UCSF has an outstanding online grand rounds series - here - and this week it posts an excellent one-hour, five-speaker review of some newest trends for AI in Medicine. The Grand Rounds makes CHAT GPT its focus, but includes other topics. Find the video at YouTube here:
https://www.youtube.com/watch?v=j-aOCuzfxUI
Faculty speakers are Drs. Bob Wachter (chairman), Aaron Neinstein, Sara Murray, Dan Lowenstein, and Atul Butte.
One of the most arresting moments for me was when Dr Butte gave Chat GPT - which isn't even specifically trained on medicine - an oncology unstructured record, and asked it to summarize the biomarkers in the record. Which it instantly could do:
I couldn't resist the urge to turn AI on itself.
I ran an autotranscript of the audio, and fed it into Chat GPT (10,000 words, but no more than the limit of 1500 words at a time). This gave the whole hour summarized by AI into ten-minute segments. Last, I asked Chat GPT to summarize this collection of six mini-summaries into one single summary of the whole hour. This is what I got:
- The discussion on Chat GPT is part of the UCSF Medical Grand Rounds and is hosted by Dr Bob Wachter, Chair of the Department of Medicine at UCSF. The session focuses on the current and potential uses of Chat GPT in healthcare and its implications.
- Four experts in the field, Drs Aaron Neinstein, Sara Murray, Dan Lowenstein, and Atul Butte, present on different aspects of the topic. Dr. Neinstein discusses the slow adoption of AI in healthcare and suggests that AI can augment the patient-doctor relationship and help with administrative tasks. Dr. Murray talks about GPT-3 and mentions potential applications in healthcare, including clinical documentation, streamlining literature review, and task automation, but also highlights risks such as bias and hallucination. Dr. Lowenstein discusses the potential of Chat GPT in medical education and research, but also highlights the need for improvement in knowledge processing. Dr. Butte discusses the advancements in AI technology and the importance of caution and ethical considerations.
- The speakers in the panel discussion touch upon the impact of AI on education and healthcare, the ethical implications of using these tools, and the potential for misinformation with the use of GPT. They have mixed feelings about the topic, with some being both excited and scared, while others are optimistic or cautious. They agree that the use of GPT in healthcare is complex and important, and will require ongoing discussion and careful consideration.
See my Google Doc summary to access the summary above, then, the six transcript sections that led to the summary, and if of interest, the Otter auto text used as input to the AI.