Jiang Daxin, the founder of Jieyue Xingchen, said that the integration of the generalization ability and multimodal understanding generation of reinforcement learning models is the key to AI development in the future.

On September 19th, at the 2024 Yunqi Conference, Dr. Jiang Daxin, the founder and CEO of Jieyue Xingchen, said that AI development is undergoing a crucial technological paradigm iteration: OpenAI’s large-scale model o1 has explored a way to make AI have the ability to think slowly like humans (to actively reflect and correct complex thinking) through reinforcement learning. The key to further breakthroughs in AI technology is to improve the generalization ability of reinforcement learning models and accelerate the integration of multi-modal understanding and generation.