Got something to say to this podcast? Send a voicemail, and you might hear yourself on the next episode!

Episode 1: What Is Health AI? From Research to Real Patient Care

What does artificial intelligence actually look like in healthcare? Sometimes, it’s a quiet alert that helps a physician choose a medication. Other times, it’s a tool that reconstructs an MRI image or helps a clinician make sense of a complex patient record.

In the first episode of A Cup of Chai, host James Mitchell joins CU Anschutz biomedical informatics researchers Nick Dwork and Yanjun Gao to explore health AI through three perspectives: medical imaging, language, and human–AI interaction.

Together, we discuss what these technologies can do, where they still fall short, and why improving patient care requires careful attention to the people using them.

In this episode

• Understanding the basics: How data science, machine learning, and large language models relate to health AI.
• Faster, clearer MRI: How AI can reconstruct images from less data—and why fetal motion presents a particularly difficult challenge.
• Making clinical records useful: How AI could summarize complex patient histories, highlight urgent concerns, and surface supporting evidence.
• Personalizing AI communication: How ClinPrompts aims to tailor AI responses to clinicians’ preferences and clinical settings.
• Designing for human oversight: Why automation complacency, cognitive load, and interface design matter for safe clinical decisions.
• Bringing AI into the workflow: What genomic prescribing alerts illustrate about delivering useful information at the moment of care.
• Looking ahead: Open-weight models, ambient documentation, wearable devices, telehealth, and the need for evaluation and safety guidance.

Featured voices

James Mitchell - Researcher studying human–computer interaction, human–AI interaction, and the design of clinical decision support.

Nick Dwork - Faculty member in the Department of Biomedical Informatics at CU Anschutz, researching MRI reconstruction and accelerated imaging.

Yanjun Gao - Faculty member in the Department of Biomedical Informatics at CU Anschutz and director of the LARK Lab, focused on language, reasoning, and knowledge.

Key takeaway

Effective health AI takes more than an accurate model. It requires reliable evidence, thoughtful design, and integration into clinical workflows so that clinicians can use its outputs well.

__

Information

The views and opinions expressed on Cup of CHAI are those of the hosts and guests and do not represent the official positions of the University of Colorado Anschutz, the Center for Health AI, or any affiliated organizations. References to institutions, products, or services do not imply endorsement. This podcast is for educational and informational purposes only and does not constitute medical advice.