Imagine using artificial intelligence to compare two seemingly unrelated creations\u2014biological tissue and Beethoven\u2019s Ninth Symphony. At first glance, living systems and musical masterpieces seem unrelated. However, a novel AI approach developed by Markus J. Buehler, MIT\u2019s McAfee Professor of Engineering and professor of civil and environmental engineering and professor of mechanical engineering, bridges that gap, revealing common patterns of complexity and order.<\/p>\n\n\n\n
\u201cBy combining generative AI<\/strong>with graph-based computational tools, this approach reveals entirely new ideas, concepts, and designs that were previously unimaginable. We can accelerate scientific discovery by teaching generative AI to make novel predictions about ideas, concepts, and designs that have never been seen before,\u201d Buehler says.<\/p>\n\n\n\n The open-access study, recently published in Machine Learning: Science & Technology, demonstrates an advancedAI<\/strong>approach that integrates generative knowledge extraction, graph-based representations, and multimodal intelligent graph reasoning.<\/p>\n\n\n\n The study uses graphs developed using methods inspired by category theory as a core mechanism to teach models to understand symbolic relationships in science. Category theory, a branch of mathematics that studies abstract structures and the relationships between them, provides a framework for understanding and unifying disparate systems by focusing on objects and their interactions rather than their specific content. In category theory, systems are viewed as objects (which can be anything from numbers to more abstract entities like structures or processes) and morphisms (arrows or functions that define the relationships between these objects). By using this approach, Buehler was able to teach the AI<\/strong>\u200b\u200bmodel to systematically reason about complex scientific concepts and behaviors. The symbolic relationships introduced through morphisms made it clear that the AI \u200b\u200bwas not just making analogies, but was also engaging in deeper reasoning, mapping abstract structures to different domains.<\/p>\n\n\n\n Using this new approach, Buehler analyzed 1,000 scientific papers on biomaterials and transformed them into a knowledge graph in the form of a graph. The graph revealed connections between disparate information and was able to find relevant ideas and key points that tie many concepts together.<\/p>\n\n\n\n “What’s really interesting is that the graph follows the scale-free property and is highly connected, which can be effectively used for graph reasoning,” Buehler said. “In other words, we teach AI<\/strong>systems to think about graph-based data to help them build better representations of the world and enhance their ability to think about and explore new ideas to enable discovery.”<\/p>\n\n\n\n Researchers can use the framework to answer complex questions, discover gaps in current knowledge, propose new material designs, predict how materials will behave, and connect concepts that have never been connected.<\/p>\n\n\n\n The AI \u200b\u200bmodel found unexpected similarities between biomaterials and the Ninth Symphony, suggesting that both follow complex patterns. “Similar to the way cells in biomaterials interact in complex but organized ways to function, Beethoven’s Ninth Symphony arranges notes and themes to create a complex but coherent musical experience,” Buehler said.<\/p>\n\n\n\n In another experiment, a graph-based AI<\/strong> model suggested making a new biomaterial inspired by the abstract patterns in Wassily Kandinsky’s painting Composition VII. The AI \u200b\u200bsuggested making a new mycelium-based composite material. “This material combines a range of innovative concepts, including a balance of chaos and order, tunable properties, porosity, mechanical strength, and chemical functionality in a complex pattern,” Buehler noted. By drawing inspiration from abstract paintings, AI<\/strong> created a material that strikes a balance between strength and functionality while also being adaptable and able to perform different roles. The application could facilitate the development of innovative sustainable building materials, biodegradable plastic alternatives, wearable technology, and even biomedical devices.<\/p>\n\n\n\n