LONDON — One of the pleasures of working for the European Broadcasting Union in Geneva was making regular visits to IMD, the Lausanne-based business school. A slide from one of their courses that stays in my mind demonstrates the rate of change in different industries. It was no surprise to know that the academics at IMD, impartial observers, selected the world of media as particularly fast-moving. And that analysis was made way back in time, over four years ago! With the introduction of Generative AI, I’m sure that their ranking has not changed. Indeed, media might even have strengthened its lead at the front of the pack.
In the last issue, I wrote about artificial intelligence and the regulations different broadcasting organizations are wisely applying to their usage of the technology. Since then, Finnish public broadcaster YLE has published its version, succinctly expressed in seven principles. They’re brief and to the point and provide a good checklist for other media organizations:
- A human is always responsible;
- We develop AI for the benefit of society, considering the environment;
- We enrich the user experience, content and our work with AI;
- Our use of AI is open, transparent and supports independent choices;
- We protect people’s personal privacy and data;
- We respect copyright;
- We continuously evaluate, develop and update.
You can see the whole reasoning behind this statement in English on its website. As a follow-up, I thought it would be valuable to concentrate on one organization, Swedish Radio, and see how AI was being employed “on the ground.”
AI as a “companion”
As is often the case, Nordic broadcasters have been among the most innovative. I was, therefore, unsurprised to find that SR now has someone with the initials “AI” in their job title: Olle Zachrison, head of AI and news strategy — not a role we could have imagined even five years ago! SR has been at the forefront of integrating AI into its workflow. While linear radio has seen only limited applications, the digital audio sphere has already undergone a significant transformation, demonstrating the potential of AI-driven innovations. It was a pleasure to speak with him about how things are developing in his organization.
So far, he says, SR has leveraged synthetic voices — based on its own presenters — in creative ways on digital platforms, such as employing them to read quotes and recommendations. SR’s own editorial algorithm, based on “old-fashioned” (not Generative AI), is already responsible for curating playlists within the app. This automation, facilitated by SR’s editorial algorithm, has led to a broader dissemination of news clips, with a special emphasis on content closely aligned with the values of the public broadcaster — “SR values.” Additionally, news audio and podcasts are automatically transcribed so they become accessible and searchable internally by production colleagues, leading to better utilization of existing content.
Structurally, he underlined the fact that SR has established a company-wide AI council with three key tasks: Identifying strategic AI applications; addressing journalistic, legal, and security issues; and facilitating organizational learning.
The main priority is ensuring that AI-generated content aligns with SR’s brand and values. Like its Finnish counterpart, SR has implemented a clear policy on Generative AI, reinforcing the importance of maintaining alignment with the station’s ethos. This policy, detailed in a blog post (www.publicmediaalliance.org/swedish-radio-publishes-policy-for-generative-ai), provides for the responsible integration of Generative AI into the production workflow.
There are several use cases worth mentioning: For instance, SR employs AI as a “companion” in newsrooms to suggest headlines or give rise to different perspectives. With this comes the need for clear guidelines. Zachrison regards it as fortunate that that SR employees are positive about innovation and digital transformation, since Sweden is, of course, a digitally advanced country.
The Swedish population can provide a valuable case study on how digital molds the way audiences consume content. What happens in Sweden today may well herald how things develop elsewhere in the future.
SR maintains a human-first approach, acknowledging the indispensable role of human creativity and editorial control. Generative AI is viewed as a tool to enhance, not replace, human efforts.
Humans first
Along with other organizations, SR maintains a human-first approach, acknowledging the indispensable role of human creativity and editorial control. Generative AI is viewed as a tool to enhance, not replace, human efforts.
Every piece of AI-generated content undergoes scrutiny from human editors and is subject to clearance by the legally responsible editor in each newsroom. This approach ensures that human intervention and editorial responsibility remain at the forefront of content creation.
To keep the audience informed, SR has produced programs exploring the possibilities and risks associated with Generative AI in daily news podcasts and linear radio broadcasts. The broadcaster expects the evolution of Generative AI in the radio industry to be transformative. In Zachrison’s view, radio and audio will become more conversational, using text chats and voice interfaces. At the same time, the launch of AI-automated radio stations, such as RadioGPT, poses new challenges for traditional broadcasters. He is also concerned about cloned voices; someone, or some organization could, for example, emulate the voice of a trusted SR presenter. SR uses its own cloned voices only in a limited way to guide listeners to more content, for instance, podcasts, rather than directly presenting content.
The challenges will certainly not disappear. In the meantime, SR underlines its human-first approach, ensuring that AI is complementary rather than replacing human creativity. Its journey with Generative AI seems to combine innovation and caution, ensuring things develop in a measured and deliberate fashion.
And now, I’ll let you into a secret. When preparing my questions for Zachrison, I wondered what ChatGPT might want to learn about the use of Generative AI at Swedish Radio. The proposed questions it generated were spookily like the ones I would have asked myself. Effectively, therefore, ChaptGPT, not I, conducted the interview.
The human element was present when I approved the list. Did I save time? Yes, a little. Might I have shaped the interview differently? Probably not. However, inspired by my success with the questions, I experimented with asking ChatGPT to construct the final article.
Frankly, I’ve spent more time playing with the text to make it sound like me than if I had written it in the first place! It’s still not entirely natural. I’ll be proceeding with caution in 2024.
The author was head of Radio at the EBU until 2020, and before that, managing editor of one of the BBC’s national stations. He currently advises media organizations.