This cheap Chinese AI is trying to compete with Silicon Valley giants! All about DeepSeek

Produced by Tarun Mishra

DeepSeek Challenges AI Norms

A Chinese start-up, DeepSeek, has unveiled DeepSeek-V3, an AI model performing at par with systems by OpenAI and Google. This milestone was achieved using only a fraction of the computer chips typically required for training advanced AI.

Efficient Training

DeepSeek-V3 was developed using 2,000 Nvidia chips, significantly fewer than the 16,000+ chips often employed by leading AI firms. The system cost about $6 million to train, far below the $100 million or more spent by U.S. tech giants.

Innovations

US trade restrictions on advanced chips have compelled Chinese researchers to optimise AI development. DeepSeek’s success underscores how limitations can drive innovation in AI training methodologies.

Open Source

DeepSeek has open-sourced its AI model, enabling global researchers and businesses to build on its technology. This aligns with a growing trend of fostering innovation through freely shared tools and resources

Global Implications

Experts warn that stifling open-source AI development in the U.S. could give China a competitive edge. The shift in the open-source ecosystem toward Chinese leadership raises concerns about long-term technological dominance.

Economic Viability

DeepSeek’s efficient approach demonstrates how small-scale operations with modest budgets can compete in AI development. This may lower entry barriers for more companies globally.

Broader Impact

DeepSeek’s achievements highlight a potential shift in AI research and development dynamics, emphasising cost-effective innovation and challenging assumptions about resource-intensive AI systems.