Toshinao Sasaki thinks about "What was so great about the Chinese-made generative AI 'DeepSeek'?" Tech and Future Vol.13

The generative AI DeepSeek was suddenly announced in China, and attracted a lot of attention for its high performance with far less cost and computing resources than conventional AI models, and for the release of open source. As a result, NVIDIA shares plummeted, and the DeepSeek shock hit the market hard. How do you feel about this?

Toshinao Sasaki (hereinafter Sasaki): First, it is important to understand that DeepSeek has two aspects: an "application" and a "generative AI model."

From a security perspective, countries are wary of DeepSeek as an application. South Korea has taken measures to stop downloads of the app, and it is currently not available for download from the Apple or Google app stores. Italy, Australia, and the United States are also quite wary, and there are talks that people should refrain from using DeepSeek.

It can be used in Japan, but the Personal Information Protection Commission*1 has warned that Chinese law may apply. China's National Intelligence Law states that "Chinese people must cooperate with government intelligence gathering activities both at home and abroad." This is a warning that if you write anything critical of China, it may be discovered by the Chinese government and you may be detained if you travel to China.

While it has a suspicious image, DeepSeek as a generative AI model is attracting a lot of attention. What's great about DeepSeek is that it is open source.