by Amy Fallon
02 May 2024
Freelance journalist Anna Codrea-Radio calls AI an ‘intern’ because “it can do some stuff” but it “can’t be completely let loose on any of your work”.
Nikita Roy, host of the Newsroom Robots podcast, prefers an “insecure intern because it sometimes does not want to tell you when it’s wrong”.
An expert-led panel discussion of writers discussed how freelancers are using AI at the recent International Journalism Festival that I went to revealed some clever but cautious ways that solopreneurs and independent journalists are using the technology. But it also shed light on its limitations.
Anna Condrea-Rado, a UK-based freelancer who has written for the New York Times, Financial Times and Business Insider among others, told the audience that she used AI like it was an ‘intern’ because as a freelancer “you are all parts of the business”.
“You are both the CEO and the HR department and also the head of finance and also the writer and the journalist and all of it,” said Condrea-Rado. “Any way I can alleviate frankly some of the more boring but necessary tasks, that I think is where AI can really come into its own.”
She explained how she had put a work contract into Google Gemini to translate all of the legalese into plain English. (Don’t copy the contract verbatim, Condrea-Rado stressed, but paste specific parts of it into Gemini and redact the company’s name).
After doing this, the robot pointed out why some terms revolving around her intellectual property weren’t favourable for her. Condrea-Rado used this to write an email pushing back on those terms to the company – and successfully had them removed from the document.
“I think what’s interesting about that example is that as a freelance journalist I would have never hired a lawyer to look over that kind of contract,” she said. “So, it’s giving that access that I otherwise wouldn’t have and it’s something that is kind of very specific to the kind of challenges that you face as a solopreneur, one person shop.”
One of the hardest parts of freelancing can be the isolation. But Condrea-Rado also has an AI hack to cope with this.
“I often will sort of chat with the model just to get a bit of positive feedback on my stories,” she said. “I’ll basically ask it ‘does this make sense?’ because by and large it’s so much friendlier than an editor. It is going to say ‘yes Anna this is really written and your point here was really clear, this part could use improvement’.”
During busy periods, she has played with Motion, an AI-powered time management tool which determines the best schedule and a calendar for users.
But Condrea-Rado’s favourite AI tool was Otter. She said that she could actually quantify the time that she had saved using this.
It’s every writers worst nightmare these days. At the session, Condrea-Rado recounted an experience where a newsletter gig that she’d been hired to work on which involved curating links and articles and writing a blurb about each one was eventually given to a machine.
However, she said that the ‘silver lining’ was that she will now be able to spot this happening next time and understood that there is “potentially a job in being a robot whisperer”.
“(That is) going into a company where I can sell a service where I’m showing the company how to use the AI,” said Condrea-Rado.
Nikita Roy, a data scientist, journalist, and Harvard-recognized AI futurist who also spoke at the event in Perugia, Italy, said that she used AI mostly as a ‘brainstorming companion’.
“One thing that is really helpful is that you can chat with YouTube videos now,” she said adding that this can be done by pasting a link into Google Gemini. This has assisted her preparation for podcast interviews and with following press conferences and summits.
Roy said that users needed to do three main things when writing a prompt. “Start off with giving them a bit of persona of who they are,” she said, giving the example of a journalist or a lawyer reviewing a contract.
Secondly, give the robots a bit of context about who you are, for instance a freelance journalist writing a particular article for a company.
Thirdly, be more specific and give them ‘output criteria’. For instance, ask them to give you the different parts of a contract that might be unfavourable and which you might want to negotiate. What are the alternatives?
Despite the upsides of AI tools that some speakers described, Chris Stokel-Walker, a tech journalist for the likes of The Economist, BBC and WIRED and author of upcoming book How AI Ate the World, hadn’t been successful with getting Chat GPT Plus to develop pitches for editors.
He said that he put together different tools to try to be more efficient with commercial work such as gigs that involved summarising events but wouldn’t AI for journalism, questioning whether it had a nose for the news.
“Sometimes you can’t even teach that, we find that very difficult to do at my university,” said Stokel-Walker.
Other tools mentioned during the discussion were search engine and information assistant Perplexity, search engine Microsoft Bing, chatbot and large language model (LLM) Claude and DALL·E 2, which can create digital images from natural language descriptions.
Stokel-Walker added that it was just as important to try AI to learn how it didn’t work as much as it did.
All panellists stressed that freelancers had to exercise caution when using AI.
“It is so important to fact check everything as a journalist and that’s one of the main reasons you don’t see right now newsrooms deploying generative AI tools to users because of those risks of hallucinations,” said Roy. “That’s something we haven’t figure out yet.”
Condrea-Rado stressed that employers needed to figure out how to work with their freelancers.
“It’s really important going forward that (with) any policy that newsrooms write, which are predominantly for their staffers, there needs to be an active ‘how does this apply to our freelancers?’” she said.
We’re always interested in how freelancers are using AI – so what about you? Have you started incorporating it into your workflow, and if so, how? Share in the comments.
Great work, Amy! I suspect AI is easily going to be as disruptive to journalism, and content creation in general, as the Internet was.
Disturbingly, the disruption is likely to unfold much more rapidly. As always, there will be some winners from that disruption (think those ex-journos prescient enough to set up now substantial content marketing agencies a decade or so) but I fear there will also be plenty of losers.
Thanks for the insights, Amy!