The rush to understand – and often, to fear – artificial intelligence often centers on what AI can do right now, fueled by headlines about chatbots and image generators. But beneath the surface of these readily available tools lies decades of foundational research, the painstaking work of computer scientists building the very algorithms that power these applications. On February 9th, UNC-Chapel Hill’s Department of Computer Science deliberately shifted the focus with its inaugural CoreAI Discovery Day, not to demonstrate what AI is, but to reveal who is building it and how. The event wasn’t about predicting a robotic future, but about showcasing the active, nuanced research happening today – research that already surpasses the capabilities of some commercially available AI systems.
The day unfolded as a deliberate counterpoint to the often-sensationalized AI narrative. Chancellor Lee Roberts opened the event emphasizing the university’s commitment to leadership in the field, noting the importance of interdisciplinary partnerships. This wasn’t simply a computer science showcase; researchers from chemistry, geography, philosophy, pharmacy, psychiatry, and radiology were present, signaling a university-wide recognition that AI’s potential extends far beyond its technical origins. Department Chair James Anderson framed the event’s core question: while everyone is talking about using AI, who is responsible for its underlying development? The answer, as presented throughout the day, is a surprisingly broad and deeply rooted community within UNC-Chapel Hill.
Original reporting: college.unc.edu.
The research presented wasn’t theoretical speculation. Computer science majors Fletcher Stuart and Ansh Aryan demonstrated virtual reality reconstruction, while other presentations detailed robots capable of autonomously navigating needles through living tissue – a feat that, according to Distinguished Professor Ron Alterovitz, already outperforms human physicians in certain bronchoscopy procedures. Perhaps most strikingly, researchers showcased AI systems capable of understanding video content with greater nuance than proprietary models like GPT-4o, a claim that challenges the perception of commercial entities holding a monopoly on cutting-edge AI. Assistant Professor Gedas Bertasius described a video understanding framework that could provide personalized coaching in skills ranging from basketball to piano, suggesting a future where expert-level instruction is democratized through AI.
However, it’s crucial to understand what the CoreAI Discovery Day actually revealed versus what headlines might imply. The demonstrations weren’t polished products ready for market; they were glimpses into ongoing research projects. The surgical robot, for example, isn’t replacing surgeons, but represents a significant step towards more precise and less invasive procedures. The video understanding framework isn’t a finished coaching system, but a powerful algorithm requiring further development and testing. The event highlighted potential, not immediate reality. This distinction is vital, as overstating the current capabilities of AI can lead to unrealistic expectations and misplaced anxieties.
The event also illuminated a critical tension: the rapid advancement of generative AI and the urgent need for tools to detect AI-generated fakes. Distinguished Professor Xiaoming Liu, recently joining UNC, presented his work on distinguishing genuine media from AI-generated content, a challenge that has become exponentially more pressing with the widespread availability of tools like DALL-E and Midjourney. This research isn’t simply about identifying “deepfakes”; it’s about preserving trust in information and safeguarding against manipulation in a world increasingly saturated with synthetic media. The fact that this work is being prioritized alongside the development of generative AI itself speaks to a responsible and forward-thinking approach within the department.
Limitations to consider include the inherent challenges of translating research findings into real-world applications. As Vice Dean Jay Aikat pointed out during the closing panel, moving from a foundational algorithm to a practical tool requires navigating complex ethical considerations and ensuring equitable access. The success of these projects hinges not only on technical innovation but also on thoughtful implementation and a commitment to addressing potential biases and unintended consequences. Furthermore, the research presented largely reflects the priorities and expertise within UNC-Chapel Hill’s Department of Computer Science; a broader perspective incorporating diverse viewpoints is essential for responsible AI development.
Looking ahead, the department plans to build on the momentum generated by CoreAI through ongoing partnerships and events, including a community open house on March 7, 2026. The focus will be on fostering collaboration across disciplines and translating foundational AI research into tangible solutions for a wider range of challenges. But the most important question emerging from CoreAI isn’t about what new algorithms will be developed, but rather how we will ensure that these advancements are used responsibly and ethically – and whether the interdisciplinary conversations sparked at UNC-Chapel Hill can serve as a model for responsible AI innovation nationwide. Will other institutions prioritize showcasing the building of AI alongside its application, and will this shift in focus lead to a more informed and equitable future powered by artificial intelligence?
