Boris Power, Head of Applied Research at OpenAI, stated that the company has allocated 80% to 90% of its R&D resources to the development of GPT-7, GPT-8, and subsequent generational models. He believes that only entirely new foundational models can achieve a qualitative leap in performance, while incremental fine-tuning within the same generation is viewed merely as a short-term investment. He also pointed out that the core challenge facing current AI assistants is how to guide users in understanding and unlocking the technology's potential, as interaction methods are undergoing fundamental changes. The focus of cutting-edge AI R&D is rapidly shifting from single-point parameter optimization to long-term foundational innovations driven by goals and featuring highly autonomous collaboration capabilities.
