Using AI to Solve Real Problems

It's been about two years since Carolyn Tang Kmet began thinking about how to create a tool that allowed users to train and educate an AI agent using their own trusted sources. The approach would limit the agent's knowledge base, but it also would reduce noise, provide greater transparency, and help users have greater control over the responses.  

"That's what sets it apart from current offerings like ChatGPT and CoPilot," said Tang Kmet, an associate professor who teaches data storytelling at the Medill School of Journalism, Media, Integrated Marketing Communications. "Imagine being able to query your own knowledge base with questions like,'“What was that recipe I looked at last week that had zucchini in it?' or 'In which lecture did I cover R2-D2 and Indiana Jones?' 

Tang Kmet turned to students in Northwestern Engineering's Master of Science in Artificial Intelligence (MSAI) program to help make her vision a reality. Tang Kmet partnered with the program on a practicum project, where students work to solve an open-ended AI problem.  

Three students collaborated with Tang Kmet for her project. The outcome was ReMindAI, a program that allows instructors to upload course materials such as slides, notes, and lecture recordings, while students can add their own files. Using natural-language queries, students can prompt the system to deliver contextual, cited responses to their questions in real time.  

The structure of the practicum is what appealed to Tang Kmet. It's also what she thinks helps best prepare students for success after graduation.  

"The opportunity to solve real-world challenges offers a kind of context and fulfillment that simply can’t be found in textbooks, research papers, or exams," she said. "It also builds a powerful sense of self-efficacy. When students see firsthand what they’re actually capable of, they begin to trust in their own intellect and abilities." 

The fact that students are navigating real problems is a differentiator in Tang Kmet's eyes. Prior to teaching at Medill, she spent her career in marketing, where it is critical to understand and anticipate customer motivation and intent.  

"It’s not just about knowing what the customer wants in this immediate moment, but also about anticipating their needs," she said. "Only when a marketer fully understands a customer’s motivations can they begin to influence what that customer wants." 

What Tang Kmet has seen throughout her career is that many times, marketers, engineers, and other innovators often create solutions in search of a problem, rather than solutions that address existing needs. In class, she often cites a study from CB Insights that analyzed 101 failed startups and found the top reason for failure was tackling interesting problems instead of ones that address a market need. 

"Time and time again, startups fail not because the tech isn’t impressive, but because there’s no market for them," Tang Kmet said. "That’s what’s so wonderful about the MSAI practicum. The entire project is grounded in a defined, real-world need."  

The practicum, which is offered during the program’s third quarter, provides students with practical experience and helps to emphasize the importance of learning by doing. The project is subject to real-world constraints. 

Tang Kmet was particularly impressed with the students' ability to balance innovation with business considerations like cost and distribution.  

"They clearly understood that AI isn’t just about building cool models," she said. "It’s a technology that can be applied to solve real problems. They approached this project with that mindset. They thought about the end-user, they understood their needs, and then they considered how AI could be integrated to solve those needs."