Science, technology, and all things awesome with Ian Steadman

RSS

The New Statesman could eventually be written by a computer - would you care?

Editors can choose to give algorithms control over parts of the editorial process - and while it's early days now, it's worth considering what that means for how we value writing.

Robots: first better dancers than humans, one day better writers? (Photo: Getty)

Would you want to read a newspaper that was written by robots?

Maybe that’s not the right question. Maybe it’s better to ask whether you’d know you were reading a newspaper that written by robots. That’s an important distinction.

The Guardian has performed a little experiment this last month with a newspaper that was semi-curated by an algorithm, called the Long Good Read. Here’s Justin Ellis from the Niemen Journalism Lab explaining how it works:

The Long Good Read began life several years ago as a digital-only experiment from former Guardian developer Dan Catt. The idea was to harvest the paper’s feature pieces and longer stories into a stream of articles best meant for RSS or a read-it-later queue. These were the stories that lent themselves to dedicated reading time, that quiet moment after work or a lazy Saturday morning. That, Kiss said, also fits the description of print: “It’s part of a noble heritage: people wanting something to read when they’re drinking their coffee or tea.”

Catt built an algorithm that scans The Guardian’s API, stripping away blog posts, multimedia, and other pieces in favor of articles over a certain length.

The robot does the legwork, leaving an editor to pick and choose what stories work for the edition before handing the process off to a different robot. In this case, it’s The Newspaper Club’s ARTHR tool, a layout program that lets people feed in content from different sources, either links or individual text and images. Tom Taylor, head of engineering for The Newspaper Club, said they use a semi-automated version of ARTHR for the Long Good Read, which allows an editor to enter story links and lets the program develop the layout on its own.

The final edition of the paper goes out to the Guardian’s quaint coffee shop in Shoreditch, east London, for people to peruse up if they so wish.

This isn’t a proper robot-written newspaper, as the individual pieces are still authored by fleshy humans, but it remains an early example of what will happen to fields like journalism, or accountancy, or architecture, or even art. Those things that we think of as “human” for their expressive creativity will be replaced by machine mimicry of a quality that we cannot complain about, and at a price that we can’t resist, in exactly the same process that affected working class jobs.

Another example: here’s the match report for Crystal Palace against West Ham last night, from the BBC. At the top of the page is a journalist's report, but at the bottom of the page there’s live text commentary provided by the sports statistics company Opta. Its staff visit games and transcribe the action, live, uploading it to a server where it is then distributed to clients around the world.

Those live commentaries convert the fluidity and chaos of a match into a set number of expressions and phrases. It makes something digital out of something analogue. There's no reason that a computer - one that’s trained in understanding how a football match works - can't do exactly the same job one day, turning recorded footage of a game into statistics.

This isn’t science fiction - it’s been happening in sports like baseball (which, with its more limited range of events, is easier for robots to understand) for a few years now. There was a weird case in 2011 where a machine managed to write “better” copy than a human journalist when reporting on the same baseball game. Companies like Narrative Science in the US have made huge leaps in developing algorithms that suck in individuals facts on one end and spit out properly-written wire copy on the other, while in the UK, it’s possible to see Summly as an obverse - turning lengthy, wordy human writing into something more concise.

In this techno-utopian future, the media will automatically compensate for spin. The journalists of the BBC, the Associated Press, Reuters, and so on, will be replaced with teams of developers that compete with each other to compile the perfect, passionless, disengaged robot writer.

It sounds a bit colourless, but it’s not as if newspapers haven’t already realised that there’s value in hiring outspoken columnists instead of news reporters. Robots can’t do investigative reporting, yet, nor can they explain to us why we should care about things like football. It’s hard to imagine the New Statesman being written by a machine, because we quite like to be opinionated here... and yet, well, there's no theoretical reason that there can’t be a Turing Test for opinion.

If a computer makes us care about something, would we care any less that it was a computer, not a human, that wrote it, or composed it, or built it, or created it? We can’t rule that out.

McKinsey & Company released a report this year detailing the global impact we can expect between now and 2025 from what it perceives as the 12 most disruptive emerging technologies. Mobile internet led the way - it’s likely to create add as much as $25 trillion to global GDP - but in second place was automation of knowledge work, something that could simultaneously add $9 trillion to global GDP while making the jobs of nine percent of the world’s population redundant. Not too shabby for a robot-curated newspaper in a Shoreditch coffee shop.