I’ve never seen a group of more than 450 media professionals collectively look so shell-shocked.
As part of the opening act in this year’s The TravMedia Summit, invited speaker Matt Barrie, CEO and founder of freelancer.com, had just taken me, and them, on a wild ride through some of the many emerging applications for artificial intelligence (AI).
Among them, he provided a convincing demonstration of generative AI mimicking his face and voice while speaking Mandarin (a language he does not speak); presented research suggesting that patients preferred AI doctors over real ones, because the answers were longer, better and more empathetic; and revealed how the latest generation of deepfakes could hold real time conversations; and much more.
Barrie spoke of a world in which software wrote itself, where AI could ace any human test from the HSC to the bar exam, and where text to video capabilities could turn a simple prompt (“Make me Top Gun 17 with Vladimir Putin fighting Tom Cruise over Paris”) into a movie.
These developments, he added, were coming on like a freight train, if not already here.
“It’s going to happen this year – your future work is going to be dramatically different, and it’s gonna be a crazy world out there,” Barrie said.
For a roomful of people who make a living writing stories, taking photographs, or otherwise creating content, it was a deeply unsettling moment.
As Barrie wrapped up, into the stunned silence, someone asked a question.
It went something like, “Do you think that readers and audiences might actually be looking for authenticity and human connection? Do you think there will be a backlash at some point as people come back to the age-old stories?”
It was the question on everyone’s lips.
Barrie sought, I think, to provide some reassurance, in that AI could make media workers “more productive”, by doing all the background research for you.
So, for example, if a travel writer was preparing a story on things to do on the Amalfi Coast, AI could draw on its vast knowledge base – including photographs, videos, travel accounts and social media comments – and provide a useful summary.
Right now, he acknowledged, there were things it couldn’t do, but just give it a few more iterations.
“These abilities that are coming out of the (large language models) are not predicted by the people who develop them, and nobody knows actually how they’re actually emerging,” he added.
But Barrie’s view seems to be that AI’s capacity for empathy and creativity might be just a matter of time.
Old school versus Otter
When I was in journalism school many years ago, I used a cassette tape recorder (yes, it was a very long time ago) to capture a mock press conference for a university assignment.
It was Journalism 101 and the lecturer had impressed upon us the importance of capturing every word with absolute, non-negotiable, 100 per cent accuracy.
We were all there with our pre-prepared questions and our cassette tape recorders in our sweaty undergraduate palms, knowing that our lecturer was also recording the press conference, and that if a single error appeared in a quote in the story we subsequently wrote, then it would mean an instant failing grade (or so I remember).
Let me tell you, the number of times I hit rewind and play on that thing, to listen and re-listen and then listen some more, would make your ears bleed.
Now, of course, I use an AI-powered transcription service like Otter, like everyone else.
It saves me oodles of time, effort, blood, sweat and tears, helps prevent burnout, and is just one example of how AI can pave the way towards huge productivity gains.
Other examples were highlighted by James Meese, Associate Professor of RMIT University, in one of several training sessions recently run by The Walkley Foundation. AI, for instance, can translate or reformat existing content.
To be clear, he was not talking here about the practice of article spinning, where a writer’s copyrighted work is ripped off and remixed – though it must be said, human agents are just as capable as AI of doing this.
Rather, I’m talking about repurposing your own material by turning one format into another – for example, chopping up an article you’ve written into a series of social media captions, or distilling it to provide a preview summary of what’s to come.
As an aside, Meese noted that Swedish daily Aftonbladet found that readers spent longer on articles with AI-generated summaries, which is a positive development. (However, I tried to get AI to write a summary of this story. It didn’t work terribly well. Perhaps I need to work on my prompt engineering.)
So as these and the other visceral impacts of AI ripple through and reshape different facets of our lives, it’s worth considering where it shines.
It can process vast amounts of data, with a speed and accuracy that’s far beyond human capacity.
It can automate mundane and repetitive tasks, like answering routine customer inquiries.
Its sophisticated algorithms can predict individual preferences and improve some decision-making. Apparently it can even fake a doctor’s bedside manner.
AI goes to the Amalfi Coast
But there’s a bunch of stuff that AI seriously sucks at too.
Though Barrie is right to suggest that it can provide a springboard for an article on “things to do on the Amalfi Coast”, AI-generated text tends to unfurl in banalities, generalities and cliches, such as the opening line of this story ChatGPT just spat out for me:
“Nestled along the picturesque shores of southern Italy, the Amalfi Coast is a destination straight out of a postcard, boasting dramatic cliffs, colorful villages, and crystal-clear waters …”
Barf-worthy, isn’t it? It goes on, reasonably enough, to include such items as “explore Amalfi town”, “take a boat tour”, and “relax on the beach”, but all sketched using more of the same broad brushstrokes that’s the signature of “desktop research”, where the writer clearly hasn’t been there, and doesn’t know what they’re talking about.
