For China's artificial intelligence innovationDeepSeekThe recently launched app of the same name quickly attracted a large number of users to download it. Its open-source artificial intelligence model "DeepSeek V3" surpasses Meta's Llama 3.1 in performance and is comparable to Anthropic's Claude-3.5 and OpenAI's GPT-4o. At the same time, the hardware computing power required behind the model is far lower than that of other competitors in the market, and the development cost is even less than US$600 million. In addition to attracting market attention, OpenAI CEO Sam Altman stated that the company will continue to focus on the current research route and emphasized that computing power remains important.
The sudden emergence of DeepSeek, which touted its ability to build artificial intelligence models at a cost of less than $600 million and emphasized that its artificial intelligence required only minimal computing power and expense to operate, made it awkward for many technology companies to claim that artificial intelligence must cost tens of millions or even over $XNUMX million to build. It even caused NVIDIA's past practice of promoting the use of GPUs as training accelerators to be questioned, impacting the stock prices of NVIDIA and other companies.
However, Sam Altman expressed his opinion earlier through "X", saying that the performance of DeepSeek's artificial intelligence model is indeed impressive, but he also emphasized that OpenAI will launch an artificial intelligence model with better performance, and at the same time stated that computing power is still an important foundation for the development of artificial intelligence technology.
deepseek's r1 is an impressive model, particularly around what they're able to deliver for the price.
we will obviously deliver much better models and also it's legit invigorating to have a new competitor! we will pull up some releases.
- Sam Altman (@sama) January 28, 2025
Mark Chen, OpenAI's Vice President of Research, also pointed out that DeepSeek controls execution costs through data optimization, hinting that OpenAI could achieve the same goal and even achieve greater cost reductions. He also emphasized that low-cost operating models do not necessarily lead to higher performance.
Congrats to DeepSeek on producing an o1-level reasoning model! Their research paper demonstrates that they've independently found some of the core ideas that we did on our way to o1.
— Mark Chen (@markchen90) January 28, 2025
As for other opinions on DeepSeek,Former Google CEO Eric SchmidtSome believe that the emergence of DeepSeek means that Chinese companies can compete with large American technology companies with fewer resources. They also call on the United States to expand the development of open source artificial intelligence models to gain an advantage at the turning point of global artificial intelligence technology competition. Former Intel CEO Pat Gelsinger also mentioned the importance of open source in his views on DeepSeek, and believes that DeepSeek will help reshape the increasingly closed artificial intelligence model building process.
On the other hand, many believe that DeepSeek highlights that small-scale AI models that have undergone data optimization and distillation can perform as well as, or even surpass, large-scale AI models in specific fields and conditions. This has led more companies to be more willing to create small-scale AI models that better meet specific computing scenarios and needs. There are also views that DeepSeek founder Liang Wenfeng's main goal is not to focus on longer-term development goals, but to gain more investment opportunities.
