These Are The Four Ways That Make DeepSek Can Change Everything

DeepSek, will change the world's AI technology (photo: x @mohritaroh)

JAKARTA - DeepSek's market movement only lasted briefly. However, the launch of this high-performance and low-cost large-language (LLM) model from this Chinese startup is expected to have a long-term impact on technology, trade, and economic relations between the United States and China.

Here are four predictions regarding the dramatic impacts that may occur, VOI quoted by Reuters:

All innovations essentially aim to generate more at lower cost or, in economic terms, in the form of unit price deflation. The AI revolution became a real example of this. Prior to the launch of DeepSek, the dominant AI model cost had fallen by about 80% on an annual basis in the last two years. DeepSek simply accelerated this trend.

The price difference of up to 30 times between DeepSek token costs and OpenAI market leaders reflects the improvement in DeepSek's algorithms as well as a more aggressive price strategy than its competitors. This deflation trend is expected to continue as AI research advances and increased competition in the industry.

As AI access is cheaper and easier, its use will increase significantly, in accordance with the Jevons paradox, which is the more efficient a technology is, the greater its demand.

When the AI model becomes more commodity, the economic value will shift to the application, so more resources will be allocated to inference or the application of AI to certain tasks compared to model training.

It will also increase demand for custom XPUs, chips specifically designed for specific application tasks, compared to standard GPUs that are better suited for AI model training.

Nvidia has indicated that this shift may have started. In May 2024, the AI leader reported that inference-related demand grew faster than model training, with inference now accounting for 40% of the total demand.

With the wider application of AI, opportunities to participate in AI research are no longer limited to entities with large funding, opening up wider innovation opportunities among academics and other industries. For example, the DeepSek R-1 model has given birth to thousands of new open-source models based on its architecture.

DeepSeekhas succeeded in achieving its breakthrough by using a much less and less sophisticated chip than its US competitors, proving that innovation can be born from limitations.

While US export controls may limit DeepSek and its competitors in China in the short term, these restrictions will not stop their progress. In addition, the policy risks isolating US technology from the Chinese market, perhaps permanently.

This export control also appears counterproductive to US governments prioritizing improving the balance of payments deficit with China. As a Chinese minister said, "If we want to buy from the US, but they limit their exports... how can they reduce their deficit?"

Indeed, US politicians from various camps in recent years have called for stricter regulations against China, not the other way around.

The decision to adjust chip export controls will be a turning point in US-China economic relations and will not be taken easily. However, the pros and cons of current chip export restrictions are becoming more balanced and need to be carefully reconsidered.

Although DeepSek's breakthrough and its open-source approach initially raised concerns among US investors, many US technology leaders have welcomed this development.

Major cloud platforms such as Microsoft, AWS, and Hugging Face have integrated various DeepSek R-1-based models, and many of the company's leaders note that cheaper LLM models will increase demand for their cloud services, which ultimately increase their revenue.

In the future, businesses will greatly benefit from increased productivity and cost savings resulting from the wider application of AI, regardless of the origin of the technology. This is especially true in the US, where wages are higher than other countries and STEM talent is still limited.

A useful historical comparison is the friction between the US and Japanese automotive industries in the 1980s. Japanese automakers eventually introduced their sleek production method to the US, which significantly increased the productivity of the American automotive industry.

The growing AI landscape offers great opportunities for collaboration between these two global powers, especially in pursuing general artificial intelligence (AGI), a mission that has the potential to change the world. However, the ongoing tension between the US and China in trade, technology, and various other issues is at risk hampering progress by dividing the global tech landscape.

It is not yet clear whether this is the beginning of a new era of technology cooperation or a period of increasing global competition. However, one thing is certain, DeepSek reminds us of how quickly everything can change in this AI era.