(Read more about why your own backyard is the best place to launch your travel writing career.)
One day, maybe very soon, technology may improve to the point where you can AI-generate a story like this and immediately hit the “publish” button.
But for now, even if you use AI for such a task (and I don’t) you’d always need to inject some personality, add plenty of detail and check AI’s work.
Because contrary to Barrie’s suggestion that it’s the perfect research partner, it gets things wrong – often, shockingly so.
Meese, on the other hand, describes AI as “unreliable … not helpful and quite risky” which I think is much closer to the money.
Not to put too fine a point on it, AI also makes shit up.
According to software testing firm Testlio, AI learns to make predictions by finding patterns in data. Where there are gaps, it learns incorrect patterns, which lead to incorrect predictions, also known as “hallucinations”, “confabulations”, or more commonly, “bullshit”.
One of the more well-known cases is the hapless US lawyer who cited non-existent cases confected by ChatGPT in a court room and was fined when the fiction came to light.
AI also cannot handle humour – even Barrie concedes it can barely get beyond “Dad jokes”.
As an example, when I asked ChatGPT to write a humorous 100-word caption about bathing in hot springs in the Queensland outback, it came up with:
“Celebrating my 1-year milestone like a seasoned hot springs connoisseur in the Queensland Outback. The water was so warm, even the kangaroos were contemplating a spa day. As I relaxed, I overheard a wallaby whispering spa secrets to a cockatoo – turns out, eucalyptus leaves make excellent exfoliators. Meanwhile, a lizard lifeguard ensured everyone adhered to the ‘no diving, just diving into relaxation’ rule. Forget about Crocodile Dundee; I felt more like a spa kangaroo, unwinding in nature’s jacuzzi. If you ever visit, just beware of drop bears trying to sneak in for a soak – apparently, they’re closet hydrotherapy enthusiasts! Cheers to bubbly adventures!”
Even by its own admission – because I asked it this question specifically – generative AI acknowledges that there are several other realms “where human ingenuity remains irreplaceable”.
Where AI continues to struggle
AI cannot replicate real human creativity, empathy, curiosity, originality, authenticity, intuition, or imagination, says Professor Charlie Beckett, director of the JournalismAI Lab in London.
In 2019, The Lab published findings of a global survey into how AI is being used in more than 70 newsrooms.
Importantly, AI lacks capacity for critical thinking, analysis, or reason. And although it won’t divulge something so outrageous as how to turn your home kitchen into a meth lab, it lacks the ability to weigh subtler ethical considerations.
“It has no judgment, it has no understanding of truth, it has certainly no understanding of emotions, or values, or morality, or politics, or ideology,” says Beckett.
“My hope is, at least, that this tech can free up time for you to do what you do best.”
Which brings me back to the Amalfi Coast, and listicles, for a moment.
Readers seem to love them, but listicles (“10 things to do on the Amalfi Coast”) are one of my least preferred formats to write in. Honestly, I’d be happy to hand them over to the AI overlords when the technology improves. That’s one job they can have.
I’d much prefer to spend my time doing boots-on-the-ground reporting.
That means going on messy, sweaty adventures like the Stinson or the Packers’ Ghost Trek, venturing into laboratories or entertainment precincts to see what’s happening there, interviewing people to gain their views on such subjects as how women navigate the traditionally masculine domain of the boardroom or how young blokes forge a living in Outback Queensland, or even just reflecting on childhood experiences like learning to swim.
Incidentally, that latter story generated 55 comments – many of them emotional and not all of them publishable – before The Guardian Australia closed the comments section.
None of the listicles I’ve ever written have generated that much (if any) feedback.
That’s because people respond to real stories – stories that reveal something of the mind, heart and soul of the person who wrote them.
Rather than viewing AI as a threat, we can embrace it as a catalyst for letting go of the grunt work, like listicles, and onto more fulfilling stories, like the thumping first-person narratives and the revealing interviews and the investigative pieces that AI can’t master (at least, not yet).
Instead of lamenting the potential loss of jobs, let’s focus on adapting to this new technological landscape and harnessing it to serve us.
Which, if I understand correctly, was always its purpose in the first place.
Over to you
What have been your experiences of working with AI? Let me know in the comments.
Wanting to pitch your own work to editors? My FREE resource, Pitching for Publication, deconstructs three pitches which won assignments and ongoing relationships with editors in Australia and elsewhere. Hit the button below to download it now.
