ChatGPT, MidJourney and co. are poised to transform content production. We spoke with Nic Newman of the Reuters Institute about automation, virtual reporters and the risks associated with AI. Nic Newman is no stranger to driving technological disruption at media companies. The digital strategist and former journalist has played a key role in shaping the BBC's online services for more than a decade. He has also been the lead author of the Digital News Report for 11 years, which is considered the most comprehensive annual survey of news consumption worldwide.
The 62-year-old is currently a Senior Research Associate at the Reuters Institute for the Study of Journalism at Oxford University. In this interview, Newman shares his perspective on what the rise of generative artificial intelligence products - technologies that generate words, images and other media themselves - means for journalism.
It also highlights how media companies can build trust with their audiences in AI and why digital twins can increase the personalisation of news.
Newman: First of all, all this technology has been around for a while. Some journalists have been sort of experimenting behind the scenes, but it hasn't really seen the light of day. What's different now is that these tools are suddenly useful and freely available, so you're starting to see real concrete use cases for journalism. That really led to this wow effect.
Technology has gotten much better, one example being ChatGPT. And it's going to keep getting better: GPT-4, the next version of OpenAI's big language models, is said to already support Microsoft's redesigned Bing search engine and Edge web browser. These interfaces allow chatbots to have really useful applications.
Part of the "existential question" is really about the fact that automation is coming. A whole lot of things that journalists do these days are repetitive processes that can be automated. That means we're going to have to rethink what it means to be a journalist.
Because until now, part of journalism has been transcribing interviews like this one. But now, as we speak, Google is transcribing using artificial intelligence. We may still need new skills to figure out where it's going to make mistakes so we can do transcriptions, but generally AI will take care of that.
As for the possibilities, there are just too many. One very important application is this massive appeal of a fragmented audience with young and old people wanting different things and also the multitude of formats available. Artificial intelligence offers journalists the ability to not only create their story, but also to version it much more cheaply and efficiently than before. Plus, it can make content more relevant, personal and engaging for different people. So this breakthrough in AI will truly deliver on the promise of personalized news that we've been talking about for 20 years.
Journalists and robots work hand in hand
You mentioned AI error detection. One of the widely reported early content experiments with AI chatbots was the US online media outlet CNET, where AI wrote at least 75 articles that contained all sorts of errors, such as incorrectly explaining compound interest. What can we learn from this and other notable experiments with AI in journalism?
The CNET example is a very good one because it shows the possibilities and the direction of travel. You can let AI add context to stories or - as CNET did - create traffic generation guides. ChatGPT does this incredibly well in principle, but the trouble is that it can also look plausibly bad, which is what happened in the case of CNET. Now it's going to get better. The other key thing is that you'll be able to train these models on reliable content. ChatGPT is currently trained on thousands of sources, some of which are correct, some of which are wrong.
There's a lot of hype around ChatGPT now, and rightly so, but it's more of a shop window. In the future I expect OpenAI, Google and other players to license their chatbots to newsrooms that will pay for the service and train the AI with their own data to make sure the chatbots are reliable and meet their own needs.
One of the problems that journalists face at the moment is that there is still a lot of effort and time involved in providing context or background for pieces of journalism. In the future, the system will do this for you, perhaps with the option for the author to manually rewrite. Examples are context boxes or summarization tools such as bullets at the top of articles. They can be quickly auto-generated and verified by the author.
How does this lead to personalized content?
Artificial intelligence can help reporters version their stories by figuring out who they want to show bullet points or context boxes to. So people who generally click on stories with context boxes will see more of them, and vice versa.
The end game is to create blocks of content with a larger scope to suit different audience needs. AI image generators such as OpenAI DALL-E, MidJourney and others will also be valuable assets here.
What are some innovative ways these AI image generators are already being used and will be used in the future in journalism?
A good example is the American start-up Semafor. Their video department does a lot of experimentation and innovation. One of their really interesting projects is a series called "Witness", for which they interviewed victims of the Russian invasion of Ukraine. They then illustrated the interviews without actual footage using eyewitness accounts as text prompts for the AI image makers combined with the artist's style. The result is this really interesting film with rather blurry and cartoonish images.
Staying on the visual side of generative AI, you'll see AI-generated illustrations everywhere this year. We're already seeing them replace images from Unsplash or other providers at the top of articles.
When it comes to audio-related applications, taking a text article and converting it into an audio article or converting an audio article into different languages is really exciting. One use case is the cloning of journalists' voices: you can train AI with the voice of your audience's favourite presenter and have a synthetic voice read, for example, text articles. A digital twin would take this a step further: A real correspondent would have a virtual version that can answer questions via chatbots or virtual assistants like Alexa. And real journalists will still do the most important work, like moderating TV news.
Another huge area is the use of artificial intelligence for inspiration. If I need to do an interview but don't have time to research it, I can ask my favourite chatbot for some tips on what might be interesting. Just like you did (earlier) with a human research assistant. Another job that AI can do amazingly well is the work of sub-editors or copy editors. When you paste in a style guide, it basically finds all the bad commas, grammatical errors and bad syntax. In the future, the subeditor role will disappear and it will be more about managing text prompts. In addition, AI will perform search engine optimization much better than any human SEO expert.
Obviously, automation raises the question of how you signal this and how to be transparent about it. What's more, anyone can train someone to sound like a digital twin that doesn't, and they can spew garbage.
Bigger roles for AI and humans
This brings us to this fear of unverified and manipulative content flooding the internet. What are the implications of the vast amount of synthetic media for fact-checking and for trust in journalism?
It will be even harder for journalists to expose the deep fakes and everything else. Of course, artificial intelligence can also be used to detect fakes, so it's this kind of battle where artificial intelligence will be used to create and detect junk content. I think detection is going to be much more important in the journalistic mix in newsrooms and news agencies. More money needs to go into it and it will be much more augmented with AI tools to help fact-checkers do their jobs.
Secondly, I think the platforms have an even bigger problem. They need to bite and identify and promote more reliable sources, which requires much better detection and reduction of falsehoods etc with their algorithms, as well as dedicating more staff and resources to detection.
Third, news organisations need to focus more on building personal relationships. Part of that is making your content more relevant with AI, but a lot of it is really about helping users build deeper relationships with your brand by making your human talent stand out more. If we can build a more direct connection between journalists and audiences, we can rebuild some of the trust that has been lost in many countries over the last decade.
Journalists are hard to replace
What role does transparency play in restoring and building trust in AI-generated content? Getting more involved in explaining the processes behind content creation_
It's going to be really challenging because every time a new technology comes along, it can take a long time to get comfortable with it, both on the production side and on the consumer side. And a lot of bad things will happen in the meantime.
Recognising the change and developing a policy to address it should be the first priority. Two of the other things you need to do are, of course, transparent labelling and understanding the legal situation.
You must also train and educate your reporters to get their buy-in. There is a deep concern that AI will replace the work of humans, but journalism is one of the professions that is unlikely to replace AI, partly because AI doesn't know what has just happened or what is about to happen.
Journalists need to focus more on things that machines can't do. For example, if there is breaking news in Ukraine, AI won't be much help. But they'll be great for background: 'ChatGPT, how did we get here? Give me a timeline!"
That's the bottom line: it will give journalists more time for breaking news and real-time analysis. We just need to make sure the chatbots are trained on the appropriate material with the right supervision.
(DW.de/RoZ)
English 
Czech