AI4OPT In Its Third year
The Artificial Intelligence Institute for Advances in Optimization is delivering a paradigm shift in automated decision-making
September marked the third anniversary of The AI Institute for Advances in Optimization (AI4OPT). Established in 2021 with a $20 million grant from the National Science Foundation, this visionary institute is a collaborative effort, bringing together UC Berkeley, Georgia Institute of Technology, University of Southern California, UC San Diego, UT Arlington, and Clark Atlanta University. Berkeley IEOR Chair and Professor Alper Atamturk and Berkeley IEOR Professor Dorit Hochbaum both serve on the AI4OPT leadership team, and Berkeley IEOR Assistant Professor Paul Grigas is a member of numerous AI4OPT research thrusts.
The institute’s mission is to transform automated decision-making on a massive scale by integrating Artificial Intelligence (AI) and Mathematical Optimization (MO). This fusion promises breakthroughs that would be unattainable by either field in isolation. According to the Director of AI4OPT, Georgia Tech’s Pascal Van Hentenryck, the very heart of the Institute’s ethos lies in its unwavering commitment to address fundamental societal and technological challenges, such as:
- How to design agile, sustainable, resilient, and equitable supply chains?
- How to operate energy systems powered by distributed renewable energy resources?
- How to deliver a step change in chip design and manufacturing, and manufacturing as a whole?
- How to create sustainable ecosystems within the food-waterenergy nexus?
Over the past year, AI4OPT has ramped up its research efforts, harnessing the potential of AI for Engineering to confront challenges entailing the evolving, complex landscapes of reliability, robustness, and scalability. Recent projects are driven by high-stakes applications with physical, engineering, and business constraints, all striving to strike a balance between efficiency, resilience, sustainability, and equity.
Because the underlying optimization problems at the core of these grand challenges surpass the capacities of current technologies, AI4OPT is organized around methodology thrusts that focus on specific challenges. They include a new generation of data-driven optimization solvers, decision-making under uncertainty,
combinatorial and reinforcement learning, end-to-end optimization, and decentralized learning and optimization. AI4OPT also remains committed to embedding ethical AI considerations from the inception of each research endeavor, rather than as an afterthought.
The research mission of AI4OPT is complemented by its educational vision, which
seeks to establish enduring pathways for AI in engineering, spanning from high school to graduate education. This vision is actualized through a “teach the teachers” approach. As part of this effort, in July, Berkeley IEOR hosted the transformative Kids Teach Tech (KTT) STEM summer camp, providing underserved youth with a two-week immersion into the world of science, technology, engineering, and mathematics. Following this, Berkeley IEOR organized “Journey Through the World of Analytics,” a two-day professional development workshop, that equipped community college faculty with essential course materials in advanced data-driven decision analytics.
As AI4OPT moves into its third year, Berkeley faculty and graduate students push the frontiers in collaborative and transformative research and inclusive education that maximizes the impact emerging from AI and optimization.
This article originally appears in the 2023 edition of Berkeley IEOR Magazine