Chinese researchers from leading universities and tech companies have published a joint paper outlining a five-stage roadmap for recursive self-improvement, aiming to develop systems that can build better versions of themselves without human intervention. The study presents the stages and frames them as a new front in the AI race with the US.
Recursive self-improvement roadmap
The paper, titled “The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement”, was published on Thursday and sets out a five-stage path for recursive self-improvement (RSI). It highlights a growing industry focus on automating the labour-intensive lifecycle of training, evaluating and fine-tuning AI models.
Authors include researchers from ByteDance, Tsinghua University and the Shanghai Artificial Intelligence Laboratory, among others. The team presents RSI as a systematic progression aimed at reducing human intervention in model development and optimization.
The study frames this work as part of broader efforts by Chinese institutions to advance automated model development capabilities and explicitly positions the research as competitive with US efforts in the field.
