In 2020, the National Science Foundation (NSF) selected UT Austin to lead the newly launched NSF AI Institute for Foundations of Machine Learning (IFML). One of the first seven institutes launched, IFML is part of the National Artificial Intelligence Research Institutes program, which today consists of 27 AI institutes connecting institutions and industry across the U.S. and around the world.
IFML uses new algorithms and architectures to ensure fairness in generative imaging, to enable robots to navigate in ever-changing environments, to deploy more robust MRI in clinical health settings, and to develop new biologics and therapeutics.
Director Adam Klivans, Professor of Computer Science at UT Austin, leads the institute in developing cutting-edge algorithms that power generative AI – research that is already having real-world impact. The institute has a prolific output with IFML researchers collaborating across an ecosystem that spans UT Austin, University of Washington, Wichita State, Microsoft Research. Stanford, Santa Fe Institute, UCLA, UC Berkeley, CALTECH, and Arizona State University.
Recent research advancements include:
- Biotech innovations from IFML’s Deep Proteins group are revolutionizing drug discovery and therapeutics with help from artificial intelligence.
- IFML researchers have developed a new framework for diffusion models that is already in several production pipelines by teams at Google.
- Generative AI and deep learning are improving the speed and image quality of MRI.
- Through collaborative efforts like ExpandAI, IFML is helping to harness the power and immense data needs of AI in ways that can be used in small, portable devices.
- To meet future demand for a highly skilled AI workforce, IFML members were instrumental in developing coursework for a new Master of Science in Artificial Intelligence (MSAI) degree program at UT Austin. The MSAI was featured in The New York Times due its potential to reshape the landscape of AI education.
IFML research in use-inspired areas—imaging, video, navigation, and protein design/engineering—are the applications that have been at the heart of some exciting recent breakthroughs. IFML senior personnel are closely connected to industrial partners in these areas, resulting in research that is aligned with the latest developments and has the potential for near-term deployment.
Our use-inspired work has become considerably more collaborative, as common foundational techniques (e.g., transformers) find applications across multiple modalities. As such, foundational research is now vital for achieving impact in diverse application areas. We use new algorithms, data sets, and architectures to ensure fairness in generative imaging, to enable robots to navigate in ever-changing environments, to deploy more robust MRI in clinical health settings, and to develop new biologics and therapeutics.
Leveraging our strong partnerships in education and broadening outreach efforts, IFML continues to address the demand for an increasingly AI-centric workforce. We strive to be a leader in AI education by providing a globally available, low-cost online Master of Science in AI.