AI makes inroads at Radio-Canada

MONTREAL —  The rise of generative AI has been a hot topic in the media for the past two years, often making headlines for its disruptive potential. However, this technology didn’t appear overnight; it is the product of years of technological evolution.

At Radio-Canada, the French-language public broadcaster, generative AI has been under scrutiny for years. The organization aims for a cautious and thoughtful integration. Maxime St. Pierre, general manager of digital media at Radio-Canada, shared the broadcaster’s perspective on the subject.

“Generative AI remains a disruptive element,” St. Pierre said, underscoring the complexity of this evolving technology. To better understand its potential and challenges, Radio-Canada partnered with Ivado, an interdisciplinary research consortium led by the University of Montreal. This collaboration has allowed the broadcaster to access experts who can help identify risks, opportunities and strategies for integrating AI across its platforms.

A cautious approach

Radio-Canada’s collaboration with academic experts has shaped its deliberate approach, focusing on “understanding the codes” of generative AI rather than rushing into untested applications. The broadcaster has conducted closed-circuit trials, such as content recommendation algorithms and synthetic voice generation. These proof-of-concept tests have allowed the team to refine their methods and ensure solutions are robust before full deployment.

The launch of ChatGPT in November 2022 marked a turning point. While many media organizations were caught off guard, Radio-Canada used the moment to define its principles for using AI. “We exchanged ideas with other francophone public broadcasters and studied nearly 25 AI charters to uphold our own values,” said St. Pierre. This effort resulted in seven guiding principles for AI use: Human accountability, transparency, ethics, copyright respect, security and privacy, user experience optimization and collaboration.

The audience at the center

Radio-Canada has since implemented practical applications of AI, including automating the transcription of radio programs into text. This serves dual purposes: making audio content searchable on platforms like Google and accessible to hearing-impaired audiences. The broadcaster is also exploring AI tools to rediscover and repurpose archival content, described as “forgotten gems,” to maximize the value of its media assets. St. Pierre emphasized the broadcaster’s commitment to ensuring content is accessible to all audiences.

In podcasting, generative AI has been used to automate episode summaries and create English translations to reach broader audiences. However, challenges remain. Tools like Whisper and OpenAI’s speech recognition model have struggled with Canadian accents and overlapping dialogue. “Our internal teams are highly skilled,” said St. Pierre, noting that Radio-Canada uses hybrid AI models that combine in-house systems with external tools.

The broadcaster’s AI program operates on two tracks: “AI Ready,” which focuses on current implementations and internal usage guidelines, and “AI Forward,” which explores innovative applications for future productions.

Public service

Radio-Canada is cautious about the risks of generative AI, particularly regarding synthetic content. All AI-generated publications are human-supervised, with strict measures to ensure legitimacy and availability on its platforms. The broadcaster is also integrating AI systems across its media divisions, supported by comprehensive training programs. “We’ve trained hundreds of employees in generative AI, prompt writing and critical thinking,” said St. Pierre, acknowledging that rapid technological advancements often render training outdated within six months.

Despite working with less than 5% of the budgets allocated by major tech companies like Google and Amazon, Radio-Canada aims to use AI responsibly to complement its public service mission. “Our goal isn’t to compete with tech giants but to advance thoughtfully and develop key products,” St. Pierre said.

This measured integration reflects Radio-Canada’s commitment to balancing innovation with responsibility. It ensures that AI effectively serves its audiences while upholding its core values.

A long-time media addict — radio, TV, web and press — the author says that not a day goes by without him listening to the radio. This article first appeared in French in RedTech’s sister publication La Lettre Pro.

More stories with a Canadian twist

Canada and Malawi boost community radio investigative skills

Calrec Type R enhances radio training at Québec college

Dielectric signs Pippin for Canada