Mark Chen, the Chief Research Officer of OpenAI, expressed his congratulations to the DeepSeek team for developing an LLM model with reasoning capabilities comparable to OpenAI’s o1 model. Chen commended DeepSeek for independently discovering core concepts utilized by OpenAI in the development of o1. He stated, “…they’ve independently found some of the core ideas that we did on our way to o1.”
Congrats to DeepSeek on producing an o1-level reasoning model! Their research paper demonstrates that they’ve independently found some of the core ideas that we did on our way to o1.
— Mark Chen (@markchen90) January 28, 2025
Additionally, Chen noted that external criticism regarding cost may be exaggerated, explaining that developing an LLM with scalable capabilities involves two axes: Pre-training and Reasoning. While optimizing both axes simultaneously to reduce costs is possible, OpenAI still aims to scale and push processing power to the highest level in both aspects.
Chen observed that with the advancement of Distillation research, the development of models with improved capabilities and cost reduction will become more distinct. This separation implies that reducing costs (particularly by increasing latency, as in the case of DeepSeek R1) does not necessarily equate to enhancing model capabilities.
In conclusion, Chen reiterated OpenAI’s commitment to developing models with reduced costs. He expressed confidence in the current research roadmap and is excited to unveil more efficient models within the upcoming quarter and beyond.
TLDR: Mark Chen praised the DeepSeek team for reaching o1-level reasoning capabilities and highlighted the importance of balancing model capabilities and cost reduction in AI research and development.
Leave a Comment