In the latest episode of the Approximately Correct podcast, Amii Fellow and Canada CIFAR AI Chair Matt Taylor discusses how AI and human beings work best when working together. Taylor discusses the importance of a Human-In-The-Loop (HITL) framework, where human beings and artificial agents work together to achieve results better than either working alone.

In his conversation with hosts Alona Fyshe and Scott Lilwall, Taylor says: “As a machine learning person, I don't want a fully robotic surgeon now.” “I want a doctor that's there to supervise any AI decisions so that their background knowledge, their common knowledge, their experience can come in and make sure the AI isn't making a silly mistake."

Taylor stresses that it isn’t just a question of providing human oversight — a true HITL approach is collaborative and iterative, making best use of what people and AI are strongest at. It also means that machine learning scientists need to work closely with subject matter experts when designing AI systems from the start of any ML process. He poins to his current work using ML to improve power grid use as an example. Taylor also touches on the importance of explainable AI (XAI) in helping people build trust in AI-aided decisions.

Tune into the full episode to hear more and subscribe to Approximately Correct on Spotify, Apple Podcasts or your favourite podcast platform.

Approximately Correct: An AI Podcast from Amii is hosted by Alona Fyshe and Scott Lilwall. It is produced by Lynda Vang, with video production by Chris Onciul.

You can hear episode three of Approximately Correct on Spotify, Apple Podcasts, Google Podcasts and other podcasting services.