The results of "The AI Scientist," a collaborative project between Sakana AI, the University of British Columbia (UBC), the Vector Institute, and the University of Oxford, have been published in the prestigious journal Nature. This ambitious initiative aims to complete the entire machine learning research lifecycle—from idea generation and experimental implementation to manuscript writing—using autonomous agents based on large language models.

The core of this research lies in the fact that AI no longer functions merely as a supplementary tool, but rather autonomously drives the process of scientific discovery itself. In tests conducted by the development team, unedited AI-generated papers were submitted to an ICLR 2025 workshop. Remarkably, these papers received evaluation scores higher than 55% of those written by humans and were officially accepted. This milestone symbolizes that AI-driven research has moved beyond the experimental phase and has reached a level capable of passing peer-review processes alongside human experts.

Furthermore, the study revealed the existence of a "Scientific Scaling Law" through the use of automated peer-review systems. This law suggests that as the capabilities of the underlying AI models improve, the quality of the resulting research papers increases proportionally. Such a discovery indicates that the advancement of computational resources and model sophistication could exponentially accelerate the pace of scientific progress. However, the technology is still evolving, with AI-specific challenges such as a lack of true originality and citations that are occasionally inaccurate still being observed.

Sakana AI is emphasizing a responsible development framework by adding watermarks to AI-generated papers and obtaining prior approval from ethics committees. As we enter a new era where the primary agent of scientific discovery expands from humans to AI, this technology is expected to become a tireless companion in overcoming diseases, protecting the global environment, and unraveling the mysteries of the universe. This paradigm shift, which dramatically transforms the process of discovery, holds the potential to fundamentally redefine the future of science.

The results of "The AI Scientist," a collaborative project between Sakana AI, the University of British Columbia (UBC), the Vector Institute, and the University of Oxford, have been published in the prestigious journal Nature. This ambitious initiative aims to complete the entire machine learning research lifecycle—from idea generation and experimental implementation to manuscript writing—using autonomous agents based on large language models.

The core of this research lies in the fact that AI no longer functions merely as a supplementary tool, but rather autonomously drives the process of scientific discovery itself. In tests conducted by the development team, unedited AI-generated papers were submitted to an ICLR 2025 workshop. Remarkably, these papers received evaluation scores higher than 55% of those written by humans and were officially accepted. This milestone symbolizes that AI-driven research has moved beyond the experimental phase and has reached a level capable of passing peer-review processes alongside human experts.

Furthermore, the study revealed the existence of a "Scientific Scaling Law" through the use of automated peer-review systems. This law suggests that as the capabilities of the underlying AI models improve, the quality of the resulting research papers increases proportionally. Such a discovery indicates that the advancement of computational resources and model sophistication could exponentially accelerate the pace of scientific progress. However, the technology is still evolving, with AI-specific challenges such as a lack of true originality and citations that are occasionally inaccurate still being observed.

Sakana AI is emphasizing a responsible development framework by adding watermarks to AI-generated papers and obtaining prior approval from ethics committees. As we enter a new era where the primary agent of scientific discovery expands from humans to AI, this technology is expected to become a tireless companion in overcoming diseases, protecting the global environment, and unraveling the mysteries of the universe. This paradigm shift, which dramatically transforms the process of discovery, holds the potential to fundamentally redefine the future of science.