(in alphabetical order)
Vrije Universiteit Amsterdam & CWI, The Netherland
Pre-reflections on a Post-trust Society
Abstract: We live in societies based on trust. In general, we trust what we read, we trust what we see and we trust what we hear. Parody and satire play with this trust. Misinformation attacks it. In our societies, people and organizations are held accountable for betrayals of trust. This betrayal can lead to social exclusion or to judicial action. But what happens in a post-trust world, one in which there is no inherent trust in what we read, see and hear? How will people function, how will members of society interact, how will parties be held accountable? This talk considers trust relationships in our current (technical) society and then considers how a post-trust world could function. We consider how responsibility and accountability can be preserved when there is no inherent notion of trust and trustworthiness. We ponder the role of semantic computing in this world The key question is: Does every technical silver lining hide a greater technical dark cloud?
Johns Hopkins University, USA
Toward a Science of Scaling Effective and Safe Medical A.I.: Workflow Integration, Real-World Validation, and Legal Thinking
Abstract: Artificial intelligence (A.I.) is rapidly transforming the healthcare industry, as evidenced by the Food and Drug Administration's approval of nearly 700 A.I. medical devices by July 2023. However, integrating A.I. into healthcare requires rethinking and redesigning clinical workflows, which presents unique challenges but also opportunities to revamp healthcare business models. This talk will explore the why, what, and how of combining machine and human intelligence in healthcare delivery. We will discuss (1) the history and current landscape of medical A.I., (2) how A.I. offers a differentiated product compared to how medicine has been practiced for centuries, and (3) how A.I. should be integrated into a healthcare ecosystem to maximize its meaningful use and impact. I will present a series of randomized controlled trials (RCTs) conducted in Bangladesh and at Johns Hopkins that provide real-world evidence that autonomous AI can improve clinical productivity, cost-effectiveness, and equity, and related service design considerations. I will also discuss recent work on the legal liability implications of physicians using A.I. in their medical practice, and the upstream and downstream implications. The discussion will touch on multiple stakeholders, and seek to inspire a science of scaling safe and effective medical A.I.
Jason H. Moore
Cedars-Sinai Medical Center, USA
Automating Biomedical Data Science
Abstract: Data science can be a laborious process with multiple rounds of method development, computer programming, application, evaluation, and interpretation. This is especially true for machine learning where dozens of algorithms are available each with multiple hyperparameters. This can be intimidating for novices and experts alike. Automated machine learning (AutoML) can simplify this process by letting the computer build an optimized analytics pipeline without human intervention. Our tree-based pipeline optimization tool (TPOT) was one of the first and most widely used AutoML tools. TPOT is open-source, programmed in Python, and makes use of the popular scikit-learn machine learning library. We introduce here a second-generation TPOT algorithm (TPOT2) with pipelines represented as directed acyclic graphs. We discuss the application of TPOT to the analysis of Alzheimer’s disease along with ideas for how biomedical knowledge in the form of an ontology-based graph can be used to inform feature selection, model selection, and interpretation.
University of California, Los Angeles, USA
Datafication and Calculation - How Far Should We Go?
Abstract: From the challenges we saw first-hand during the COVID 19 pandemic to questions of the climate and risks associated with habitability, we live in a time where the brand of 'division' has spread far and wide. Tied to this are the common critiques we see of technology - which tend to catastrophize and block a real re-imagining of how the powerful digital tools of today and tomorrow may bring us together. This talk will give examples around the world where digital technologies, far from isolating us or treating us merely as quantifiable individuals, bring us all together. The powerful ingredient here is human creativity and the compassion so many of us have that recognize that we are all in it together, as a global community. From policy I have worked on with major global figures to jobs of the future to the re-imagining of technology to transform waste into new forms of value and the voices of all peoples worldwide, I will give share numerous examples that point a path forward where technologies truly serve interconnection; one that celebrates and supports lives around the planet rather than cynically views it as something that is in perpetual crisis. This talk will provide us all with a concrete and powerful vision of a digitally mediated future that lifts all life up on our planet.