In the age of ChatGPT, Professor Ming Shao is working to make AI smarter and safer.
At UMass Dartmouth, our faculty are at the forefront of the technology shaping our future.
Meet Ming Shao, associate professor of computer and information science. Dr. Shao's research is focused on machine learning and artificial intelligence (AI). In collaboration with other researchers, he explores how AI can be used to advance all kinds of endeavors, from fisheries science to medicine.
Tell us about your research. What is your primary focus? Why is this topic worth investigating?
"My principal area of focus is computer vision, machine learning, and AI in general. Deep learning and large language models have significantly advanced the ways we can leverage large amounts of data for real-world applications. We have witnessed the popularity and success of ChatGPT, Stable Diffusion, and many other AI-driven tools. However, these AI algorithms and products are also vulnerable to adversarial attacks.
"An 'adversarial attack' refers to different techniques used to 'fool' a machine learning model with deceptive or misleading data that will cause the model to malfunction. These kinds of attacks can lead to catastrophic consequences, especially in human-centered AI. There is a strong demand for fair, safe AI."
Science is collaborative
Dr. Shao works with researchers in other disciplines, exploring how machine learning and computer vision can be applied in different fields, including fisheries management, neuroscience, and healthcare.
"I have been actively collaborating with Professor Pingguo He from UMass Dartmouth's School for Marine Science and Technology (SMAST) to promote AI and computer vision in an innovative electronic monitoring program for New England groundfish based on wireless video transfer, edge-based AI, and an intelligent discard chute.
"I am also working with Dr. Yalda Shahriari from the University of Rhode Island to extend my multi-view graph representation learning research. This expansion involves combining and analyzing various types of brain signals for patients diagnosed with amyotrophic lateral sclerosis (ALS).
"Additionally, I have been working with the faculty of Harvard Medical School to integrate our expertise in biofabrication, tissue engineering, microfluidics, non-invasive bioanalysis, and machine learning to develop a generalized, self-dose-optimized 'multi-sensor-integrated multi-organ-on-a-chip' platform."
What kind of change do you hope to bring about in the world – either with your research or through teaching?
"My research goal is to understand the flaws and weaknesses of existing AI models and develop secure AI for human-centered computing and data analytics tasks. I hope to develop AI that can enable trustworthy computing in changing environments. This will help ensure more reliable results when we trust computers to perform complex data analysis in areas such as social science, health informatics, chemistry, and biology. These research outcomes and discoveries will also be included in my teaching and mentoring activities to engage students in the recent advances of AI and data science."
"I'm currently involved in a project that aims to create a better way for computers to learn from different sources of information. This will help the computers automatically spot attempts to trick or confuse them and adjust how they learn when there's extra noise in the data. The goal is to make computers more resistant to harmful tricks and create ways to defend against these attacks. All of this is being done to make sure that computers can keep learning and improving their skills over time, while also being understandable, fair, and safe.
"In today's world with so much data, this learning process is getting more complicated. The data comes from various places like cameras that look at things from different angles or data that's presented in different ways like audio, visual, and text. This makes it difficult for current models to produce consistent results. Also, data comes in huge amounts from sources like the internet, and there's sensitive information like personal photos and health records that need special handling. So, my project is about building a strong and smart learning system that can handle all these challenges and keep learning even as new data keeps coming in."
What do you most enjoy about working with UMass Dartmouth students?
"I have been working with UMass Dartmouth students for seven years in undergraduate and graduate courses and research projects. Their creativity and talents are evident in their exciting research projects and remarkable coursework.
"I oversaw the National Science Foundation Research Experiences for Undergraduates (NSF-REU) program for the past three years at UMass Dartmouth and found that our undergraduate students actively seek research opportunities as early as their sophomore years. They demonstrate their strong research capabilities and many of them publish papers. Another impressive finding is that many of our graduate students have industry experience and thus can quickly adapt to research projects and their theses. Some of my former advisees are currently working at Amazon as data/applied scientists or are getting their PhDs. I have supervised six PhD students, and most of them have continued in academia."