China's AI Breakthrough Sparks Deepseek Shock

Despite boasting performance on par with top-tier models, Deepseek was developed at a fraction of the cost, raising questions about its impact on the global AI landscape and the potential for renewed friction between the United States and China.

The disruption, now being referred to as the "Deepseek Shock," echoes the historical "Sputnik Shock" of 1957, when the Soviet Union successfully launched the world’s first artificial satellite, leaving the United States scrambling to catch up in the space race. In this case, Deepseek’s advancements in AI technology have stunned the industry, positioning China as a formidable competitor in the field. The company behind this innovation is Deepseek, founded in 2023 by Lian Wenfeng, a former hedge fund entrepreneur born in 1985. After achieving success in algorithm-driven stock trading, Lian shifted his focus to artificial intelligence, leading to the development of Deepseek’s advanced AI model.

What makes Deepseek’s AI particularly remarkable is its rapid development timeline and significantly lower costs. The core model was completed in just two months, with a total development cost of 900 million yen—less than one-tenth of what OpenAI reportedly spent on creating ChatGPT. Yet, its performance is said to be comparable to ChatGPT’s September 2023 version. Like its Western counterparts, Deepseek’s AI is capable of engaging in interactive dialogue and answering a wide range of questions. For instance, when asked how one should approach negotiations with former U.S. President Donald Trump in under 30 characters, the AI responded, "Maintain a strong and firm stance to gain the upper hand." However, as a Chinese-developed model, it also enforces certain restrictions. When questioned about the 1989 Tiananmen Square incident, it declined to answer, instead responding with, "I cannot provide an answer. Please change the topic."

A major factor behind Deepseek’s cost efficiency is its innovative approach to AI training. The company leveraged publicly available open-source AI models, allowing it to develop its system without relying entirely on proprietary technology. Using a "teacher-student" learning model, Deepseek’s AI was trained by querying larger, more advanced external AI models, which acted as "teachers" to its smaller-scale "student" AI. This method enabled Deepseek to bypass the need for expensive, high-performance semiconductor chips that are typically required for training large-scale AI systems. Additionally, Deepseek has made its own AI model open-source, allowing researchers and businesses to refine and customize it for their own applications. AI expert Takahiro Yasuno noted that this breakthrough could democratize AI development, making it accessible to companies without significant financial resources. He also emphasized that Deepseek’s AI could serve as a foundation for developing new applications tailored to specific needs.

Despite its impressive achievements, Deepseek has also come under scrutiny over allegations of improper data usage. According to a report by Bloomberg, the company may have utilized proprietary data from other firms without permission to train its AI model. The U.S. government has expressed concerns, with officials suggesting that American intellectual property may have been compromised. Given the already tense relations between the United States and China over technological competition, Deepseek’s emergence could further inflame geopolitical tensions.

When asked about the controversy, Deepseek responded by emphasizing the importance of international cooperation. "There are ways to ease tensions through international cooperation and dialogue," the company stated. As the global AI race accelerates, Deepseek’s rise marks a turning point, demonstrating China’s growing capability in cutting-edge technology and reshaping the competitive landscape of artificial intelligence.

Source: KTV NEWS