Pilot Programs to Enable Human-led Generative Knowledge Discovery
CAMBRIDGE, Mass.–(BUSINESS WIRE)–Unreasonable Labs, the company building superintelligence for knowledge discovery, today announced it has launched from stealth and closed a $13.5 million funding round led by Playground Global, with participation from AIX Ventures, E14 Fund, and MS&AD Ventures. The company, founded by a former senior research scientist at Google DeepMind and an engineering professor at MIT, has built a foundational AI discovery engine designed to accelerate breakthroughs by composing new knowledge across chemistry, materials science, biology, and beyond.
The funding will be used to scale Unreasonable’s core technology, which pairs state of the art large language models (LLM) with a map of relationships and neurosymbolic mathematical abstractions that allow real-world patterns across disparate fields to enable generative discovery. Unreasonable’s unique pairing of capabilities creates a cross-disciplinary AI engine that moves beyond simple data retrieval to generate genuinely novel, validated scientific hypotheses, simulate solutions, and design experiments.
“We are at a turning point where AI can be both an assistant to the scientist and a catalyst for the science itself, but LLMs alone cannot solve for scientific discovery,” said Yuan Cao, PhD, co-founder and CEO of Unreasonable. “To go from idea to impact, you need to move unreasonably fast and we exist to make that possible. Unreasonable will enable R&D teams to solve in weeks what previously took years.”
“Current AI models can only retrieve what is already known, which prevents even the most impressive reasoning models from generating novel discoveries,” said Prof. Markus Buehler, co-founder and CTO of Unreasonable. “Genuine discovery requires a deeper understanding, to connect disparate ideas to ultimately synthesize new insights. By pairing models with neurosymbolic mathematical abstractions we create a machine that does not just parrot the world as it is, but actively shapes the world. At Unreasonable, we’ve built that machine to invent a better future.”
Moving From Generative Text to Generative Discovery
Modern R&D teams face an unprecedented information bottleneck. Scientific and technical data are compounding exponentially, yet critical insights remain trapped in isolated silos—scattered across millions of research papers, lab notes, and proprietary simulations.
Current AI models, while impressive, have largely used sophisticated search tools for knowledge aggregation and summarization for tasks like scientific discovery. They can interpolate from existing knowledge, but – unlike the way humans create knowledge – they struggle to connect distant concepts or navigate the high-dimensional complexity required for true invention. Unreasonable was founded to bridge this gap, moving from “generative text” to “generative discovery.”
The Unreasonable Approach: Composing New Knowledge
Unreasonable’s discovery engine is built to overcome the limitations of closed AI models, whose representations of the world are constrained by finite training data. While most AI for science tools are narrowly trained on specific datasets—such as protein folding or battery chemistry—Unreasonable has built a general framework that incentivizes models to reason beyond their intrinsic world representations.
Unreasonable’s framework pairs state of the art LLM models with a map of relationships and neurosymbolic mathematical abstractions that allow real-world patterns across disparate fields to enable generative discovery.
These mathematical abstractions transform unstructured data into a verifiable, structured network of entities and their relationships. The set of rules that organize that vast data in turn enables scientists to edit the reasoning process. Together, this framework allows machines to comprehend and deduce meaning from data in a logical and systematic manner, enabling a more precise, causal reasoning than a standard LLM while simultaneously providing the necessary rigor and explainability for genuine scientific breakthroughs. In doing so, Unreasonable’s framework enables AI to reason creatively, a uniquely human characteristic that is a prerequisite to genuine knowledge discovery.
To enable generative discovery, Unreasonable has built the essential operating system for scientific research and development, to provide a critical orchestration layer that stands alone in the field. Specifically, the system is designed to:
- Ideate Across Disciplines: Synthesize vast quantities of unstructured data from disparate scientific fields into a unified representation of world knowledge.
- Generate Novel Hypotheses: Connect seemingly distant but analogical concepts (e.g., applying a structural principle from aerospace engineering to a challenge in synthetic biology) to suggest entirely new experimental paths.
- Validate via Simulation: Integrate directly with high-fidelity physics-based simulations to test and refine hypotheses before they ever reach a wet lab or a production line.
- Interact with Physical Hardware: Translate discovery into executable protocols and ingest experimental data, creating a feedback cycle from hypothesis to physical prototype.
A Founding Team at the Intersection of AI and Physical Science
Unreasonable unites two worlds that have traditionally remained separate: elite artificial intelligence research and deep physical engineering expertise. The company’s founders represent the pinnacle of these fields.
- Yuan Cao, Co-founder and CEO, was previously a Senior Staff Research Scientist at Google DeepMind where he drove AI research and development across multiple areas and made core contributions to the Gemini model family and Search AI Mode.
- Prof. Markus Buehler, Co-Founder and CTO, an Engineering Professor at MIT, has spent decades pioneering AI-driven discovery and computational methods for materials and complex physical systems.
World-Class Advisors
In addition to the co-founders and initial team, Unreasonable is supported by the following list of advisors, all of whom are highly regarded leaders in their fields:
- Kostya Novoselov, Nobel Prize in Physics in 2010
- Robert Langer, Institute Professor at MIT, biotech innovator with 40+ companies and 1,000+ patents
- Thomas Wolf, Co-founder and Chief Science Officer, Hugging Face
“At Playground, we look for the unreasonable founders who tackle seemingly impossible problems,” said Sasha Ostojic, Venture Partner at Playground Global. “The intersection of AI, reasoning, and knowledge discovery is the next meaningful frontier for human productivity. The Unreasonable team has the rare combination of technical pedigree and ambitious vision required to redefine how we discover new materials, medicines, and energy solutions. To paraphrase George Bernard Shaw, all progress depends on the unreasonable.”
Looking Ahead: From Pilots to Breakthroughs
With this capital, Unreasonable will expand its technical team, hiring top-tier talent across machine learning, simulation engineering, and computational science. The company has already initiated pilot collaborations with leading industrial partners in the energy transition, materials science, and pharmaceutical sectors.
By productizing scientific intuition, Unreasonable aims to create a world where new knowledge and solutions are discovered through human-led, AI-driven exploration of the possible.
About Unreasonable
Unreasonable is an AI company building superintelligence for knowledge discovery. Unreasonable’s operating system is designed to accelerate R&D and human ingenuity for the physical world across disciplines. Founded by world-leading experts from Google DeepMind and MIT, Unreasonable has offices in Palo Alto, California and Cambridge, Massachusetts.
About Playground Global
Playground Global is a deep tech venture capital firm with $1.2 billion under management, backing early-stage startups solving foundational challenges in next-generation compute, automation, energy transition, and engineered biology. Founded in 2015 and based in Palo Alto, Playground partners closely with technical and scientific founders to turn breakthrough ideas into lasting companies. The firm’s portfolio includes PsiQuantum, MosaicML (acquired by Databricks), d-Matrix, Agility Robotics, Ideon, Ultima Genomics, Strand Therapeutics, and xLight. Learn more at www.playground.vc.
Contacts
Media Contact: [email protected]
