Cover art by Katrina McGaughey

In November 2022, OpenAI released ChatGPT, a chatbot application powered by (what was then) its most advanced text-generating AI system: GPT-3. ChatGPT’s ability to produce clever, well-written text in response to users’ prompts had many journalists wondering: how will this new breed of text-generating AI, known as large language models, affect the journalism industry, especially newswriting?

In January, we got our first taste of what this AI-driven future might look like. US online tech news site CNET was caught quietly experimenting with an as-yet-unspecified large language model tool. They used it for over 70 financial explainers—but without publicly announcing the initiative. Even by CNET’s own admission, the results were mixed.

It’s an issue that didn’t make it into the magazine, so we thought we’d devote this, the first episode of the sixth season of Pull Quotes, to large language model tools and their potential to disrupt the journalism industry.

First, we talked to a computer scientist, who explained how these tools are built and how they work. Then, we turned to a newsroom digital developer. His experience developing other forms of text-generating AI tools means he is ideally suited to assess the new technology’s potential impact on news journalism.


Alona Fyshe is an associate professor of computing science at the University of Alberta’s faculty of science. In her research, she combines computational linguistics, machine learning and neuroscience to study how large language models process language—compared to how the human brain does so.

Lucas Timmons is a digital developer for data-driven content at Torstar Corporation, where he works on newsroom automation. Among other things, he creates tools that source and process data, which they can then use to automatically generate stories. Before then, he was the head of the Canadian Press’s digital data desk.

Further reading:

I, Reporter: How “robojournalism” can save local news—the Review of Journalism

The Threat of “Deep Fake” Text Generation: Pull Quotes Series 3, Episode 10—the Review of Journalism

How The Canadian Press is using bots to build the future of election coverage—J-Source

COVID-19 with automated journalism: leveraging innovation in times of crisis—Columbia Journalism School’s Tow Center for Digital Journalism

How do computers understand meaning and language? | Alona Fyshe—the Walrus

GitHub Copilot: Your AI pair programmer

Create production-grade machine learning models with TensorFlow

Please check out the Spring 2023 issue of the Review of Journalism at Even better, buy a print copy. When you do, look out for Charlize Alcarez’s feature about another newsroom use of AI: the Globe and Mail’s AI-powered content curation tool—Sophi.
You can find the Review of Journalism on Twitter, Instagram and TikTok.

This episode of Pull Quotes was written by Tim Cooke; produced by Angela Glover; and fact-checked by Stephanie Davoli. Silas Le Blanc was the sound-designer and editor.
Tara De Boer was the production assistant. The executive producer was Sonya Fatah.

About the author

Photo of Silas Le Blanc
Silas Le Blanc
+ posts
Photo of Tara De Boer
Tara De Boer
+ posts

Sign Up for Our Newsletters

Keep up to date with the latest stories from our newsroom.

You May Also Like
The CBC Toronto building as seen from one of the higher floors.

Forever Temporary

They work hard. They stick around. And yet many of them are never offered jobs. How CBC treats its temporary workers—and how some are fighting back.
A sign for the former Toronto Star offices at 1 Yonge Street.


For generations, the Toronto Star has had a reputation as a progressive paper. But does its newsroom culture live up to its own ideals?