Earlier this year, Professor Jason Corso and Professor Enrique Dunn in the Department of Computer Science were awarded a DARPA grant of nearly $5.9 million dollars to research and develop the first Multi-Directional Loosely-Linked Archetype Models for Perceptually-Enabled Task Guidance program—or in short, MILLY. This project is a collaboration between multiple institutions, with co-investigators in the University of Rochester, University of Michigan, and Purdue University. The program’s goal is to help any individual easily complete a task in a new environment by having the program “continuously learn through a variety of sources…and practice ‘active learning’ in order to adapt on the fly.”
Among the 15 research centers and labs across multiple disciplines on campus, the MILLY team received the grant through Stevens Institute for Artificial Intelligence (SIAI). The Stute had the chance to interview Professor Dunn to learn more about MILLY, and what he is most excited about in this project to broaden the horizons of artificial intelligence (AI).
In Dunn’s words, MILLY will be a program that is a “tutor, assistant, and guide.” MILLY takes in those wide ranges of data inputs or resources to automatically learn how to internally present a task. The program will come up with a sequence of instructions for the individual to complete a task and will use machine learning to learn the sequence. Dunn further emphasized that there will be two important ways that MILLY will be different from other alternative programs available today. First, through its learning capabilities, it will be fully automated modeling of tasks. Second, the automated analysis that interprets the users’ and the sequences’ actions based on observations. Current alternative designs and programs require the continuous addition of new data points and a programmer in the back finding the inefficiencies in the program — MILLY takes care of that. Ultimately, MILLY will be a program that can remove the programmer from the automation task, and make task automation and task learning more efficient.
Working with different co-investigators across institutions across the U.S. reinforces the diverse backgrounds and capabilities that can really set MILLY apart from others. The DARPA award intends to support the multiple Ph.D. students across various institutions and provides the resources necessary to advance MILLY’s developments. More importantly, Dunn is excited about how this project can make AI and augmented reality (AR) opportunities more practical in the future. As MILLY is developed more over the years, and people become more appreciative of AI and AR opportunities in the future, he is hopeful that AI can help carry out many more daily tasks in a very practical manner. While it is difficult to determine the exact users of MILLY early in its development, Dunn intends for MILLY’s predominant use to be education. He hopes for MILLY to help educate individuals in tasks across various industries—whether it be working on fixing and building something manufacturing or performing surgery in hospitals—hands-on, eliminating the need for aids and manuals.
That’s where Dunn’s knowledge and background in computer vision and 3D imaging come into play. As mentioned before, the beauty of the MILLY project is the diversity in the investigators’ skills that help MILLY come to life. He intends on implementing his past and present expertise in providing MILLY the spatial context regarding where and how certain tasks are to be performed and in what sequence.
Dunn emphasizes that this is a long-term project, intended to be worked on for the next four years, and is excited for people to know that it will actually affect lives in a positive way. He looks forward to working with professionals and the various researchers, learning from them, and expanding the opportunities available in AI research.
He encourages those interested in learning more about how they can be involved in AI/AR research projects, to contact faculty, and actively look at websites to see what research projects are available for students. He states that the Stevens Computer Science Department has around 20 faculty members who are currently involved in various research projects in various disciplines in computer science, who students may be interested in working with.
The Stute asked Professor Dunn what he hopes for the broader community to know and appreciate about AI going forward. He said, “AI can be used for good, and MILLY is one of many examples.”
Be First to Comment