Man vs Machine

As my mum always told me, bit of politeness goes a long way, especially when writing a prompt for ChatGPT. So, here goes:

In literally the blink of an eye – too fast for me to be able to measure it – my ChatGPT rival came up with an ok sort of headline, 507 words divided into six bite-sized sections and this portentous conclusion: “As we navigate the evolving landscape, the marriage of human expertise and AI capabilities promises to redefine the future of journalism, ensuring that the industry continues to deliver high-quality, relevant, and timely information to audiences worldwide.”

Ok, job done then. That’s me off for some pre-Christmas cheer and feet up while the team at InPublishing Towers lovingly prepare my (their) golden words for the next issue.

But, of course, it won’t do, will it?

The sections were all perfectly serviceable, in an online encyclopaedia sort of way, but there were no verifiable facts or expert comment and opinion, two of the basic constituents of journalism as we know it.

Indeed, in the first section called ‘Automated Content Creation’, we are told that, “news agencies are leveraging AI algorithms to generate routine and data-driven news articles, allowing journalists to focus on more complex and nuanced storytelling”.

No evidence or examples of that, so we’ll have to take your word for it then.

So, in search of some definitive answers, I strapped on my generative AI boots and signed up for the 2023 JournalismAI Festival – an online celebration that attracted no less than 700 people from all over the world.

I lined up with delegates from Algeria, Vietnam, Argentina, Singapore, Nigeria, Australia, Korea and Kyrgyzstan among others to hear some very clever people talking about some very clever things. Strangely a poor turnout from North America – make of that what you will.

Kicking off was everyone’s favourite AI expert, Charlie Beckett from the London School of Economics, who as the director of JournalismAI was fundamental in making it all happen. The presentations came thick and fast: Using AI to detect bias in real-time in English and Arabic content, a Kenyan team who looked at lessons learned in localising articles with AI and an interesting session on ‘analysing UK MPs’ financial interests’ led by former BBC journalist Clare Spencer.

It was all pretty technical but Louis Harkell, senior reporter at Undercurrent News and the founder of ‘multi-lingual stock exchange filing aggregator’ FilingReader, cut to the chase with this cogent comment:

“From today’s great presentations it seems AI is not yet capable of 100% independently automating original reporting, from conception to publication, or a particular part of the production process (researching, interviewing, transcribing, writing). This seems to preclude large-scale, mass-produced journalism by AI. A human is always needed ‘in the loop’.

“The exception to this is perhaps scraping and re-writing news, which isn’t really journalism but plagiarism, as it doesn’t require any original reporting by the AI. And this is low-quality journalism that readers perhaps will be able to see through anyway…”

Ah, the P-word makes an appearance.

The concept that AI is not able to come up with original journalism, but can beg, steal or borrow from you and me was also explored in a session called: JournalismAI 101: FAQs, Mythbusters, and AI Tools.

Addressing the issue that AI can still feel like a buzzword in newsrooms, with many people unsure about what AI in newsrooms would actually look like, the session aimed to tackle some common misconceptions, address the challenges of AI, and “equip you with all of the answers you need to feel confident about the future of AI in journalism”.

Helping hand

Laura Ellis, head of technology forecasting at the BBC, offered her take on how AI can help: “I’d say headline generation… always hard for humans and AI will give you options. Translation and transcription where AI is getting better and better and offering new possibilities to summarising big documents when researching new topics.”

And Jane Barrett, global editor for media news strategy at Reuters, cited “production tasks more than newsgathering” and “areas where tired brains start losing their sharpness. Which could be writing headlines, checking scripts for repetition and comparing texts.

“For now, I think it’s useful to stick to data and language tasks rather than giving away the keys to any knowledge and news tasks.”

The festival roamed over large prairies of current knowledge from the outwardly arcane ‘Decoding complex timelines with Temporal Knowledge Graphs’ to the more accessible session on photography, image generation and AI – ‘pushing the boundaries of what photography is and what isn’t’.

But you couldn’t help but feel empowered by the capable people holding our webinared hands through it all.

I also got the feeling that my rival ChatGPT writer had snuck in at the back. Despite the lack of basic journalism, the ideas in the auto-generated piece were echoed throughout the festival.

So, yes to ‘Automated Content Creation, Enhanced Data Analysis, Personalized News Delivery, Fact-Checking and Verification, Language Translation and Accessibility and Audience Engagement and Social Media Monitoring’ as defined by the bot.

Perhaps the key lies in using this as a jumping off point with real people adding that original journalism. Now, ‘talking to people’, wouldn’t that be an idea…

View from the Gurusphere

Out there in the gurusphere, there’s no shortage of people with something to say. Ian Carter, editorial director of Iliffe Media Group, has some sage advice: “If you are worried about a robot taking your job, my advice would be to swiftly retrain as an organiser of meetings about AI as it really does seem to be a growth industry.”

Carter is also a member of the Editors’ Code committee and chairs the working party for the BBC-funded democracy reporter scheme and I can personally confirm he is not a robot.

“AI can and inevitably will be used to crunch huge datasets and produce financial reports, property listings and the like – and that means publishers have to double down on what the robots can’t do,” he says.

“Let the AI do the donkey work, but don’t take it as an opportunity to do away with the reporters. Instead, set them free to do what AI can never do – fostering the relationships that lead to the stories that make a difference.”

Gary Rogers, currently senior newsroom strategy consultant at Fathm, and previously big cheese at the BBC and ITN believes there is certainly more talking than doing in the world of AI.

“I believe at issue is how some have jumped to a solution that GenAI should be used to originate journalism. This is madness. But AI tools placed smartly in the hands of journalists can produce great results,” he says.

“New AI technologies have the capacity to do more if under clear editorial control. Journalism has always moved, albeit sometimes slowly, to adopt new tech. This is not the binary of machine v journo but finding the way for journalists to use these tools to the benefit of readers.”

As the founder of Help Me Investigate, a platform for crowdsourcing investigative journalism, Paul Bradshaw helps media organisations and journalists adapt their work processes to new media along with his day job as leader of the MA Data Journalism course at Birmingham City University.

Bradshaw gave a talk to the BBC Local Democracy Reporters conference called, ‘How to work with a bullshitting robot’.

Part of his advice was: “Don’t think ‘Learning how to use AI’. Think: ‘Learning how to write effective prompts’. Four-year-olds come up with great ideas. You’re the one that provides the reality check. You’re not so perfect either – let it help you, but don’t let it drive.”

Back at Iliffe Media, Paul Fisher is group audience development manager and the man tasked with making AI happen in the newsroom. “There’s a lot of hype, but no one really has a plan,” he says. Fisher points to practical applications like improving SEO headlines and other “mundane, repetitive tasks” but can’t see AI actually doing any writing.

“We were caught with our trousers down 20 years ago with the internet, which is why there’s so much talk now to make sure it doesn’t happen again.”

Verdict

It’s not difficult to worry about where our collective trousers are right now. The editorial side of the publishing industry – especially newspapers – has been notoriously slow to get on board with technological developments.

Maybe that’s just a reflection of the generation that was pulling the strings in the largely unchanged ‘legacy newsroom’ while the internet revolution was raging outside. Or maybe it was short-sighted owners and management who hadn’t the foresight to invest in the nascent technology and the people to run it.

But AI is offering some different opportunities, with the possibility of automating some tasks and streamlining the operation. That will lead to a saving (fewer people!) which may just help publishing companies weather the current storms – and those around the corner.