According to the Securities Times, on July 17, Richard Sutton, winner of the 2024 Turing Award and founder of enhanced learning, said at the main forum of the 2026 World Artificial Intelligence Conference that currently large models and multi-modal generation only complete model fitting of historical human data; they can only reproduce existing knowledge; they do not have the ability to independently discover new knowledge; problems such as factual biases and logical flaws are difficult to eradicate. The market has two extreme perspectives on AI: excessive hype and excessive fear. The global industry should adhere to multilateral cooperation for mutual benefit and win-win situation. In his view, the static labeling data model has reached the ceiling. Empirical labeling is the core route of next-generation AI. Currently, the stock of manual labeling resources continues to decline and cannot support long-term iteration of the model. Empirical intelligence can rely on environmental feedback to continuously correct cognition, and autonomous evolution is an irreversible technological trend.

According to the Securities Times, on July 17, Richard Sutton, winner of the 2024 Turing Award and founder of enhanced learning, said at the main forum of the 2026 World Artificial Intelligence Conference that currently large models and multi-modal generation only complete model fitting of historical human data; they can only reproduce existing knowledge; they do not have the ability to independently discover new knowledge; problems such as factual biases and logical flaws are difficult to eradicate. The market has two extreme perspectives on AI: excessive hype and excessive fear. The global industry should adhere to multilateral cooperation for mutual benefit and win-win situation. In his view, the static labeling data model has reached the ceiling. Empirical labeling is the core route of next-generation AI. Currently, the stock of manual labeling resources continues to decline and cannot support long-term iteration of the model. Empirical intelligence can rely on environmental feedback to continuously correct cognition, and autonomous evolution is an irreversible technological trend.