The true value of DeepSeek is underestimated!
DeepSeek-R1 has undoubtedly brought a new wave of enthusiasm to the market. Not only are the relevant so-called beneficiary targets rising sharply, but some people have even developed DeepSeek-related courses and software in an attempt to make money from it.
We believe that although these phenomena have a certain chaotic element, and we must be aware of the risks involved, it is undeniable that they reflect the public’s curiosity and enthusiasm for DeepSeek.
Previously, I analyzed the significance of the emergence of DeepSeek-R1, but today I would like to discuss in depth the real opportunity behind it, which is to promote the popularization and prosperity of AI applications. At the strategic level, I have always emphasized that continuous investment to improve performance is crucial.
When the technology has reached a certain stage of development, performance tuning and energy efficiency should become the focus in order to reduce costs and enhance competitiveness. DeepSeek has caused such a stir because it has trained a DeepSeek-R1 model with performance comparable to the OpenAI o1 model at a cost far lower than that of American AI giants such as OpenAI, Meta, and Anthropic. This has shown everyone the possibility of China’s technology industry breaking through US containment.
Moreover, some time ago, many experts believed that the Scaling Law was about to fail. As the size of AI models increases, it will become more and more difficult to obtain high-quality data, and the marginal effect of performance improvement will gradually weaken.
In addition, the sharp increase in the demand for computing power for large AI models will also bring serious energy consumption and environmental problems. This makes people feel that DeepSeek’s approach has great hope of reaching the top of large AI models.
However, I still agree with Huang Renxun’s view that the Scaling Law is still valid. Increasing investment in capital and computing power can still continuously improve model performance, and the ceiling for this kind of improvement is definitely much higher than performance tuning and energy efficiency. In other words, when we have optimized all the details that can be optimized, and then want to further improve performance, we can only rely on increasing investment.
Therefore, in the long run, relying solely on performance tuning may not be able to keep up with competitors who keep pouring money into improving performance.
Therefore, I think we still need to take a cool-headed look at the cutting-edge competitiveness of DeepSeek. But on the other hand, the actual value of DeepSeek may have been underestimated.
Leading AI companies such as OpenAI have invested a lot of resources in training and optimizing models, but have not solved the problem of application and developed the application market to support the development of these models.
High operating costs, complex computing processes, and data security and privacy issues have resulted in a continuous high demand for financing, which also limits the further expansion and application of these companies in the field of AI.
Can DeepSeek solve this problem? This requires a careful insight into the delicate balance between open source and closed source, performance improvement and market application.
On the one hand, DeepSeek’s open source approach is different from other models.
In the traditional sense, open source means that the code is completely open, and anyone can freely use, modify and distribute it, while the open source developer cannot profit from it. However, in the field of AI, open source is not just about opening up the code, but more importantly, about model training and optimization.
DeepSeek makes the model structure public and provides open source models that have been fully trained and optimized, which not only lowers the threshold for users, but also ensures the performance of the model. At the same time, DeepSeek also continuously collects user feedback and data through online services to continuously optimize model performance.
In the future, it may even be possible to adjust model parameters in real time based on user usage, thereby providing more efficient and personalized services.
In the future, similar to Meta, DeepSeek’s open source strategy will also attract developers and researchers from around the world to participate, forming a larger collaborative ecosystem. This model of cooperation will greatly promote the innovation and application of AI technology. At the same time, DeepSeek will also gain more technical support and business opportunities from this collaboration, achieving a win-win situation.
On the other hand, DeepSeek is expected to solve the problem of inclusiveness in the current AI application process. At present, many companies that do AI applications have already achieved considerable revenue, which shows that AI technology is already mature enough.
For example, Palantir, whose stock price has skyrocketed recently, has largely improved its operational efficiency and thus its profit margins by building its own AI platform. Not only did its revenue in the fourth quarter reach 800 million US dollars, far exceeding market expectations and shocking many people, but the number of users also increased significantly by 43%.
However, these successes still seem to belong only to large software companies. When we look at smaller companies and individuals, opportunities for entrepreneurs and startups are still limited.
The emergence of DeepSeek has broken this deadlock. Through innovative architecture and training methods, DeepSeek has successfully reduced the cost of developing and using AI models, making it possible for more people to try and use AI technology. This approach will not only promote the popularization of AI technology, but also help discover new application scenarios and needs.
Many companies have already developed low-cost applications using DeepSeek’s open source models, which further proves the feasibility and commercial value of the DeepSeek model. More new discoveries or applications may continue to emerge as DeepSeek develops, while the open source model allows more users to implement local deployment, further addressing the issue of data security.
In the future, with the emergence of low-cost, high-performance AI solutions, more and more people will begin to use AI technology, and new needs and application scenarios will continue to emerge, thus promoting the development of the entire AI industry.Whether it is the AI agent or the even more distant future, the development of AI will never stop.
To sum up, DeepSeek will help promote the emergence of some new trends in the current AI industry, that is, the development of general-purpose technologies has matured, and the development of supporting technologies and the application and commercialization of technologies will become even more important.
In the future, with the development of multimodal technologies and the continuous expansion of application scenarios, AI technology will play an important role in more fields, and it will also provide more development opportunities and space for emerging AI companies such as DeepSeek.
