Agenda

April 17, 2018

8:00 AM -6:30 PM

 Time   Speaker Session Description
8:00 AM   Registration and Breakfast  
9:00 AM Robert Harrington Welcome  
9:15 AM Abraham Verghese/ Prober Honoring John Hennessy for his recent Turing Award  
9:30 AM Jonathan Chen Overview for the Day  
9:45 AM Abraham Verghese What This Computer Needs is a Physician What are the functions we expect and need from our (human) physicians? How should clinicians be driving adoption of AI technologies into medicine to avoid unintended consequences that impact real lives?
10:10 AM Eric Topol Healthcare Ex Machina: How Artificial Intelligence Will Transform Medicine With increasing hype for machine learning applications that can outperform human diagnosticians at a fraction of the cost, what transformative evolutions in medicine can we expect? Where could such advances go wrong and what would have to change about the existing culture and industry of healthcare to allow such change?
11:00 AM   Break  
11:05 AM Daniel Yang Diagnostic Excellence meets AI  
11:15 AM Lawrence Tierney + Tanya Gupta Human Diagnostic Intelligence - A Live Demonstration We will observe how a classical human expert physician and medical educator unpacks a live and interactive presentation of a mystery clinical case. What cognitive processes, base knowledge, and resources does an expert need to make smart medical judgements? How can this be effectively reproduced and disseminated (trained) to other practitioners?
12:00 PM Mark Graber Diagnosis - The Beauty and the Beast What are the cardinal features of diagnostic excellence? How good (or bad) are we now compared to what is possible? What role can technology play in disseminating expertise? What still requires a human practitioner that technology will not conceivably address (for decades to come)?
12:30 PM   Lunch  
1:15 PM Dean Lloyd Minor Precision Health vs. Physician Burnout Electronic medical data can enable grand visions for a new kind of healthcare in Precision Health, but the advent of electronic medical record systems is mostly recognized as a key source of physician burnout. What is Stanford's vision for using the best technology, data, and (artificial) intelligence to improve patient care without squeezing the humanity out of practitioners and patients alike?
1:30 PM Art Papier 20 Years Bringing Visual Diagnosis Decision Support Tools to Real-World Practice Ideas, prototypes, and commercial products for clinical decision support have existed for decades. What challenges, false-starts, dead-ends, and triumphant milestones have occurred in the industry attempting to bring these into real-world practice? Is "free" not cheap enough to get practitioners to use tools? What integration into workflows and existing systems is necessary? Will decision support technology "deskill" clinicians and interfere with patient relationships and confidence in their clinicians, or can it actually be used to enhance both?
2:00 PM Fei Fei Li Using AI to Illuminate the Dark Space of Healthcare Will advancing AI improve the average while systematically exacerbating existing biases? Who needs to be (and can be) involved in learning, building, and shaping a future society with pervasive AI? How do you assess an AI technologist's success and what does someone need to achieve that?
2:30 PM Erich Huang Creating a Data Science Culture in Healthcare Research How has the culture and infrastructure of classical healthcare research for evidence-based medicine evolved? What is the need, the promise, and the pitfalls of reframing the culture of medicine into a data science?
3:00 PM Rob Califf Who's Responsible? Regulatory Guidance on Clinical Decision Support and Software as a Medical Device How does one balance the drive for innovation against immediate patient safety? What are the issues at stake to decide what to regulate? What are the current standards of proof needed for deployment of software impacting medical care? When does Software as a Medical Device require formal proof of efficacy, and by what measure?
3:30 PM   Break  
3:45 PM Rich Caruana Friends Don't Let Friends Release Black Box Models in Medicine The talk is about the risks of deploying black-box machine learning models such as deep neural nets in mission-critical applications such as healthcare.  I’ll show that there are surprising landmines hiding in many data sets that sometimes cause machine learning to learn patterns that could be risk for patients.  Then I’ll show that we now have methods for training machine learning models that are transparent and editable so that these problems can be discovered and fixed before deploying the models.
4:15 PM Bob Kocher How to Make Our Crazy, Expensive, Amazing, and Uneven Health Care System Better Faster How do healthcare policy and economics drive forces inthe industry? In a largely fee-for-service world, what business use cases for using AI and data-driven approaches are actually incentivized in healthcare? What is the difference between clinical and operational ("back office") applications, and what are the opportunities in each? How many startup company pitches include claims of using "AI" or "Big Data" and how does one vet credible opportunities vs. empty buzzwords? In a moving landscape of healthcare policy reform and advancing information technology, where are healthcare IT and services investors placing their bets for the future?
4:45 PM Margaret Levi  AI, Automation, and Society Should people worry about AI systems taking their jobs? If not that, how else could such technologies be misused in ways that are both easily and not so easily predictable? What are the risks of consolidation of data within a few commercial entities? What expectations can and should people have for "understandable" algorithms driving decisions about their lives?
5:15 PM Jonathan Chen Closing  
5:20 PM Reception    
6:30 PM Close