It’s the early-90s. Our hack has just entered the office, more than a little hungover (carousing for scoops is part of the job and it can’t be done from behind a desk). They’re probably wearing faded denim, shoes adrift of their natural partner – the suit, and the obligatory shoulder-padded jacket. Then someone “techy” whispers the word ‘internet’. And media changed forever.
The internet ushered in some major, now obvious, changes. Some of which were: major job losses in print media (as money-making classifieds became all but obsolete). The belief that content should be free. The democratization of access to news and, as an offshoot, the polarization of that news, catalyzed by an increase in alternative news sources able to reach audiences via aggregators like Google and Facebook.
Now, more or less 30 years in, with the internet an immovable part of modern life, the question of how to monetize journalism, as its survival mode, is still a pertinent one. And the answer is inevitably bound up with AI in more ways than one.
It helps grow audiences and monetizes them
Advertising enables publications to get articles in front of potential new audiences and to monetize when these audiences are big enough. And it’s at this intersection that journalism and AI are deeply entwined. Mike Kaput, of the Marketing AI Institute, in his blog post entitled Artificial intelligence in advertising says:
“AI is critical to the infrastructure that underlies advertising products on many platforms, though you may not always see it. Modern programmatic platforms often use AI to manage real-time ad buying, selling, and placement.”
Basically, the platforms that offer advertising – Google, Facebook, Instagram and digital ad exchanges (see the link to understand what the hell they are) – use AI to manage the purchase and sale of ads at scale in real-time. So, essentially, it’s AI that decides, as Kaput puts it: “how ad spend gets used, who sees ads, and how effective overall campaigns are.” And in the world of journalism when giving content away for free – as is still the rule rather than the exception, despite a shift towards paywalls and subscription models – the effectiveness of advertising is critical, both when trying to grow an audience and when selling ad space. Let’s take the New Yorker as an example. It makes money from subscriptions and ad revenue. The two are reliant on each other and both are reliant on AI: the publication uses digital channels to reach new consumers in the form of paid posts and paid search keywords and a large audience allows it to sell lucrative ad space. That ad space is served programmatically. And so the cycle continues.
As well as managing the economics of digital advertising, AI also processes customer data from multiple sources at scale which allows for hyper-targeted advertising based on preferences. Ultimately, this gets people to pay attention to ads and take action off the back of them by improving relevance in the most cost-effective, resource-lite way.
It has replaced reporters and editorial staff
If institutions, like MSN and Associated Press, can lay off, or free up, reporters by using AI that can do the same job – the former to pick news and content that’s presented on MSN.com and the latter to report companies’ quarterly reports – they’re improving their bottom line. If Phrasee can write a subject line or an advert that results in more clicks than one authored by a marketer, it makes sense to go with the automated version. No more need be said, really. Except that it might feel like the dawn of an age everyone dreaded; an age where machines overtake humans. AI engineers everywhere eyeroll, so far from the truth as that currently is. While there are pockets of automation, and admittedly the idea that humans might imminently become jobless is entertaining-scary, it far from reflects the current state of AI, which improves human efficiency rather than replaces the need for the species altogether. Which brings us onto the next point...
It frees journalists to do deep analysis
AI has undoubtedly changed journalism, but it’s not replacing the need for human brains and in particular their ability to make complex emotional connections and judgments that AI cannot...yet. Not at scale anyway. Lots of news organisations use versions of the aforementioned technology to automate aspects of reporting – The BBC has Juicer, the Washington Post has Heliograf, and nearly a third of the content published by Bloomberg is generated by a system called Cyborg – but they can’t use these systems to write pieces that require imagination or complex analysis. As this LSE study describes, “Even the newsrooms [...] that are furthest ahead in the adoption of AI described it as additional, supplementary and catalytic, not yet transformational.” AI may power distribution and automate elements of reported content but, far from pit humans against machines, it highlights the value of human insight and lets humans focus on doing just that.
It can facilitate or prevent the spread of fake news
When it comes to AI taking over basic reporting or the curation of news articles, the careless use of data or machine learning that’s trained on bias data can result in inaccuracies, distortion and discrimination. As happened with Microsoft’s AI journalists mis-identifying mixed-race members of Little Mix last year. Personalization, too, when done inadvertently ignorantly or intentionally so for a particular outcome, can result in confirmation bias – whereby the content served is based on data that fans or confirms assumptions held by the viewer. Otherwise known as fake news. On the flipside, just in the same way AI can be trained on bias data, if it’s trained on impartial data (which, it is important to note, doesn’t really exist because the world is full of bias) or rather data that is thoughtfully handled, it can reduce the chances of misinformation.
The changes brought about by AI in this piece, fall into two categories, as laid out by Sebastian Raisch and Sebastian Krakowski in their paper AI and Management – automation and augmentation. If Associated Press can use AI, as it does, to produce corporate earnings reports, that frees up its journalists to do higher value work that requires complex storytelling, fuelled by curiosity. This is automation and augmentation working in tandem. And it’s as a result of this synchronicity, according to Raisch and Krakowski, that AI does its best work – eliminating tension and optimizing different skillsets.
As for our early 90s journo – they look a little different today. The shoulder pads are goners; switched for a jaunty bob hat and Pategonia. And the hangover has been replaced with a speciality coffee. The internet also means immediacy is now king. Reporting on something first for our 20s journo wins them valuable clicks. And in a game where turnaround time for dull work can be entrusted to AI, that is a welcome addition. Still, there’s a way to go. There remains only a few people in newsrooms who know the ins and outs of how AI can support their work. So change is likely to continue to be a slow burner, or incremental, for some time to come.