Huang Renxun claims that currently AI is thinking for itself, practicing for itself, and implementing reinforcement learning.

On July 17th, at the Third China International Supply Chain Promotion Expo, NVIDIA CEO Huang Renxun stated: In the first decade, we focused on pre-training. We collected a lot of data, and even used AI to prepare some data for this pre-training. Later, we focused on human reinforcement training, where human feedback is added, allowing humans to teach AI to do things or for humans and AI to work together. Now, AI is capable of thinking and practicing on its own through reinforcement learning, generating a lot of data to complete tasks and make inferences by itself.