If you’re still part of the dwindling local newspaper readership, there are probably way more AI-generated articles than you realise…
‘Written by a Robot’ is a colourful way of saying ‘generated by an AI algorithm’.
The main two Robot giveaways are Style and Subject.
Robot Style in Local News
As you scan an article, do you sometimes find your eyelids dipping, impatient to pick out the relevant numbers, instead of lingering on delightful turns of phrase or local features?
If so, chances are it was written by a robot.
It will be as dull as ditchwater.
As bland as blancmange.
All the cut-through of a cookie-cutter.
As lively as a lecherous Lothario off his lithium.
Robot journalists can’t yet reliably generate orotund phrases in context. But it won’t be long before they can.
A robot writer is unlikely to use the word ‘orotund’. If it did, a robot editor would probably replace it with a more common word. ‘Fancy’, maybe. Or ‘florid’.
In this particular case, you and other human editors may agree. But what if all local newspapers end up using the same algorithm?
Would any robot generate a headline like ‘Panic as Man Burns Crumpets’? That’s the title of a bittersweet memoir recently published by British local news feature writer Roger Lytollis.
The book lays out with painful clarity the transition of the Cumberland News from shining light of local journalism to anodyne advertising platform.
The tipping point was in 2018, when it was taken over by corporate conglomerate Newsquest Media Group Ltd.
Newsquest is owned by Gannett, the USA’s #2 local newspaper agglomerator – here’s a list of the hundreds of local titles they’ve gobbled up there. Gannett is owned by New York private equity group New Media Investment Group. NMIG also owns GateHouse Media, LLC, America’s #1 local newspaper agglomerator.
Such corporations are working with AI developers on robot editors to adjudicate on such matters as using ‘orotund’ or ‘fancy’.
What do you think? How precisely can you draw the line between ‘colourful’ and ‘pompous’?
In isolation, replacing ‘orotund’ with ‘fancy’ may be a good editorial decision in this case, but that’s not the point.
As it improves, AI will get much better at both writing and editing. Robot-spotting will get trickier. Gannett newspapers are already full of robot articles, automating the low-hanging fruit of replacing finance and sports reporters. Why? Money.
- Slapdash finance and sports hacks working under time pressure tend to paste numbers/scoresheets together with plausible, grammatical sentences.
- Expensive, ethical, responsible finance and sports reporters selectively use carefully contextualised financial data, or sports stats, to illustrate or support their broader analysis.
Robots are already pretty good at the former. AI boffins are working hard at the latter.
At the cutting edge, GPT-3 developers have opened up the possibility of unleashing such a mega-tsunami of plausible robot-written content, they’re worried it might break the internet.
GPT-3’s developers seem to be trying to keep their genie in the bottle, but the past few thousand years of human history don’t provide many successful examples of controlling technological progress.
Some computational linguists, unimpressed by the ‘intelligence’ displayed by GPT-3’s Big Data form of AI, have dubbed it no more than ‘stochastic parroting’.
GPT-3’s AI credentials may be as irrelevant as its creators efforts to contain it. A few years back a couple of exotic pet parrots escaped their London cages, and now huge flocks of parrots can be seen in the city’s parks. And they’re only real world parrots.
For the moment, Gannett’s bean-counters, and their software engineer sidekicks, appear to have no qualms about the prospect of unleashing a tsunami of poor quality content on the internet.
They’re too busy embracing this exciting new cost-cutting strategy – or as they might put it:
downstream-distributed localised grammatical text incorporating geographically-defined data subsets derived from larger universal datasets with customised geographical fields correlating to the geographical scope of the tasked masthead.
Maybe their robo-editors will replace that with ‘robot-written content’, in the millions of robo-articles with which they flood across the internet..
Robot Subject Matter in Local News
For Robot- Spotters, the other giveaway is subject matter.
As Robot articles harvest local data from national datasets, they’re quite easy to spot once you know what to look for.
Any article in your local rag about trends in house prices, crime statistics, or, recently, Covid stats, are probably written by Robot.
If you’ve ever used mail merge features, or dabbled with macros, you’ll get the idea. When you see a headline like:
Local Covid Cases Rocketing
Then imagine something like:
‘[INSERT GEOLOCATION HERE] Covid Cases [IF UPTREND THEN INSERT CHAR STRING =’Rocketing’][IF DEVIATION WITHIN RANGE x THEN INSERT CHAR STRING= ‘ Plateauing’][IF DOWNTREND THEN INSERT CHAR STRING=’Plunging’)
Massive agglomerators, like Gannett, are working with AI developers to cut costs through automation.
This applies to both the ‘curation’ role that used to be filled by editors in selecting relevant stories for an individual reader, and to news articles themselves.
Replace an editor with a robot, and you cut one human salary. Replace an entire class of reporters, and you could cut hundreds, thousands for the bigger conglomerates of human salaries.
Clearly, this is of interest to the bean-counting hedge funds, now the ultimate owners of so many local titles. Ignoring such cost-cutting potential would be a dereliction of duty to their investors.
When it comes to replacing reporters with robots, the generation of sports and finance reporting is just the beginning.
They’re basically numbers strung together with grammatical sentences, which current AI is getting quite good at. More topics will follow.
The push behind the automation of journalism is obvious. If you can’t raise revenue, you have to cut costs.
Robots are much cheaper than humans. They work 24/7, don’t get sick or unionise, and are much easier to manage.
The prose they generate is currently quite easy to spot as it is utterly bereft of vivacity, colour, elan, piquancy, local flavour, or any sense of community/landscape.
Ignore The Byline
The byline is the name of the reporter at the top of the article following the word ‘by’, usually accompanied by a smiling mug shot .
A byline was once valuable journalistic currency. It meant it was Proper Local News. Real reporters employed by the paper got them, agency reports didn’t.
For sub-editors writing in their free-time, with dreams of becoming reporters, freelancers looking for their big break, or new recruits desperate to make their name, their first ‘byline’ was a big big deal.
Post-internet, this changed. Knowing that a byline implied authenticity, papers began to plaster them on any old article to boost credibility.
As advertorials stopped being labelled as such, and advertising began its creeping dominance over journalism, this currency became further devalued.
For a while, a byline retained its basic, most functional meaning – that it was written ‘by’ someone. That someone, increasingly, may have been a corporate PR, but at least they were human. No longer.
Don’t be fooled by the appearance of a byline and smiling photo adorning the top of a robot-written article.
As we demonstrate in our pilot Review Group, conglomerates like Gannett no longer have any compunction about pretending robot-written articles were written by humans.
This must be pretty humiliating for the human involved, who’s trained for years as a professional writer.
Assuming, of course, the person represented by the byline and photograph actually exists.
These are the kind of tips we offer in our See Through News Newspaper Review Project, a public media literacy project to raise awareness of the under-reported concentration of media ownership in the local newspaper sector.