The self-improving AI model is a topic of conversation – and for good reason

If you feel like AI is evolving too fast to keep up, a group of Chinese researchers have bad news: They’ve developed a model that “evolves” on its own, creating better versions of itself with each self-analytical loop.

ASI-Evolve, developed by researchers at Shanghai Jiao Tong University, operates on an infinite loop that reflects how people would put this type of technology through its paces. Essentially, it creates variations of AI models, changes the way they are trained, and adjusts the data they learn from. It then conducts its own experiments to find out which clone performs better and uses those results to guide its next experiments.

“ASI-Evolve extends standard evolutionary agents with two key components: a cognition base that feeds accumulated human priors into each round of exploration, and a dedicated analyzer that distills complex experimental results into reusable insights for future iterations,” the researchers write. “As far as we know, ASI-Evolve is the first…