Generative AI is at the very top of the hype cycle right now. This means, according to the market research company Gartner that invented and branded the graphic representation technique, that generative AI is at the ‘peak of inflated expectations’ and is about to start falling into the ‘trough of disillusionment’. Gartner gives it two to five years to reach the ‘plateau of productivity’ via the ‘slope of enlightenment’ and I’d expect it to be the sooner end of that time span than the later.
I see a few signs that generative AI may already be tipping into the trough of disillusionment and in places perhaps starting to creep up the slope of enlightenment. At the Press Gazette Future of Media Technology Conference in London in September I sensed a split over the future impact of AI on publishing.
Publishers are worried that AI, in the hands of the tech giants, will eat what little is left of their lunch. Would it become just another way for Google to keep users on its own site instead of leaving for others? Some thought so. Could publishers license their original content to tech giants as AI fodder? Or could they block it? Is checks or cheques the way forward? Views differed. Sites can shut out Open AI but not the Google crawlers. The only option was to work with the AI platforms, thought one, while another thought the tools and the scraping that feeds them have to be regulated.
Publishers were sceptical of AI’s role in editorial. We heard from publishers who were already using it in some limited ways in the background: some for channelling content to users, for summations for storage or archiving content; others for translation and almost everyone transcription. For words, it could be helpful for the marketing of content, improving SEO or tweaking headlines.
Many were using it for research support, with one describing it as a super-librarian that’s brilliant at suggesting where to go for information. We also heard that it’s useful for hypothesis testing or scenario analysis because you can put any question to it for consideration, such as ‘what needs to happen for electric vehicles to be 100% of cars on the road by 2030?’ and you’d get an answer no matter how hypothetical the question.
However, in journalism itself, it tends to be limited to quite formulaic, repetitive articles, like sports results, election results or product reviews with specifications. By automating market reports, for example, Associated Press had multiplied its output from hundreds to thousands, leaving staff more time for source-based journalism.
The most strident publisher believed it is a tool that could close the gap between the low and high performers, whether it’s in writing words or code. They all thought journalists should be exploring what they can with AI, but they were quick to list the drawbacks.
The one million pound question
One panel was asked if you had a million pounds to invest, would you put it into AI or human staff? ‘Both’ was the general consensus. “AI can save time,” said one speaker, “but accuracy still trumps speed”. Everyone at the conference thought that human input was still necessary. Generative AI for journalism is still a ‘human-in-the-loop’ technology.
AI is not about to replace the entire journalist profession, we heard over and again from publishers. Perhaps not, but the technologies are developing at breakneck speed and many of its weaknesses will be quickly overcome, perhaps all of them in time. This logic was behind the alternative view from the conference, which came mainly from the suppliers of AI. You could argue they are biased of course but then equally they ought to know what they are talking about.
“AI is going to be a much bigger part of the equation in the future than people think in this room at the moment,” said one. There were plenty of AI publishing tools on show to improve workflows, marketing, data analytics, monetisation and more, with new ones appearing all the time for an ever greater range of automation and insights.
AI has its drawbacks for editorial but that won’t stop it appearing in updates to content management systems and other workflow tools within months.
Microsoft 365 Chat is launching to enterprise users on 1st November, the software giant has announced. AI will automatically draft texts in Word, answer questions about data sets in a spreadsheet, make a PowerPoint slide deck from a text document and I expect much more. It’s just the beginning. AI is here to stay. Publishers are already employing AI tools on the business side. Most editors and journalists have done little with it but tinkering so far, but that is about to change and fast. Will it start taking more creative jobs from humans? That’s further off but I wouldn’t rule it out, when we look at other kinds of AI in the pipeline to come after generative AI.
Several publishers pointed out that AI is no good at the primary sourced stories that’s the mark of strong journalism. Sure, there’s no AI that can meet a contact in a pub and persuade them to spill the beans on a secret scandal that rocks the government. But AI could, for example, be doing original primary source research with data, discovering patterns and spotting trends that humans would never have seen and turning them into new and surprising stories. Let’s hope so.
There is one scenario that frightens me though, and will chill to the bone anyone old enough to remember Windows 97 and its paperclip help character. Just imagine an artificially intelligent ‘Clippy’. Scary. Let’s pray that Microsoft doesn’t go that way again.
This article was first published in InPublishing magazine. If you would like to be added to the free mailing list to receive the magazine, please register here.