A high-performance, low-cost artificial intelligence model recently unveiled by Chinese startup DeepSeek has sparked a wave of attention around the world. The DeepSeek-R1 is said to offer performance comparable to US alternatives, but the company claims that its development cost less than a tenth of the cost of those models. Chinese technology news portal 36Kr interviewed the company's founder, Liang Wenfeng, in July 2024.
Below are translated and edited excerpts from this interview.
DeepSeek-V2, a large language model introduced in May 2024, has caused a shock in Silicon Valley. Why do you think this happened?
In the US, innovation is happening on a daily basis. In this context, DeepSeek-V2 is nothing special. People in Silicon Valley were surprised because this model was developed by a Chinese company. So far, Chinese companies have only followed suit, but this time we have entered the field as an innovator.
DeepSeek publishes its innovations through open source code. Aren't you worried about the risk of other companies misusing it? How do you want to build your competitive advantage?
In the case of a disruptive technology that changes established thinking, competitive advantage can only be protected in a closed environment for a limited period of time. Although the US OpenAI keeps its source code private, it cannot prevent competitors from catching up. We will grow in the process, accumulate a large amount of know-how, and create a corporate structure and culture to incubate innovation. That is our strength.
Many Chinese AI model developers are struggling to commercialize their technologies. Why did DeepSeek decide to focus on basic research first?
Because we believe the most important thing is to ride the wave of global innovation. Chinese companies have so far been good at applying innovative technologies invented by someone else to develop products quickly and make money. But we should not think that this is the only way. Our original belief is not to pursue short-term profit, but to be at the forefront of technological development and help the whole ecosystem grow.
It is often said that the US is good at creating new technologies, while China is good at applying them to products and services. What is your view on this issue?
China should stop relying on the achievements of others and gradually shift to the role of innovator in line with its economic growth. During the past 30 years of information technology, China has chosen to make money instead of actively participating in innovation. Innovation is not necessarily driven by business, but can arise from sheer curiosity and the desire to create.
In the area of AI model development, even the most advanced technologies may not have an absolute competitive advantage. What is DeepSeek's approach to this challenge?
We believe that Chinese AI technology will not forever follow in the footsteps of its predecessors. US semiconductor company Nvidia has built its current position not just through the efforts of a single company, but through the support of technology communities and industries in the West. China's AI industry needs to create a similar ecosystem. China's domestic chip development has stagnated because of a lack of support from technology communities and access to the latest information. This is why China needs people at the forefront of technology development.
Do you plan to raise funds? One media outlet reported that your parent company is considering spinning off DeepSeek.
In the short term, we do not have a plan for fundraising. For us, a bigger problem than securing funding is the US ban on high-end chip exports.
Some say DeepSeek employs "genius" engineers. Is that true?
Our employees are not genius engineers. They are recent graduates from leading Chinese universities, PhD students and young people who have just finished their studies.
Yu Lili / gnews.cz
English 
Czech 
Spanish 
French 
Italian 
Russian 
Chinese 
German