Have you ever thought about the connections between seemingly unrelated fields, like biological tissue and Beethoven’s “Symphony No. 9”? It might seem implausible at first, but a groundbreaking AI technique developed by Markus J. Buehler, the McAfee Professor of Engineering at MIT, bridges this divide by exposing shared threads of complexity and order.
“By uniting generative AI with graph-based computational methodologies, our approach unlocks a realm of novel ideas, concepts, and designs that were once thought to be impossible. We can hasten scientific discovery by training AI to make innovative predictions about ideas and designs we’ve never encountered,” Buehler explains.
This open-access research, recently published in Machine Learning: Science and Technology, showcases a sophisticated AI framework that integrates generative knowledge extraction with graph-based representation and multimodal intelligent graph reasoning.
At the heart of this work is the use of graphs inspired by category theory, a mathematical discipline focused on abstract structures and their interrelationships. This unique perspective allows the AI model to comprehend symbolic relationships in science, aiming to understand and unify complex systems by emphasizing the interactions between objects rather than their specific details. The incorporation of morphisms — the functions defining relationships between these objects — fosters a deeper reasoning capability in the AI, facilitating the mapping of abstract structures across diverse domains.
Buehler applied this innovative method to analyze a dataset consisting of 1,000 scientific papers on biological materials, transforming them into a comprehensive knowledge graph. This graph illuminated the interconnections of various insights, revealing clusters of related concepts and critical ideas that link numerous themes together.
“What’s particularly fascinating is that the graph exhibits scale-free properties, boasts high connectivity, and is well-suited for graph reasoning,” remarks Buehler. “Essentially, we teach AI to interpret graph-based data, enhancing its capacity to develop more accurate world models and explore groundbreaking ideas for discovery.”
With this framework, researchers can tackle intricate queries, uncover gaps in existing knowledge, propose innovative material designs, and even predict material behaviors while linking previously unassociated concepts.
The AI model astonishingly identified unexpected parallels between biological materials and “Symphony No. 9,” indicating that both domains adhere to intricate patterns of complexity. “Just like cells in biological systems organize to perform specific functions, Beethoven’s 9th symphony carefully arranges musical notes to create a rich yet cohesive experience,” explains Buehler.
In a different experiment, the graph-based AI model suggested deriving a new biological material inspired by the abstract patterns in Wassily Kandinsky’s artwork “Composition VII.” The AI conceived a mycelium-based composite material that blends disorder and order, featuring adjustable properties, porosity, mechanical strength, and intricate chemical functionalities. “The resulting material exemplifies a harmony of strength and versatility, suitable for a range of applications,” Buehler notes. This approach could lead to the creation of sustainable building materials, biodegradable plastics, advanced wearable technology, and even cutting-edge biomedical devices.
This advanced AI model allows scientists to glean insights from music, art, and technology, revealing hidden patterns and opening doors to a world of innovative possibilities in material design, research, and even in the realms of music and visual art.
“Graph-based generative AI provides a much higher degree of novelty and exploration compared to traditional methods, establishing a versatile framework for innovation by uncovering hidden links,” concludes Buehler. “This research not only advances the understanding of bio-inspired materials but also paves the way for a future where interdisciplinary studies fueled by AI and knowledge graphs become essential tools for scientific and philosophical exploration.”
Photo credit & article inspired by: Massachusetts Institute of Technology
