Title: "The future of AI-driven science"

Abstract: Since Marie Curie discovered radioactivity at the turn of the 20th century, the fundamentals of research have changed little. A scientist has an idea and designs an experiment. They go into their laboratory, close the door and conduct the experiment — typically involving manual tasks that are repeated over and over again with only slight variations. They record and analyze the data. Often, insights are gleaned only after dead ends are explored, failures are overcome and significant time and effort are invested. The process is time-intensive, prone to human error, requires tremendous resources, and often not reproducible.

At Carnegie Mellon University (CMU), we are throwing out this dated model and radically changing how science is conducted. Across much of science, automated instrumentation is beginning to replace repetitive tasks in the lab, and to allow new scientific questions to be asked through the collection of vastly larger data sets. The key lies in partnering computation, robotics and data analytics with innovative scientific research, which will make the discovery process more transparent, less prone to error and more reproducible — and it will accelerate innovation.  By automating the process of experimentation and data analysis in a remote controlled laboratory driven by artificial intelligence (AI), robotics, and machine learning CMU scientists and researchers will be able to focus on asking tough questions and designing breakthrough experiments that lead to world-changing advancements.  How AI-driven science changes the scientific landscape is the topic of this presentation.  Specifically, we will think about how science is conceived, experiments designed and analyzed, data and protocols are shared, as well as the impact of this framework on open science.