Can self-alignment harness the "stampede" of AI

　　Since Open AI launched ChatGPT Big Model, AI Big Model has entered the public eye, the discussion fever has soared, and Internet giants such as Baidu, Tencent and Ali have announced their investment and layout roadmaps for AI Big Model.

　　However, in the midst of the fervor, there are also many tech giants who have expressed their concerns about AI big models. The use of artificial intelligence to write papers and create fake pictures ...... brings convenience to humans at the same time, the "young" artificial intelligence technology has also exposed many security risks. According to media reports, Jeffrey Hinton, who is known as the "Godfather of AI", has resigned from Google in order to talk about the dangers of AI technologies such as AI Big Models in complete freedom.

　　Is the wild growth of AI big models really out of control? Researchers from the Language Technologies Institute at Carnegie Mellon University, the International Business Machines Corporation Research Institute, and other institutions have proposed a completely new approach - self-alignment. This approach combines principle-driven reasoning with the generative power of big models, enabling "self-regulation" of AI big models with minimal human supervision.

　　Aligning with human values

　　Currently, even the most advanced AI big models are not self-regulating and still rely on human instructions and annotations to adjust accordingly.