Why Healthcare Needs Its Own Foundation Models | Ortet Founder & CEO Kyunghyun Cho
One of the world’s leading AI researchers has big plans for healthcare. And Thoreau just committed $500 million to bring that vision to life.
This week, we are the first to sit down with Ortet founder and CEO Kyunghyun Cho to talk about the frontier AI lab for health. After helping shape some of the technologies behind modern AI and selling his last AI startup to Genentech, Kyunghyun is now taking on a much bigger challenge: building a foundation model that can learn across biological, clinical, and operational data to deliver predictably better care.
We cover:
Why healthcare needs its own foundation models, and the “cold start” problem he thinks it can solve
What Ortet plans to do differently from general-purpose AI companies
Why Ortet secured $500 million before it had even really gotten started
What buying AI compute and a lobster roll have in common
How KC’s own experience with cancer shaped his vision for more predictable healthcare
About our guest:
Kyunghyun Cho is the founder and CEO of Ortet and the Glen de Vries Professor of Health Statistics and a professor of computer science and data science at New York University. Through his AI research, he co-developed the attention mechanism and the gated recurrent unit, both foundations of modern AI technology. Prior to founding Ortet, Kyunghyun built an AI platform for lab-in-the-loop molecular design, co-founding Prescient Design and, following its acquisition, establishing Genentech’s frontier research team. He previously served as a research scientist at Facebook AI Research (FAIR) after completing his postdoctoral fellowship at University of Montreal under the supervision of Prof. Yoshua Bengio. He received MSc and PhD degrees from Aalto University under the supervision of Prof. Juha Karhunen, Dr. Tapani Raiko and Dr. Alexander Ilin. Kyunghyun received the Samsung Ho-Am Prize in Engineering in 2021 and has served as a (co-)Program Chair of ICLR 2020, NeurIPS 2022 and ICML 2022, and also on the boards of ICML and ICLR.
—
Tune in: