The Professor Behind a Generation of AI Scientists and Founders

Russ Salakhutdinov, PhD., is the UPMC professor in the Machine Learning Department, in the School of Computer Science at Carnegie Mellon University. He is also the CSO at Sooth Labs,; President Elect ICML Board; Ex-VP of Research at Meta; Director of AI Research at Apple,

Russ Salakhutdinov's students at Carnegie Mellon University are building the forefront of AI and Robotics. Zhilin Yang founded Moonshot AI; Jimmy Ba co-founded xAI; Devendra Chaplot is at Thinking Machines and Mistral; Nitish Srivastava co-founded Vayu Robotics and Perceptual Machines. His alumni populate Google DeepMind, OpenAI, Anthropic, Tesla, X AI, Meta's Superintelligence Lab, and the list goes on. His students wrote the Dropout paper and the Adam optimizer papers, which trained every major model in the last decade. And Russ will tell you "it's not about him". Today on Lab to Startup, we're going to learn about the magic that happens in his lab. Because one advisor doesn't produce this many founders and researchers by accident. We also explored a few other topics like AI in drug discovery, building Sooth labs, and Geoffrey Hinton's views about AI as an existential threat.

Russ received his PhD in machine learning from the University of Toronto in 2009 working for Nobel Laureate, Geoffrey Hinton, considered the God Father of AI.

Shownotes:

• Reasons for such level of success: Luck, student conviction, environment, freedom to explore

• Russ’s beginnings at Geoffrey Hinton’s lab: Freedom to explore, colleagues with Ilya Sutskever, Alex Krizhevsky, Yann André Le Cun, amongst others

• Power of conviction

• Russ’s lab absolutely believes in

• Early days were based on models; and now, we are in the age of scaling

• Some things work better at scale

• A little less innovation in industry compared to academia. In industry, innovation happens around scale

• What it takes to believe in something before the rest of the world does

• How can a professor/advisor help these self driven, high conviction students working in their labs?

• Paper reviews are complex, especially when they reject it

• Publishing is good, but people are looking for impact! Don’t just optimize for publications

• Scientific community doesn’t understand the role of scaling

• Balancing being a skeptic and also supporting a student as a professor

• Building Sooth Labs

• Teaching about AI to kids

• Coming up with ideas

• AI in drug discovery

• Supporting students when things are not going well

• Geoff Hinton now warns the world about existential AI risk.