What is DeepSeek and why does it matter?

In 2022, the emergence of ChatGPT significantly impacted the AI landscape, and now DeepSeek, a new reasoning model from the Chinese company DeepSeek, is creating waves with its open-access approach. DeepSeek-R1, a model comparable to OpenAI's o1 and o3-mini, utilizes a chain-of-thought process to tackle complex problems, offering performance on par with other leading models at a lower cost by using H800 chips instead of the high-spec Nvidia H100 GPUs. Despite U.S. export restrictions on AI technology, DeepSeek has achieved this through innovations like the "mixture of experts" model, efficient training and inference techniques, and distillation. Their models, including DeepSeek-V3 and Janus-Pro-7B, perform competitively against major counterparts like GPT-4 and DALL·E 3, challenging Silicon Valley's dominance. The open nature of DeepSeek's models, however, has stirred controversy, particularly regarding data privacy and censorship issues, as well as allegations of using ChatGPT outputs without permission. As a result, their chatbot app has gained rapid popularity, although it faces scrutiny due to its Chinese origins and potential geopolitical implications.