Now there are horror stories about news publishers. Mass-produce error-filled AI-generated content is well known. Other uses of AI are helping journalists make headlines for the right reasons.For example, New York Times Using computer vision for satellite imagery To count bomb craters in Gaza.
But beyond the most high-profile wins and faceplants, a modest but impactful application of generative artificial intelligence is quietly transforming newsrooms around the world. Generative AI presents new opportunities to enhance the reporting process and the storytelling itself. From assisting with copy editing to surfacing insights buried under vast mountains of data, the latest breed of AI can, and in some cases already has, given journalists powerful tools to improve their craft. doing. And as technology evolves, it promises to reshape journalism and the news business with the potential to enhance accuracy, efficiency, and depth of coverage.
A 2023 survey of 105 news organizations in 46 countries found that the majority of news organizations See the potential benefits for journalism with generative AI tools Like ChatGPT. In a survey by JournalismAI, a global initiative to keep news organizations informed about AI, nearly three-quarters of respondents said such AI applications would bring new opportunities to the field. did. Additionally, 85% of respondents, including journalists, technologists, and newsroom managers, have experimented with using AI for tasks such as creating images or generating article summaries.
Semafor, which launched six weeks before ChatGPT, is one such experimental newsroom. The media startup uses two AI tools, its in-house editorial bots and MISO, the latter of which helps it find stories that will power news. signal, a breaking news feed curated by journalists and AI-generated AI, highlighting news from publications around the world. Semafor's approach to AI is to have a clear picture of AI's current capabilities and limit its use to those areas, said Gina Chua, Semafor's executive editor.
“Basically, if you major in English, you can do a lot of things,” Chua said.
Beyond those tools, she Experimenting with what is known as search expansion generationor RAG. The technology allows the AI chatbot to generate large amounts of relevant information on the fly, prioritizing a library of data provided by Chua. This approach helps combat so-called hallucinations, a phenomenon in which generative AI models make things up.
Chua built the RAG tool using the Department of Justice's definition of a hate crime and fed the chatbot with specific scenarios, such as a white man attacking an Asian woman. The chatbot can label the call as a hate crime or not and explain the reason behind the call. She gave her second model the Trans Her Journalists Association's style and reporting guidelines. She then asked her AI to provide feedback about published articles and the level of misinformation about trans issues they contained. Both work well, and Chua would like to consider building her RAG model to help Semafor's journalists navigate how to report on sensitive subjects and how to quickly understand the content of a story. thinking about.
Chua created these chatbots in his spare time using easy-to-use tools. He said that most journalists agree that he needs to experiment with AI and that for certain tasks he should not dismiss it as not being as good as they are.
“They're probably not as good as humans in some ways,” she said. “It's a mistake to say, 'I want this to be human.' The trick is to say, 'I want this to be as good a tool as possible, but how do I complement it, and how does it complement me? “It is to say.”
Chatbots that drive subscriptions
The most common way people interact with generative AI is through chatbots such as: OpenAI's ChatGPT or google gemini. Newsrooms are also building chatbots for readers, such as his Skift, a business news website covering the travel industry.
Jason Clampett, Skift's co-founder and president, said Skift CEO Rafat Ali told an audience of travel experts, “This is going to be big, and we need to start working on it now. '', he recalls, when ChatGPT was introduced only a few weeks ago.
Mr. Crump took that advice to heart, and Skift's engineers quicklyAsk SkiftThe chatbot, trained on Skift's 30,000 news articles and reports, can also answer users' travel questions and suggest existing Skift articles for further reading.
“This is a great way to think about how we cover stories,” Crump said. “Other news sites might have to write about something as knowledgeable observers, and we can say, 'Oh, I see what this is about.'”
Ask Skift currently answers thousands of questions each week. The chatbot works much like his Skift paywall, allowing users to ask him three questions for free before prompting them to become paid subscribers. Ask Skift's paywall already translates into annual subscriptions of about $365, Clampett said.
He said Skift monitors reader questions to identify trends and potential story ideas. This creates new coverage such as: Why have travel costs risen so much?.
“In much the same way that you would look at Google Trends to find out what people are trying to do and what they're looking for,” Clampett said, “this is how we know that this is what people are looking for, and here's how we take this theme even further.”
Ask Skift is just the company's first generative AI experiment. The engineering team has also released an app that lets users ask questions in Slack, the widely used office chat platform. Clampett says the technology is getting easier to use every day, and more ideas are being explored.
“A lot of it is just trial and error and figuring out how to get a little bit better each time,” he said.
“Impatient to get ahead”
High-profile failures in sports illustrated is just the tip of the iceberg. Nearly a dozen other news organizations have published AI-generated articles containing errors, and the issue has been extensively documented by tech site Futurism.
Felix M. Simon, a communication studies and doctoral student at the Oxford Internet Institute, said failures like this stem from companies rushing to deploy generative AI capabilities without a thorough vetting process in place. He said it seems like it.
“What you can see from the outside is that in each case, there was a rush to look ahead and implement AI as quickly as possible,” he said.
Simon said journalist involvement was essential during production and before publication. The need for human oversight is also one of the reasons why journalists must learn how to work around AI tools.
“There's going to be a learning curve,” Simon said. “If you want to work with these systems in the first place and identify their strengths as well as their weaknesses, you need to get used to working with these systems.”
“AI is a tool. It's just a tool.”
While human oversight remains important, Simon cautioned against complacency about the potential for job losses to AI. That would not be a reason for journalists to be replaced in the short term. Deeper systemic issues, such as pressure to cut costs and a lack of sustainable business models, led to mass layoffs of journalists long before the advent of generative AI.
However, as AI technology advances, it may eliminate the need for human assistance, reducing the need for staffing even with current capabilities. Management could also use AI to justify further job cuts, Simon said.
Some worry that relying on big tech companies will exacerbate news organizations' already dire woes. As Rodney Gibbs, senior director of strategy and innovation at the Atlanta Journal-Constitution, wrote at the Nieman Institute late last year: Conversations about ChatGPT have echoes of “transition to video” In the age of digital media, Facebook, many newsrooms have lost their way.
It's a concern Nikita Roy, who heads the AI Journalism Lab at Craig Newmark Graduate School of Journalism, has heard but doesn't give much weight to. He says newsrooms aren't relying on technology companies to build audiences this time around. Rather, newsrooms are the customers who buy the product, giving news organizations more power in that relationship.
Importantly, AI tools are so easy to use that even smaller newsrooms can benefit, Roy said, and if used correctly, AI could benefit an industry that is perpetually struggling.
“AI is a tool,” she says. “It's just a tool. But it's a tool that helps us do more work and reach a larger audience, and those are the two things we have to do. .”
