CAMBRIDGE, MASSACHUSETTS — The MIT-IBM Watson AI Lab supported MIT faculty as they built research teams and advanced theory and applications in AI and engineering. The lab brought together researchers in the AI realm to accelerate research and blend work across disciplines.

Jacob Andreas, an associate professor in the Department of Electrical Engineering and Computer Science at MIT, a member of CSAIL, and a researcher with the MIT-IBM Watson AI Lab, launched his first major project through the lab shortly after joining MIT. The project worked on language representation and structured data augmentation methods for low-resource languages. Andreas studies natural language processing.

In describing that early period, Andreas said, "The MIT-IBM Watson AI Lab has been hugely important for my success, especially when I was starting out." He also said, "It really was the thing that let me launch my lab and start recruiting students."

Andreas’s initial project through the lab provided the compute resources required to navigate the transition in natural language processing toward larger language models. "I feel like the kind of the work that we did under that [first] project, and in collaboration with all of our people on the IBM side, was pretty helpful in figuring out just how to navigate that transition," he said. Andreas’s group pursued multi-year projects on pre-training, reinforcement learning, and calibration for trustworthy responses using the computing resources and expertise within the MIT-IBM Watson AI community.

Yoon Kim, an associate professor in the Department of Electrical Engineering and Computer Science at MIT, a member of CSAIL, and a researcher with the MIT-IBM Watson AI Lab, develops methods to improve large language model capabilities and efficiency. Before joining MIT, Kim held an MIT-IBM postdoctoral position where he pursued neuro-symbolic model development.

Speaking about the lab’s role in his work, Kim said, "Having both intellectual support and also being able to leverage some of the computational resources that are within MIT-IBM, that's been completely transformative and incredibly important for my research program." He added, "This is an impetus for new ideas, and that's, I think, what's unique about this relationship."

Other MIT faculty working with the MIT-IBM Watson AI Lab include Justin Solomon, an associate professor in the Department of Electrical Engineering and Computer Science at MIT, a member of CSAIL, and a researcher with the lab. "Crucial … from its beginning until now," Solomon said. "I think these are all really exciting spaces," he said.

Solomon’s research team focuses on theoretically oriented, geometric problems in computer graphics, vision, and machine learning, and lab work with IBM allowed his group to fuse distinct AI models trained on different datasets for separate tasks.

Chuchu Fan, an associate professor of aeronautics and astronautics at MIT, a member of the Laboratory for Information and Decision Systems, and a researcher with the lab, combined formal methods with natural language processing and developed both autoregressive task and motion planning for robots and large language model–based agents for travel planning, decision-making, and verification. "I think these early-career projects [with the MIT-IBM Watson AI Lab] largely shaped my own research agenda," Fan said.

"That work was the first exploration of using an LLM to translate any free-form natural language into some specification that robot can understand, can execute. That's something that I'm very proud of, and very difficult at the time," she said. "would be impossible without the IBM support," she said. Fan’s team also improved large language model reasoning through collaboration with IBM support.

Faez Ahmed, an associate professor of mechanical engineering at MIT and a researcher with the lab, collaborated through the lab to develop machine learning methods to accelerate discovery and design within complex mechanical systems. Ahmed’s Linkages project employs generative optimization to solve engineering problems in a way that is both data-driven and precise, and his team is applying multi-modal data and large language models to computer-aided design.