FutureHouse, a nonprofit backed by former Google CEO Eric Schmidt, has released its first set of AI tools.
These tools have been designed to act as AI scientists and support researchers in scientific discovery. This new development is a step towards the company’s mission of building an AI “scientist” within the next ten years.
The platform features four tools (Crow, Falcon, Owl, and Phoenix).
And each one serves a specific function. Together, they aim to accelerate key steps in the research process.
How the Tools Work
| Tool | Function |
| Crow | Searches scientific papers and answers related questions. |
| Falcon | Performs deeper searches across specialized databases. |
| Owl | Identifies past research on a chosen topic. |
| Phoenix | Assists in planning chemistry experiments. |
They operate on a shared system. They draw from a large set of open-access scientific literature and also use step-by-step reasoning to check sources before offering results.
According to FutureHouse, researchers can even combine these tools. In theory, this should help them work faster and more effectively.
The Importance
AI in science is not new. Several startups have developed similar tools, often with support from venture capital. Tech leaders are also entering this space.
Earlier this year, Google introduced “AI co-scientist,” a system meant to help form hypotheses and create experiment plans.
Leaders at OpenAI and Anthropic have said AI could significantly speed up scientific progress. In medicine especially, they see high potential.
Yet many researchers remain skeptical. They often find current AI tools unreliable. Some struggle with complex reasoning, while others produce false or misleading information, known as hallucinations.
Can These Tools Make a Real Impact?
Right now, there is no public record of any scientific discovery made using FutureHouse’s tools. The team acknowledges that the tools, especially Phoenix, may still produce errors.
To address this, FutureHouse has released the tools under a “rapid iteration” approach. This means users can begin testing them and provide feedback. The team hopes this early launch will improve future versions.
Also read: AI in Science: Unearthing Rare Earth Elements
Caution from the Scientific Community
Past efforts to combine AI and science have had mixed results. In 2023, Google announced that its system, GNoME, helped synthesize 40 new materials. But later reviews found that none were truly new.
This is a common problem: AI tools often look promising but fall short in practice.
