AI Testing and Evaluation: Learnings from Science and Industry

Microsoft’s Eric Horvitz and guests Bruce Wittmann, Tessa Alexanian, and James Diggans discuss the Paraphrase Project—a red-teaming effort that exposed and secured a biosecurity vulnerability in AI-driven protein design. The work offers a model for addressing AI’s dual-use risks.Show notes

For the series finale, Peter Lee, Carey Goldberg, and Dr. Zak Kohane compare their predictions to insights from the series’ most recent guests, including experts on AI’s economic and societal impact, leaders in AI-driven medicine, and doctors in training.Show notes

Former Washington State Secretary of Health Dr. Umair Shah and Mayo Clinic CEO Dr. Gianrico Farrugia explore how healthcare leaders are approaching AI when it comes to public health, care delivery, the healthcare-research connection, and the patient experience.Show notes

Next-generation physicians Morgan Cheatham and Daniel Chen discuss how generative AI is transforming medical education, exploring how students and attending physicians integrate new tools while navigating questions on trust, training, and responsibility.Show notes

In the series finale, Amanda Craig Deckard returns to examine what Microsoft has learned about testing as a governance tool. She also explores the roles of rigor, standardization, and interpretability in testing and what’s next for Microsoft’s AI governance work.Show notes: https://www.microsoft.com/en-us/research/podcast/ai-testing-and-evaluation-reflections/

Drawing on his previous work as the UK’s cybersecurity chief, Professor Ciaran Martin explores differentiated standards and public-private partnerships in cybersecurity, and Microsoft’s Tori Westerhoff examines the insights through an AI red-teaming lens.Show notes

Daphne Koller, Noubar Afeyan, and Dr. Eric Topol, leaders in AI-driven medicine, discuss how AI is changing biomedical research and discovery, from accelerating drug target identification and biotech R&D to helping pursue the “holy grail” of a virtual cell.Show notes

Professors Daniel Carpenter and Timo Minssen explore evolving pharma and medical device regulation, including the role of clinical trials, while Microsoft applied scientist Chad Atalla shares where AI governance stakeholders might find inspiration in the fields.Show notes

In this episode, Alta Charo, emerita professor of law and bioethics at the University of Wisconsin–Madison, joins Sullivan for a conversation on the evolving landscape of genome editing and its regulatory implications. Drawing on decades of experience in biotechnology policy, Charo emphasizes the importance of distinguishing between hazards and risks and describes the field's approach to regulating applications of technology rather than the technology itself. The discussion also explores opportunities and challenges in biotech’s multi-agency oversight model and the role of international coordination. Later, Daniel Kluttz, a partner general manager in Microsoft's Office of Responsible AI, joins Sullivan to discuss how insights from genome editing could inform more nuanced and robust governance frameworks for emerging technologies like AI.

In the introductory episode of this new series, host Kathleen Sullivan and Senior Director Amanda Craig Deckard explore Microsoft’s efforts to draw on the experience of other domains to help advance the role of AI testing and evaluation as a governance tool.