Graph-based AI model maps the future of innovation

Imagine leveraging artificial intelligence to explore the connection between two seemingly unrelated realms — biological tissue and Beethoven’s *Symphony No. 9*. At first glance, it might seem that a living system and a timeless musical masterpiece have no common ground. However, a groundbreaking AI method developed by Markus J. Buehler, the McAfee Professor of Engineering at MIT, reveals an unexpected bridge between the two. This novel approach uncovers shared patterns of complexity and order, offering a new lens through which we can explore and connect disparate fields.

“By combining generative AI with graph-based computational tools, this method opens up possibilities for generating entirely new ideas, concepts, and designs that were once unimaginable,” says Buehler. “We can accelerate scientific discovery by teaching generative AI to make novel predictions about ideas and concepts that have never been conceived before.”

This open-access research, recently published in *Machine Learning: Science and Technology*, introduces an advanced AI method that integrates generative knowledge extraction, graph-based representation, and multimodal intelligent graph reasoning. The approach provides a framework for understanding and discovering complex systems across various disciplines, all through the lens of AI.

The core of this AI technique lies in graph representations inspired by category theory, a branch of mathematics focused on abstract structures and the relationships between them. Category theory offers a framework that transcends specific content, enabling the AI to identify unifying principles across a wide range of systems. Instead of focusing on the individual components of a system, category theory emphasizes the relationships between objects and how they interact. This method is particularly powerful for teaching AI to systematically reason over complex scientific concepts, allowing it to connect abstract ideas in innovative ways. The symbolic relationships defined by morphisms (functions that map between objects) enable the AI to move beyond simple analogy-making and engage in more profound reasoning, uncovering deeper patterns across domains.

Buehler’s team used this novel method to analyze 1,000 scientific papers on biological materials, transforming them into a comprehensive knowledge map. This map, represented as a graph, illustrates the connections between various ideas and concepts in the field. The AI’s ability to identify clusters of related information, as well as key points that tie disparate ideas together, marks a significant advancement in our understanding of complex scientific systems.

“What’s really striking about this approach is that the graph exhibits a scale-free nature, with a highly interconnected structure,” says Buehler. “This allows for effective graph reasoning, meaning the AI can think in terms of relationships and connections, helping it build better models of the world and explore new ideas to drive discovery.”

This graph-based AI framework holds the potential to answer challenging questions, uncover gaps in current knowledge, suggest novel material designs, predict material behavior, and connect concepts that have previously been unlinked.

One of the most surprising outcomes of the research was the AI model's ability to draw unexpected parallels between biological materials and *Symphony No. 9*. The AI recognized that both systems exhibit patterns of complexity, with each following a coherent structure that allows for functionality and expression. “Just as cells in biological materials interact in intricate but organized ways to perform specific functions, Beethoven’s *Symphony No. 9* arranges musical notes and themes in a manner that creates a complex yet harmonious musical experience,” explains Buehler.

In another intriguing experiment, the AI model recommended creating a new mycelium-based biological material, drawing inspiration from the abstract forms and patterns in Wassily Kandinsky’s painting *Composition VII*. The AI suggested a composite material that balances chaos and order, while also possessing adjustable properties, porosity, mechanical strength, and complex chemical functionality. "This material combines innovative concepts that bridge art, science, and engineering," notes Buehler. "It’s adaptable, strong, and capable of performing a wide range of functions."

By using the visual and conceptual language of abstract art as a design prompt, the AI was able to propose a material that could have wide-ranging applications, from sustainable building materials and biodegradable alternatives to plastics, to wearable technology and advanced biomedical devices. The creation of such a material would mark a significant leap in both material science and the integration of AI into the design process, demonstrating the power of artificial intelligence to generate unexpected yet profoundly useful innovations.

With this advanced AI model, scientists can draw insights from a diverse array of fields — including music, art, and technology — to analyze data and uncover hidden patterns that could unlock a world of innovative possibilities in material design, research, and even creative domains like music or visual art.

“Graph-based generative AI offers a level of novelty, exploratory capacity, and technical detail that far exceeds conventional methods, creating a versatile framework for innovation by exposing hidden connections,” explains Buehler. “This research not only advances the field of bio-inspired materials and mechanics, but also paves the way for a future where interdisciplinary research powered by AI and knowledge graphs becomes an essential tool for scientific and philosophical exploration.”

“Markus Buehler’s work on bioinspired materials has transformed vast amounts of data into knowledge graphs that map the interconnections between various topics and disciplines,” says Nicholas Kotov, the Irving Langmuir Distinguished Professor of Chemical Sciences and Engineering at the University of Michigan, who was not involved in the study. “These graphs serve as valuable information maps, helping us identify central themes, novel relationships, and potential research pathways by exploring complex connections across different areas of bioinspired and biomimetic materials. Tools like these are likely to become indispensable for current and future researchers.”

This research was made possible through support from MIT's Generative AI Initiative, a gift from Google, the MIT-IBM Watson AI Lab, MIT Quest, the U.S. Army Research Office, and the U.S. Department of Agriculture.