DeepSeek outpaces AI competitors days after release, Tech Analyst says: Video
Japanese tech analyst Lee Mu-yueh has said Chinese startup DeepSeek is a rising star in the artificial intelligence industry, asserting that the company is outpacing competitors.
DeepSeek recently made waves on Wall Street with the launch of a powerful new chatbot developed at a fraction of the cost incurred by other AI tech giants.
According to Lee, what sets DeepSeek apart is its decision to release the model as open-source, allowing researchers, developers, and users unrestricted access to the underlying code. “It’s not just a competition between the United States and China,” he told AFP, “but a battle between open-source and closed-source.”
Speaking at the Generative AI Conference, Lee elaborated on DeepSeek’s innovative advancements. “DeepSeek has launched an incredibly affordable model using an open-source approach. They’ve discovered a way to outpace other companies in the industry,” he said. While acknowledging OpenAI’s leading o1 model, Lee emphasised that DeepSeek not only matches it in several respects but also introduces a critical functionality: internet connectivity during inference—a feature o1 has yet to offer.
Lee praised DeepSeek’s efficiency in resource management during AI training. “Typically, training AI models requires a tremendous amount of GPU resources. DeepSeek has figured out how to squeeze the most out of a limited GPU allocation, focusing it effectively on AI model training. This optimization is something they’ve done remarkably well,” he explained.
A notable aspect of DeepSeek’s success is its development team, comprised solely of local Chinese university students. “None of the team members returned from overseas,” Lee remarked. “Their success stems from two key aspects: a strong financial and quantitative background, and a strategic approach to user engagement.”
Lee also addressed concerns about government influence on the AI model’s outputs. According to him, DeepSeek’s network service operates in two stages: the primary AI model generates an output, and a secondary model checks for alignment with Chinese government values.