New AI Model PodGPT Blends Research and Podcasts for Smarter Health Answers
Kolachalama Lab Launches PodGPT to Bridge Science and Everyday Health Communication
Medical misinformation is a persistent challenge, and scientific communication often misses its mark with the general public. PodGPT offers a fresh approach by combining traditional written sources with the natural flow of spoken conversations. This AI model is trained not only on textbooks and journal articles but also on thousands of hours of science and medicine podcasts, where experts discuss topics from heart disease to public health.
Vijaya Kolachalama, PhD, Associate Professor of Medicine and Computer Science and a founding member of the Faculty of Computing & Data Sciences at Boston University, explains, “By integrating spoken content, we enhance the model’s grasp of conversational language. It learns how real experts talk about science, making its responses more relatable and easier to understand.”
Unlike many large language models (LLMs) that rely almost exclusively on written text, PodGPT taps into the richness of human speech. This boosts its ability to interpret questions and provide precise, context-aware answers across various STEMM fields—science, technology, engineering, mathematics, and medicine.
The development process involved compiling over 3,700 hours of publicly available podcasts. Using advanced transcription software, the team converted this audio into text to train the AI. PodGPT was tested on subjects ranging from biology and math to medicine, including multilingual content, which improved its versatility and accuracy.
This approach has practical implications for making complex scientific and medical information more accessible. PodGPT can help individuals better understand new diagnoses, research health concerns, and stay informed on public health matters without needing specialized training. It also holds promise for breaking down language barriers, delivering information in a way that feels natural and culturally relevant around the globe.
Kolachalama notes, “This opens avenues for using multimedia content like video or audio lectures to build smarter, more human-like AI tools. It also supports science communication across multiple languages, helping people everywhere access reliable information.”
Potential applications include improving information delivery on Alzheimer’s disease, mental health, cancer, and global health education. The core goal is straightforward: provide expert-backed answers that resonate with everyday users and support informed decision-making.
The study detailing PodGPT’s development and results is published in npj Biomedical Innovations.
Your membership also unlocks:
