The Commercial Real Estate Alpha Site

via Titan – Click here for complete story and to look into Titan further

That was fast … DeepSeek, a Chinese AI startup, is making headlines with the release of its R1 model, raising concerns in the U.S. about the shifting dynamics in artificial intelligence. Founded by hedge fund manager Liang Wenfeng, DeepSeek has developed AI models that rival those from American companies like OpenAI and Anthropic. Notably, DeepSeek’s R1 model matches the performance of OpenAI’s GPT but was developed at just 3% of the cost. How does DeepSeek stack up with Western AI platforms? For one, R1 employs reinforcement learning, which means it can self-improve without human supervision. This approach contrasts with models like GPT and Claude, which rely on supervised fine-tuning. Further, the open-source nature of DeepSeek’s models creates an advantage as people can collaborate, share ideas, and improve the technology at a faster clip. Meta’s chief AI scientist recently admitted that “open-source models are surpassing proprietary ones.”  Then there’s the cost advantage. The development cost of DeepSeek’s R1 was just $5.6 million, significantly lower than the $100 million to $1 billion range cited by other AI developers. This can be attributed to the self-learning nature of the technology, as well as lower development costs in China (e.g. access to cheap labor and infrastructure). DeepSeek excels at problem-solving, creative tasks, and research, but as one tester pointed out, there are limitations due to censorship.

DeepSeek’s release largely took Silicon Valley by surprise, and U.S. tech leaders are expressing concern that China’s rapid progress in AI could threaten the nation’s edge in economic and national security—a competition that’s poised to shape the future of global power.