FutureHouse, a nonprofit supported by Eric Schmidt with the goal of creating an “AI scientist” in the next decade, has introduced a new tool designed to assist in “data-driven discovery” in the field of biology. This tool, called Finch, was released shortly after FutureHouse launched its API and platform.

Finch works by taking biology data, primarily from research papers, and a prompt (like “What can you tell me about molecular drivers of cancer metastasis?”). It then runs code, generates figures, and analyzes the results. FutureHouse co-founder and CEO Sam Rodriques described Finch as being similar to a “1st yr grad student.”

According to Rodriques, the ability to perform these tasks in minutes is like having a superpower. He added that Finch has uncovered some exciting findings, proving to be valuable for internal projects at FutureHouse.

FutureHouse, like many other startups and tech giants, believes that AI tools such as Finch will eventually automate various steps in the scientific process. However, evidence supporting this claim is currently lacking. Despite its efforts, FutureHouse has not yet made any significant scientific breakthroughs using its AI tools.

In the field of biology, especially drug discovery, AI companies see great potential. The market for AI in biology was valued at $65.88 billion in 2024 and is projected to reach $160.31 billion by 2034.

Although AI has shown some success in drug discovery, it has not provided an instant solution in the lab. Companies like Exscientia and BenevolentAI, which utilize AI for drug discovery, have experienced setbacks in clinical trials. The accuracy of leading AI systems for drug discovery also varies widely.

According to Rodriques, Finch sometimes makes errors. To address this, FutureHouse is enlisting bioinformaticians and computational biologists to evaluate and enhance Finch’s accuracy and reliability during its closed beta phase. Interested individuals can sign up here.