Dirk Colbry

Building the Future: The AI Catalyst in Accelerating Interdisciplinary Scientific Discovery 

Senior Specialist, Department of Computational Mathematics, Science and Engineering 
colbrydi@msu.edu 

This enlightening presentation explores the transformative role of AI in modern science. Using the analogy of building a “house of science,” Dirk Colbry will share with the audience how computers are being used to accelerate science.

The audience will learn how a supercomputer is built and how accelerators (GPUS) can be used to calculate thousands upon thousands of mathematical operations per second (aka petaflops). This technological foundation is crucial for modern scientific inquiry.  

Although the technology is changing rapidly, modern AI tools struggle with the types of data analysis tasks that are common in observational science. For example, auto-generated results can be hard to interpret, are subject to (AI) hallucinations, and often require significant amounts of training data to achieve scientifically usable results. Colbry’s research focuses on accelerating scientific progress by developing tools and techniques that will significantly reduce the time required for manual data annotation, accelerating scientific discoveries and reducing the “mean time to science.” 

“One of the most significant challenges in using AI for scientific research is the generation of ‘training data,’ also known as ‘data annotation,’” Colbry said. “This process is often a bottleneck, slowing down scientific advancements due to its labor-intensive nature.” 

Colbry’s team has proposed a solution to this problem. They aim to develop an AI-driven user-friendly interface that simplifies data annotation, supporting workflows such as image segmentation, anchor point selection, classification, regression, and unstructured text analysis. By integrating a search algorithm, their software will seek automated solutions for these tasks, reducing the time and cognitive load on researchers. 

The team collaborates with academics across the university to apply their techniques in various fields, including education, smart agriculture, and animal phenomics. These interdisciplinary efforts demonstrate the broad applicability and potential impact of their work. 

“This presentation highlights the transformative power of AI in scientific research, showcasing how our tools can enhance efficiency and effectiveness across disciplines,” Colbry said. “By bridging the gap between manual and automated processes, we aim to accelerate scientific discoveries and foster a collaborative research environment, ultimately building a future where AI is a catalyst for groundbreaking advancements.”