Future House

Future House is a nonprofit research organization that develops artificial intelligence (AI)-based tools to accelerate scientific discovery. Established in 2024 with the support of former Google CEO Eric Schmidt, the organization aims to develop a functional “AI scientist” within a decade. Its work focuses on building AI systems that can automate research processes in biology and other complex scientific fields. The organization is headquartered in the United States.

Founding

Future House was co-founded by CEO Sam Rodriques and Chief Science Officer Andrew White. The organization is staffed by a multidisciplinary team of technical and operational experts. Members of the team include Sam Cox, Michael Skarlinski, Jon Laurent, Lauren Jaeger, Siddharth Narayanan, James Braza, Michaela Hinks, Ryan-Rhys Griffiths, Tyler Nadolski, Kiki Szostkiewicz, Geemi Wellawatte, Jasmine Dhaliwal, Mayk Caldas, Ludovico Mitchener, Ali Ghareeb, Albert Bou, and Remo Storni.

Mission and Research Layers

Future House’s main goal is to develop an AI scientist that can automate scientific research and accelerate the discovery process. This vision is structured across four layers:

1. The first layer includes humans and their research goals (e.g., full understanding of brain function or gene delivery technologies).
2. The second layer involves the AI scientist who generates hypotheses and updates models with experimental data.
3. The third layer comprises AI research assistants tailored for specific biological workflows.
4. The fourth layer includes predictive tools such as AlphaFold, laboratory experiments, and application programming interfaces (APIs).

Platform and AI Tools

In 2025, Future House launched its first platform and API, consisting of four AI tools: Crow, Falcon, Owl, and Phoenix. These tools assist researchers in literature review, data mining, literature comparison, and experiment planning.

• Crow is a general-purpose literature review and question-answering tool.
• Falcon specializes in comprehensive literature analysis and access to scientific databases.
• Owl focuses on identifying prior studies in specific domains.
• Phoenix is designed for experiment planning in chemistry and is based on the organization's ChemCrow system.

All tools offer access to high-quality open-access scientific literature and domain-specific utilities. They support multi-step reasoning processes, enabling in-depth evaluation of each source. The platform is accessible via both a web interface and an API.

Tool Performance and Criticism

Crow, Falcon, and Owl have outperformed existing frontier models in information retrieval accuracy and summarization quality, at times exceeding the performance of PhD-level researchers.

However, Future House has not yet achieved a scientific breakthrough or original discovery using these tools. The reliability of AI systems in addressing complex scientific problems remains a topic of concern, with risks stemming from their tendency to produce incorrect or fabricated information.

Benchmarks and PaperQA2

To evaluate the scientific proficiency of AI models, Future House released BixBench, a bioinformatics-focused benchmark dataset comprising 53 scenarios and 296 research questions. Claude 3.5 Sonnet and GPT-4o scored only 17% and 9% respectively, demonstrating that autonomous scientific discovery is still in its infancy.

Conversely, the organization's literature tool PaperQA2 achieved the highest accuracy in the RAG-QA Arena benchmark for scientific tasks. PaperQA2 uses multi-stage techniques such as query expansion, source re-ranking, and contextual summarization to answer scientific queries with high accuracy.

Future Outlook

Future House plans to expand the BixBench dataset, conduct performance comparisons against human researchers, and develop new models with improved scientific reasoning abilities. Its ultimate objective is to create an AI scientist capable of automating tasks such as hypothesis generation, experimental design, data analysis, and scientific writing. These systems are expected to significantly accelerate research processes in fields such as medicine, climate change, and emerging technologies.