Episode 18: Artificial Intelligence
Author: Eoghan Colgan @eoghan_colgan
Special Guest: Greg McKelvey @DrGregMcK
25/07/18
Listen
Guest Bios
Greg Mckelvey
Dr. T. Greg McKelvey Jr. MD, MPH, is the CMO at KenSci, where he leads teams of doctors and data scientists to apply healthcare machine learning at scale across 3 continents. Prior to his role at KenSci, Dr. Greg served as Medical Director at Epsilon. Dr. Greg trained in Occupational & Environmental Medicine and Biomedical & Health Informatics at the University of Washington (UW). Dr. Greg received his Masters of Public Health (M.P.H) from the Johns Hopkins Bloomberg School of Public Health and his Doctorate of Medicine (M.D.) from Dartmouth Medical School.
Show Notes
Greg McKelvey is Chief Medical Officer at KenSci, a Seattle-based company exploring and formulating the impact of Machine Learning in predicting disease management risk. Eoghan and Greg discuss the basics of Artificial Intelligence in healthcare - what it means, current applications, ethical issues and predictions for the future.
Take Home Points
What does it mean?
Artificial Intelligence
using technology to replicate intelligence
when a machine mimics "cognitive" functions that humans associate with other human minds, such as "planning", "deciding", "learning" and "problem solving"
Machine Learning
a subset of artificial intelligence in the field of computer science that often uses statistical techniques to give computers the ability to "learn" (i.e., progressively improve performance on a specific task) with data, without being explicitly programmed
Artificial Intelligence in daily life
examples include:
search engines such as google
all social media sites
online shopping and advertisements
face recognition software
spam filters
autopilots and GPS units
Artificial Intelligence in Healthcare
current applications include image recognition which is at or above human capability
diabetic retinopathy, radiology scans, skin cancers etc
current research in Glasgow:
analysing data of COPD patients
predicting which patients will require admission
predicting how long they will likely stay in hospital
using wearables/remote-monitoring to intervene and avoid admissions
Prediction for the future
machines will take over mechanical/repetitive tasks
data entry
information retrieval
they can handle huge amounts of data to provide additional support:
predict patient flow and patient outcomes
offer diagnostic support such as image analysis
doctors will do more things that are uniquely human
problem-solving
creativity
patient interaction and communication
An uncertain future
the impact of artificial intelligence on humans is very uncertain
reflected in the polarising views held by luminaries in the field
it is unlikely there will be a world without the need for doctors but no-one really knows for sure
some fields of medicine will be more impacted than others depending on how big a fraction of their current role is automatable e.g. radiology
Ethical issues
tech giants have independent regulatory advisory boards
local work (such as the COPD project) is covered by local institutions who protect patient privacy/data etc
Recommendations for those with an interest
we should probably all have some degree of interest in the field as it will likely impact us all
be hands-on, sceptical and inquisitive
doctoral-level training is not required and there are fantastic ways to develop without formal training
MOOC's and other online courses
e.g. https://www.edx.org/course/artificial-intelligence-ai-columbiax-csmm-101x-3
https://hackr.io/tutorials/learn-artificial-intelligence-ai
practice with data sets using open-source machine-learning platforms from google/amazon/microsoft etc
(see below)
learn to code - even a basic understanding could be helpful if you intend to be more involved in AI
Links
OPEN-SOURCE MACHINE-LEARNING PLATFORMS
Google AI
https://cloud.google.com/products/machine-learning/
Microsoft AI
https://www.microsoft.com/en-us/ai/ai-platform
Amazon Machine Learning
https://aws.amazon.com/machine-learning/
Big Data and Machine Learning in Healthcare
https://jamanetwork.com/journals/jama/article-abstract/2675024
AI’s Birth, 1956
“We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer.” https://www.aaai.org/ojs/index.php/aimagazine/article/download/1904/1802
What Machine Learning Can and Can’t Yet Do
http://science.sciencemag.org/content/358/6370/1530
AI & employment
https://ideas.ted.com/will-automation-take-away-all-our-jobs/
Getting Started in ML
https://docs.microsoft.com/en-us/azure/machine-learning/studio/data-science-for-beginners-the-5-questions-data-science-answers
https://machinelearningmastery.com/start-here/
Healthcare Data Sets
https://archive.ics.uci.edu/ml/datasets.html
Artificial Intelligence vs Stupidity
https://www.newscientist.com/article/dn26716-fear-artificial-stupidity-not-artificial-intelligence/
Do the Amish use Google?
http://amishamerica.com/do-the-amish-use-computers-and-the-internet/
KenSci
https://www.kensci.com/company/about/
KenSci and GG&C