But with the improvement of LLM (Large Language Model) reasoning abilities, especially in grasping human so-called common sense or established theories, is it possible to achieve more interesting applications based on this foundation?
Recently, I have extensively used models such as China's Kimi, DeepSeek, Tongyi, and the US's Gemini, OpenAI, etc. I found that current large models mainly serve as high-level personal assistants. They struggle to be effective in industrial production and team collaboration.
The biggest issue is their accuracy, which requires human judgment. However, they significantly enhance human decision-making efficiency and information processing capabilities, aligning with my initial assessment.