Can We Trust Scientific Discoveries Made Using Machine Learning? with Genevera Allen

800 800 The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence)

On today’s show, we’re joined by Genevera Allen, associate professor of statistics in the EECS Department at Rice University, Founder and Director of the Rice Center for Transforming Data to Knowledge and Investigator with the Neurological Research Institute with the Baylor College of Medicine.

Back in February, Genevera gave a talk at the American Association for the Advancement of Science meeting titled “Can We Trust Data-Driven Discoveries?,” that caused quite a stir amongst many members of the ML community. In our conversation, Genevera details the goals of her talk and gives us a few use cases outlining the shortcomings of current machine learning techniques. We also discuss reproducibility, including inference vs discovery, and the lack of terminology for many of the various reproducibility issues.

About Genevera

From the Interview

“More On That Later” by Lee Rosevere licensed under CC By 4.0

1 comment
  • Gireesan Namboothiri P

    That was an amazing talk. I came across such cargo cult research (as quoted by Feynman) in my career. Two of them on top are, assuming that a passive infrared sensor can detect strict boundaries in open ground and build a complete target tracking system; making a sound filter with pipes for an exhaust system. Both lacked the science to back up the assumptions to make it practical. Studies like these should be questioned by the stakeholders/field experts on the assumptions/ discoveries made in Machine Learning

Leave a Reply

Your email address will not be published.