Chinese companies are currently dominating the robotics industry and are likely to continue doing so.
Hugging Face added 2 petabytes of data to its platform in a single week.
Over the past few years, US-based Frontier Labs have increasingly shifted towards releasing their models behind closed-source APIs rather than as open source.
Deepseek, Qwen, and Kymi are Chinese organizations that are contributing massively to the field of open-source AI.
The safety concerns about releasing AI models, such as those raised about GPT-2 six to seven years ago, have historically proven to be overblown.
The Mythos model was recently deployed to its first international organization, located in South Korea.
Releasing AI models in open source is the safest approach because it enables a wider community to build both AI systems and the necessary safety and protection systems.
Hugging Face has shipped almost 10,000 units of its Le Robot Reachy Mini, making it one of the most widely distributed robots of the year.