際際滷

際際滷Share a Scribd company logo

AI in Medical Education A Meta View to Start a Conversation

Download as PPTX, PDF
Philip Bourne
Philip Bourne

- AI has the potential to significantly impact medical education and healthcare. - Chatbots and large language models can provide a rich training ground for students to learn, while augmented reality may change the student-patient dynamic. - AI tools like predictive analytics and imaging analysis can assist in research, diagnosis, and personalized treatment, but models are still limited and education of implications is needed. - If developed responsibly with oversight, AI could help democratize healthcare and create new industries, but history shows technology disruptions can also lead to deception if misused. The impacts and timeline of AI in medicine remain uncertain.

1 of 17
AI IN MEDICAL EDUCATION A META VIEW TO START A CONVERSATION Philip E. Bourne PhD, FACMI, February 26, 2024. /pebourne
I have no formal medical school training I was the first Chief Data Officer of NIH I have taught many pharmacy students I tend to see everything through the lens of data AI is only part of the story I complained that medical school curricula were not in keeping with DISCLAIMER/BIA S
THE (NON-MEDICAL) STUDENTS HAVE SPOKEN Its here, its a tool A tool like never before How to relate to the honor code Dont get caught up in the hype Wish professors knew more Want to know more about the implications Soon to be posted
HOW DID WE GET HERE? AI consumes data of all types: 90% of the worlds data was generated in the last 2 years Medical data is a mess Kudos to GW Bush Improved computer technology UVA was behind but is catching up Breakthroughs in algorithms and hence software providing persistent models This has got everyone's attention A zetabyte is 1012 gigabytes
CONSIDER OUR FAVORITE EXAMPLE - ChatGPT is one of many forms of AI A Large Language Model G generative ability to generate language, images, video, code P pre-trained unsupervised learning on vast amounts of content The training is done by neural networks that mimic the brain learning by adjusting weights of each neuron/node). Training stops when the right result is achieved. That network is then a model that can produce {mostly} the right answer from data it has never seen before T Transformers allow for parallel computation and treats text etc. as tokens ChatGPT
Diagnostic and image analysis Predictive analytics Personalized medicine Drug discovery Robot assisted surgery Virtual health assistants Clinical trials research Wearables Healthcare operations Mental health applications THE NOW - EXAMPLES
WHERE ARE WE HEADED? The current deep neural networks are equivalent to a rice grained size of the cerebral cortex and we are yet to explore most aspects of brain morphology Terry Sejnowski https://www.pnas.org/doi/full/10.1073/pnas.1907373117
What Does This Mean in Practical Terms with AI
AR changes the student- student; student-patient dynamic LLMs provide a rich training ground Student-mentor relationships will be different, but remain important. PEDAGOGY Images by DALL-E
RESEARCH A Biomedical Researcher The Holy Grail of Molecular Biology Food production Energy production Drugs Achieved by DeepMind (a Google spin off) not academia 30 interdisciplinary scientists working together not competing Compute power beyond a university
THREATS OPPORTUNITIES Creating a new health care industry
How Disruptive Will It Be?
THE 6 DS (PETER DIAMANDIS) Digitization Disruption Demonetization Dematerialization Democratization Time Volume, Velocity, Variety Digital media becomes bona fide form of communication Deception
KODAK A 6DS CASE STUDY Digital media becomes bona fide form of communication
WILL HISTORY REPEATS ITSELF? MEDICINE Digitization Deception Disruption Demonetization Dematerialization Democratization Time Volume, Velocity, Variety AI impact minimal Models reach human capacity Augmented reality, sensors Quantum computing Digital media becomes bona fide form of communication Learning modalities change Knowledge workers must adapt job market shifts Robotics Research practice changes
THANK YOU Philip E. Bourne February 21, 2024, Leadership Meeeting
TWO DIFFERENT FUTURES ACCORDING TO DALL-E

More Related Content

AI in Medical Education A Meta View to Start a Conversation

  • 1. AI IN MEDICAL EDUCATION A META VIEW TO START A CONVERSATION Philip E. Bourne PhD, FACMI, February 26, 2024. /pebourne
  • 2. I have no formal medical school training I was the first Chief Data Officer of NIH I have taught many pharmacy students I tend to see everything through the lens of data AI is only part of the story I complained that medical school curricula were not in keeping with DISCLAIMER/BIA S
  • 3. THE (NON-MEDICAL) STUDENTS HAVE SPOKEN Its here, its a tool A tool like never before How to relate to the honor code Dont get caught up in the hype Wish professors knew more Want to know more about the implications Soon to be posted
  • 4. HOW DID WE GET HERE? AI consumes data of all types: 90% of the worlds data was generated in the last 2 years Medical data is a mess Kudos to GW Bush Improved computer technology UVA was behind but is catching up Breakthroughs in algorithms and hence software providing persistent models This has got everyone's attention A zetabyte is 1012 gigabytes
  • 5. CONSIDER OUR FAVORITE EXAMPLE - ChatGPT is one of many forms of AI A Large Language Model G generative ability to generate language, images, video, code P pre-trained unsupervised learning on vast amounts of content The training is done by neural networks that mimic the brain learning by adjusting weights of each neuron/node). Training stops when the right result is achieved. That network is then a model that can produce {mostly} the right answer from data it has never seen before T Transformers allow for parallel computation and treats text etc. as tokens ChatGPT
  • 6. Diagnostic and image analysis Predictive analytics Personalized medicine Drug discovery Robot assisted surgery Virtual health assistants Clinical trials research Wearables Healthcare operations Mental health applications THE NOW - EXAMPLES
  • 7. WHERE ARE WE HEADED? The current deep neural networks are equivalent to a rice grained size of the cerebral cortex and we are yet to explore most aspects of brain morphology Terry Sejnowski https://www.pnas.org/doi/full/10.1073/pnas.1907373117
  • 8. What Does This Mean in Practical Terms with AI
  • 9. AR changes the student- student; student-patient dynamic LLMs provide a rich training ground Student-mentor relationships will be different, but remain important. PEDAGOGY Images by DALL-E
  • 10. RESEARCH A Biomedical Researcher The Holy Grail of Molecular Biology Food production Energy production Drugs Achieved by DeepMind (a Google spin off) not academia 30 interdisciplinary scientists working together not competing Compute power beyond a university
  • 11. THREATS OPPORTUNITIES Creating a new health care industry
  • 13. THE 6 DS (PETER DIAMANDIS) Digitization Disruption Demonetization Dematerialization Democratization Time Volume, Velocity, Variety Digital media becomes bona fide form of communication Deception
  • 14. KODAK A 6DS CASE STUDY Digital media becomes bona fide form of communication
  • 15. WILL HISTORY REPEATS ITSELF? MEDICINE Digitization Deception Disruption Demonetization Dematerialization Democratization Time Volume, Velocity, Variety AI impact minimal Models reach human capacity Augmented reality, sensors Quantum computing Digital media becomes bona fide form of communication Learning modalities change Knowledge workers must adapt job market shifts Robotics Research practice changes
  • 16. THANK YOU Philip E. Bourne February 21, 2024, Leadership Meeeting
  • 17. TWO DIFFERENT FUTURES ACCORDING TO DALL-E

Editor's Notes

  1. Demis Hassabis and John Jumper winners of 2023 Lasker Award Basic Biomedical Research