How do we ensure AI makes safe choices? with Revan MacQueen | Approximately Correct Podcast

Ever wonder what exactly plants get up to all day? It’s much more than just sitting around to soak up the sun. Now, machine learning is helping to unlock the mysteries of how plants change over the course of a day, and the impact it could have on how we grow our food. Biochemist  Dr. Glen Uhrig joins hosts Alona Fyshe and Scott Lilwall to talk about how his lab is using machine learning to study how plants grow in different lighting conditions of the course of the day. This could lead to applications with big impacts for indoor farming in northern climates, and perhaps even feeding astronauts during space missions. Listen to discover the potential that artificial intelligence has in advancing scientific research in biochemistry, agriculture, and other fields.

Should powerful foundational AI models be kept under lock and key, or shared openly with the world? In this episode of Approximately Correct, we sit down with Joelle Pineau, a professor at McGill University, former head of AI research at Meta and current chief AI officer at Cohere.. She led the development and release of the early versions of Meta’s Llama models, a series of open-weight models that challenged the closed-door approach of other AI teams. Pineau argues that openness is the fastest path to better, safer AI, and that diversity in foundational models is essential to developing a limited ‘algorithm monoculture’.

On this episode of Approximately Correct, we talk with Michael Littman about the importance of making AI accessible and fun for everyone. A former division director for the AI division at the National Science Foundation, Michael shares his unique perspective on AI policy, communication, and his career in reinforcement learning. He also discusses his new role as Associate Provost of Artificial Intelligence at Brown University, where he is working to coordinate AI research and teaching across the entire university.

This AI doesn't replace jobs; it collaborates with human experts to get the job done.On this episode of Approximately Correct, we talk with Revan McQueen on the future of industrial control and AI's role in making it safer and better

Is it possible to make building a video game as easy as writing a story? What if artificial intelligence could be more than just a tool, and instead become a true creative partner?In this episode of Approximately Correct, we dive into computational creativity with Amii Fellow and Canada CIFAR AI Chair, Matthew Guzdial. We explore how AI is being developed to collaborate with artists and designers, breaking down the technical barriers that can stand in the way of a great idea. Learn about the future of human-AI collaboration, the philosophical questions behind AI art, and how research that starts with video games can end up solving problems in finance and even medicine.

How can we teach robots to safely navigate our unpredictable world?On this special live episode of Approximately Correct recorded at Upper Bound 2025, we talk with Mo Chen about combining classical and modern AI to create smarter, safer, and more robust robots.

Is AI this year's MVP? How is machine learning changing sports?On this episode of Approximately Correct, we talk with Chicago Blackhawks' David Radke about how machine learning is transforming the analysis of sports like hockey.

Could AI-powered ultrasounds save lives in remote areas? On this episode of Approximately Correct we talk with Dr. Jacob Jaremko about AI's revolutionary impact on medical imaging, particularly ultrasound technology, and its potential to transform healthcare.

In the latest episode of Approximately Correct, we’re taking the time to celebrate with Amii Fellow, Chief Scientific Advisor, and Canada CIFAR AI Chair Rich Sutton, newly-minted winner of the A.M. Turing Award, a prize that is often referred to as  the “Nobel Prize of Computer Science.”

Discover the secret to training AI with less data! On this episode of Approximately Correct, we talk with Amii Fellow and Canada CIFAR AI Chair Lili Mou about the challenges of training large language models and how his research on Flora addresses memory footprint concerns.