University of Iowa AI Lightning Talks March Session, Theme “AI in Research: Inside Faculty Workflows” featured staff and faculty who are implementing artificial intelligence into their workflows. During the March 12th talk, staff members showed attendees how they are leveraging AI in their research areas.
The UI began hosting semi-monthly talks starting in October 2024. We provide a platform for leaders. across campus Share how you are currently using and implementing AI and other projects, concepts, influences, and interests.
The talk will feature four to five speakers speaking on AI-related topics, followed by a Q&A session with the audience.
Sugand AlaviAn assistant professor in the Tippee College of Business, she began her presentation with a discussion on “Data Analysis with AI for Non-Coders.”
Alavi’s research allows students to more easily understand complex datasets. Her project allows students to use AI tools to analyze and manipulate datasets before learning Python coding. Python has traditionally been a computer language that requires students to have some level of fluency in order to work with dense data, often requiring students to practice for months before being able to apply their knowledge. Using AI in this example reduces the steep learning curve of Python.
During his presentation, Alavi demonstrated how students can use AI to access information from a set. Without AI, a deep understanding of coding was required.
In a traditional environment, students could “take a month to get to this point,” Alavi said, demonstrating what AI can deliver in seconds.
An audience member asked how students can deal with the possibility of receiving incorrect information from AI tools.
Alavi said this AI tool is not always suitable for certain tasks, and the need for human validation of results is always important.
related: UI team develops AI-powered camera to detect violence in real time
Aislin Conradan associate professor of social work at UI, is working on a project with. Karim Abdel MalekUI professor of mechanical engineering.
Their project is a device that uses AI-powered cameras to detect incidents of abuse, such as domestic violence, and can alert local authorities.
“This device is a nanny camera on steroids,” Conrad said.
When the camera detects a disturbance, it will turn on and start recording. The footage uses an AI model trained on skeletal movement patterns to determine if abuse is occurring.
They may then initiate actions, such as calling local authorities, according to the device owner’s program.
Conrad envisions it being deployed in places where vulnerable populations congregate, such as nursing homes and daycares.
Conrad and Marek are working on a physical device that is still in the early stages of manufacturing. They have completed a proof of concept and are working on developing a prototype. Next steps include testing and working with third parties to validate the system.
“Human review is still important. There are ethical considerations to be made with devices like this. It’s very important to think about safeguards,” Conrad said.
She said the project is “an opportunity for leaders to lead the way in ethical human protection.”
“Through this research and commercialization, we hope to bring violence to light,” Conrad said.
Tori ForbesUI professor in the Department of Chemistry and director of the MATFab facility, presented “AI-enhanced Experiment Optimization: An Accessible Workflow for Chemists.” Recently accepted into the Journal of Chemistry-Methods.
Forbes said chemistry has less accessible data than other fields. Journals publish only the data that yield results and leave other data unpublished. Their field, radiochemistry, is particularly affected by this lack of data.
As a result, the problem arises that it is difficult to train an AI model without a dataset.
“We decided we needed to start thinking about how experimenters could add data to the field so that we could start developing more of these AI tools,” Forbes said.
To do so, Forbes and her colleagues need to consider how chemists work with data and how they can produce end products that can be more easily understood, shared, and analyzed.
Forbes’ process begins by taking a photo of the physical data, scanning it into Google as a PDF, and then using readily available AI services to convert the data into an accessible digital format.
They aimed to use existing AI applications to enable this process to be carried out in a variety of institutions, including smaller institutions that do not have access to advanced technology.
These existing AI applications are often inaccurate when providing insights into data and fail to account for the safety of experiments, so it’s important for humans to “stay in the loop” to review data transformations, Forbes said.
When the students applied Forbes’ procedures, they observed that the undergraduate students were able to gain a deeper understanding of their research, and that for the graduate students, it was “very helpful in starting to organize their ideas.”
Kenneth Knepplea clinical professor of urology and deputy chief health information officer at UI Health Care. It’s a platform called an “AI-powered clinical data intelligence system” for reviewing patient medical records.
Patients often come to a new facility after receiving care elsewhere and come with pages of patient data that new providers must review before providing quality care.
Nepple’s work “pioneered the ability to do custom prompts and custom summaries within charts.”
Nepple said the AI tool could be particularly applicable to older patients and those with complex conditions, who tend to have the most complex medical records.
