InfoQ 中文T2·68 pts
- LLMs have eliminated rule-based and experience-driven traditional NLP methods through data-intensive brute force, causing top linguists to change fields.
- The key indicator of AI evolution is no longer benchmark scores but the model's ability to maintain long-term autonomous reasoning without human intervention (Cognitive Runtime).
- Physical world tasks progress slowly due to the lack of low-cost self-feedback mechanisms found in Go or math, while excessive reliance on AI for complex problems is eroding human independent modeling and deep thinking capabilities.