Artificial Intelligence (AI) is currently deployed to aid scientists in their scientific endeavors. However, it does not possess the capacity to independently conduct scientific research and discover new knowledge. Sakana AI, a research and development company based in Tokyo, has introduced the first fully AI-driven system for automated scientific discovery applied to machine learning research.
The new generative AI model system, dubbed The AI Scientist, automates the entire research process: generating innovative ideas, formulating methodology, conducting tests, and presenting its findings in a full scientific manuscript. It is capable of evaluating generated papers and writing feedback akin to human accuracy. Furthermore, the system can learn from its own reports, developing and improving ideas across multiple reports to improve the next generation of ideas, mirroring the human scientific community. The system is designed primarily for institutional and research-focused applications, and runs on more advanced computing environments, like servers or cloud-based platforms.
In its first demonstration, The AI Scientist conducted research in diverse subfields within machine learning research, discovering novel contributions in popular areas. While still in its early stages with certain constraints such as generating inaccurate or misleading papers, this advancement marks a significant step toward realizing the full potential of AI in the entire research process, including that of AI itself.
