Show Notes
Back in late 2022, Alona Fyshe was working on a TED Talk about how we evaluate the performance of Large Language Models. Little did she know, there was something big on the horizon.
In this episode, Alona tells co-host Scott Lilwall about how the release of ChatGPT shook the field of Natural Language Processing (and made her rethink her entire talk.)It’s a story about how we test machine learning models and a tremendous shift in the public’s expectations of what AI can do.
You can see the final version of Alona’s talk here: https://www.ted.com/talks/alona_fyshe_does_ai_actually_understand_us
Production Credits: Lynda Vang, Keely Booth - Producer, Chris Onciul - Video Production, Jen Tomski - Social Media
Music Credits: Main Theme - Brooklyn Bridge by Lunareh
More from Approximately Correct: An AI Podcast from Amii
How do you build a robot that had empathy? On this episode of Approximately Correct, Amii Canada CIFAR AI Chair Angelica Lim joins hosts Alona Fyshe and Scott Lilwall to discuss the complex world of social robotics. As the director of the ROSIE Lab at Simon Fraser University, Lim is focused on making robots friendlier, more fun, and more useful, allowing them to interact with humans more effectively and react to our social cues. She dives into the challenges of adding empathy to artificial intelligence, why cultural diversity if important when it comes to training social machines, and why making robots into artificial friends might not actually be the ideal goal.
There are all kinds of big promises about how artificial intelligence will change everything. But there's a very real transformation quietly happening in healthcare that is happening just under the surface. In this episode of Approximately Correct, hosts Alona Fyshe and Scott Lilwall sit down with Amii Fellow and Canada CIFAR AI Chair Dr. Amber Simpson to explore how machine learning is making a tangible difference in cancer detection, research and everyday medicine.
How do we know where we are, and where we are going? While we use our eyes to take in the information, it is our brains that are doing the heavy computing as we navigate the world. Now, new research using machine learning models is perhaps giving us a glimpse into what is going on inside our minds, and how the cells needed to navigate the world might form. On this episode of Approximately Correct, we are joined by two Amii Fellows and Canada CIFAR AI Chairs — Marlos C. Machado and Quinn Lee — who combined their expertise in machine learning and neuroscience to better understand the mysteries of navigation.
What happens when you give an economics final to ChatGPT? On a special episode of Approximately Correct, filmed live at Upper Bound 2026, we're joined by Canada CIFAR AI Chair Kevin Leyton-Brown. His work focuses on the underexplored intersection of machine learning and microeconomics — two fields that might seem wildly different, but have a lot in common. Tune in for a wide-ranging discussion on multi-agent systems, the reasoning abilities of LLMS, and how AI is a crucial part of Canada’s economic future.
In just five years, Upper Bound has grown from a 400-person birthday party to one of Canada’s largest AI-centred events, drawing in thousands of people each year. But it isn’t the only thing that has changed: the way we talk about and use machine learning is also completely different in 2026.