OpenAI evaluated the artificial intelligence (AI) model of the Chinese startup "DeepSeek" as "not new" and began to restrain it. Although it acknowledged that China's AI technology has caught up with the United States to a considerable extent, it expressed concerns about the potential for misuse in strengthening authoritarian regimes.
According to Japan's Nihon Keizai Shimbun (Nikkei) on the 3rd, Sam Altman, the CEO of OpenAI, stated in an interview with the publication that DeepSeek's new AI model is "certainly a good model, but models of this level have already existed," adding that "OpenAI will continue to develop even better models in the future."
DeepSeek claimed that its AI model "V3," released in December last year, has similar performance to OpenAI's latest model while incurring significantly lower expenses. The development expense of V3 was $5,576,000 (approximately 810 million won), which is about 5.5% of OpenAI's development expense for "GPT-4" (approximately 1.45 trillion won).
In particular, DeepSeek announced that it developed its model utilizing the Nvidia "H800" chip released in 2022 instead of the latest high-performance chips. This chip was launched as a low-spec product exclusive to China due to U.S. export restrictions. In contrast, global big tech corporations like OpenAI have been using cutting-edge chips like the Nvidia "H100" for AI model development.
As DeepSeek developed high-performance AI models at a low cost and released them as open-source, some analysts suggested that OpenAI could be in crisis. In response, CEO Altman noted on X (formerly Twitter) on the 27th of last month that "DeepSeek R1 is an impressive model," but expressed confidence, stating, "We will provide a much better model."
Meanwhile, after allegations were raised that DeepSeek illegally collected OpenAI data for AI model training, OpenAI and Microsoft (MS) launched an investigation.
CEO Altman acknowledged that China's AI technology has "caught up with the United States to a considerable extent" but expressed concerns, stating that "there is a possibility that an authoritarian country could misuse AI to reinforce its regime."