The global AI arms race is intensifying, with businesses and nations vying for supremacy in artificial intelligence. Amid discussions peppered with the term 'safety,' the potential impact on humanity is profound, especially if missteps occur. While the stakes are sky-high, so too are the sums of money being discussed—ranging from millions to hundreds of billions of dollars. This disparity has sparked a debate about the true cost of AI advancement.
Enter DeepSeek, an AI lab based in Hangzhou, Zhejiang Province. Despite being relatively unknown, DeepSeek has disrupted the AI landscape with its AI chatbot application featuring DeepThink R1 reasoning mode. This chatbot distinguishes itself by explaining its thought process, offering users a glimpse into its 'reasoning'—a feature that appears both eye-opening and organically intelligent, especially for those not versed in the technicalities of AI. Astonishingly, DeepSeek achieved this with an investment reportedly around $6 million, a fraction of what industry giants are spending. This development has warped AI power rankings and challenged the conventional wisdom on what it takes to succeed in the AI arena.
DeepSeek's achievement is drawing comparisons and implications that could reshape the AI industry's future. Training their AI on a fraction of the resources utilized by leading U.S. companies like OpenAI, which ignited the generative AI revolution with ChatGPT, DeepSeek's success challenges the notion that massive investment is a prerequisite for innovation. NVIDIA, once known primarily for gaming GPUs and now a key player supplying AI chips, played only a modest role in DeepSeek's endeavor. The limited number of NVIDIA H100 chips used contrasts sharply with U.S. efforts to build expansive AI supercomputing clusters with hundreds of thousands of advanced chips. DeepSeek's approach underscores a potential shift towards more resource-efficient AI development.
Reference(s):
Million? Billion? Trillion? China's DeepSeek sparks AI costs debate
cgtn.com
