Slate had an interesting piece the other day about “robot” journalism, doubtless prompted by the AP’s recent announcement that it would be turning to software from Automated Insights to produce thousands of corporate earnings stories. (Business news organizations like Reuters, Forbes, and probably Bloomberg have been automating stories as well, but this underscores how mainstream the technology and practice is becoming.)
The Slate piece was a smart look at the advantages and disadvantages that humans and machines bring to the exercise of producing stories, and it summarizes the state of play well. Machines are faster, more reliable, better at crunching reams of data, and all that. Humans are better at bringing in broader context, understanding what might be interesting to other humans, and so on. Machines are critically handicapped by the quality of the data available to them. Machines are cheaper than humans. All of which is true – or mostly true, anyway. (Machines can be very expensive to run.)
Slate’s conclusion? Humans are better at producing journalism, as the quote above notes.
Well, sure – but is that the right question?
Horses are more empathetic than cars. They won’t – by and large – go off a cliff by themselves if you take your eye off the road or doze off. They’re superior at jumping over obstacles. They don’t need messy fossil fuels to run. In short, horses are better than cars at providing horse-type locomotion. But as a method of mass transportation, they suck.
So – yes, humans are better at producing human-type journalism, and God knows, we need much more of that. But is it the only thing we need, and what are machines better at producing that the world needs?
I don’t mean to suggest that we get rid of human journalists and replace them with machine-writing software. Nor am I suggesting that the Slate piece is off-base. But I do think we need to stop thinking purely in terms of the kinds of journalism we produce – or like to believe we produce – and think more broadly about the information needs of society and how technologies like language generation can help meet them.
It’s true, as NY magazine’s Kevin Roose notes in another recent piece about robot journalism, that:
The stories that today’s robots can write are, frankly, the kinds of stories that humans hate writing anyway.
But to a large extent, that’s because that’s how we’ve defined their mission so far. We turn to machines to do tasks we’d rather not do, and as a result, sometimes push them to do tasks they’re less than optimized for. It might be more useful to look at what they do well, and how that might mesh with the goal of better informing society. As Slate notes:
Software programs can scan a spreadsheet in a fraction of a second. They can rapidly compare the numbers in that spreadsheet to all sorts of other data sets, from past earnings reports to historical averages to the recent performance of other companies in the relevant industry sector. Over time, this ability should allow them to spot trends that human reporters would overlook. For instance, (Automated Insight’s CEO Robbie) Allen says, Automated Insights’ software might notice that a given company’s revenues have beaten estimates every third quarter for the past five years. “Certain kinds of analysis are completely missed all the time today because humans aren’t ideally suited to that kind of thing.”
More importantly, those capabilities can create different types of stories from the current one-to-many, “push” model journalism that humans are capable of. Machines can already produce personalized reports at scale; they could in theory, at least, publish “news on demand,” pulling data on the fly to answer specific questions people have when they have them, much the way Siri or its Android counterpart give you useful responses when you ask whether you need an umbrella tomorrow; or help you understand what the key factors are in how a result came to be, as Narrative Science‘s Larry Birnbaum explained at a panel I moderated.
In other words, not mimic humans, but do things that humans couldn’t possibly do. Not be second-rate people, but first-rate machines.
To reiterate: This isn’t to suggest that human-created journalism isn’t needed, or that machines aren’t a good adjunct to human newsrooms. Indeed, integrating humans and machines in a “cybernetic newsrooms” is one great possible outcome, and one we need to pursue more aggressively.
But we can and should also go beyond that, and approach the coming age of machine-generated language capabilities from the point of view of what society needs, and not so much from the notion of how they stack up against human abilities.