‘AI Scientist’ Aims to Automate Scientific Discovery

Tokyo-based R&D company Sakana AI has released "The AI Scientist," an AI system designed to fully automate scientific research. The company claims this system is the first of its kind, capable of handling multiple aspects of the research process independently.

In its research paper, published Monday, the team says its approach “generates novel research ideas, writes code, executes experiments, visualizes results, describes its findings by writing a full scientific paper, and then runs a simulated review process for evaluation.

“In principle, this process can be repeated to iteratively develop ideas in an open-ended fashion, acting like the human scientific community,” it said.

The AI Scientist utilizes advanced Large Language Models (LLMs) to interact with the user, proposing and implementing new research directions, particularly in the field of machine learning. The AI Scientist is also fully open source, released under the Apache 2.0 license, which makes it legal to use, modify, and commercialize.

According to Sakana AI, each research paper generated by the system costs less than $15 to produce. This affordability is paired with a sophisticated simulated peer-review process that the system uses to evaluate its own work, emulating traditional peer-review methods.

As a proof of concept, The AI Scientist was enlisted to complete the entire research process of different established investigations, and drafted research papers.

Sakana AI's AI releases include an image generation model, a vision language model named EvoVLM, and another model designed to generate images of Japanese Ukiyo artwork. All are available on its Hugging Face repository.

The firm says it's main goal is “to create a new kind of foundational AI model based on nature-inspired intelligence.”

The AI Scientist announcement was met with some controversy. Omniscience, another AI startup, accused Sakana AI of not being the first to introduce such a system. Omniscience claims that its AI model, Omni, predates The AI Scientist.

Omni retrieves relevant information from users’ files as they write up their ideas, helping them create or write about different topics. One example of its work was an essay titled “The Unconstitutionality of California Senate Bill 1047,” which was written using Ommniscience’s model—and later analyzed with counter-arguments by a professional, human team.

The team replied to Sakana with a copy of its announcement of The AI Scientist, just replacing its name with their product. “Copying the text shows that the exact same thing can be said about our system, released months before,” they argued.