Self-improving AI the key step towards AGI but safety is key: Z.ai founder | South China Morning Post

In the account reported by the South China Morning Post, Tang said that using AI to build smarter systems would make AGI more achievable, and that it would need to be done safely.

The report identifies Tang by both roles — Z.ai co-founder and Tsinghua computer science professor — and attributes to him the description of recursive self-improvement as the main research problem. The safety point was made in the same remarks as the capability claim: in the report's phrasing, the approach would make AGI more achievable though it would need to be done safely. Both statements came from the Saturday event in Shanghai.

AGI, or artificial general intelligence, is the objective named in the article, and recursive self-improvement is described there in terms of using AI to build smarter systems. Safety is presented in the piece as a condition attached to that work.