Amber Campbell, an undergraduate Computer Science and Engineering student, reflects on her experience attending the 2024 e-HAIL Symposium alongside her peers and mentors at CHEPS.
On September 13, I had a chance to attend the e-HAIL (E-Health and Artificial Intelligence) Symposium, which provided me the opportunity to learn more about generative AI in healthcare through a variety of engaging presentations. This included Dr. Susan A. Murphy’s keynote speech on reinforcement learning, interacting with researchers presenting posters on topics such as Tik-Tok as a search engine for medical advice and eye-tracking software as a tool for analysis of physician-AI interactions, as well as attending a panel discussion with Michigan-affiliated panelists from a variety of disciplines such as Michigan Medicine, CSE, and Health Sciences.
I think one of the most impactful takeaways from this symposium was the need for training on using AI tools in the healthcare field and the risks associated with taking AI responses to inquiries about medical care without a grain of salt. A commonly cited concern of using generative AI in healthcare is the possibility for AI tools to produce hallucinatory responses. Particularly after attending OpenAI CEO Sam Altman’s fireside chat in which he claimed that hallucinations produced by ChatGPT were not a top concern for OpenAI, I was interested to see the opinions on hallucinations from a healthcare perspective.
One panelist actually countered Altman’s defense for a lack of concern about hallucinations—that humans are adaptable and often able to parse the reliability of responses from AI tools such as GPT—by pointing out how long it took for the medical community to adapt to new technologies like electronic health records. This intrigued me and has made me more curious about the ways in which the medical community will go about training physicians and trusting AI to revolutionize the way medical care is given in a technology-driven future.
— Written by Amber Campbell, CHEPS Student