Last November, I brought up the concept of automated news. When I presented the piece, what I showed you was slapdash, barely even intelligible combinations of words that vaguely resembled news articles, things that were created purely to generate hits.
This has been identified as a problem. However, the problem being addressed isn't that computers are generating news articles. The problem being addressed is that they're doing it badly.
The Big Ten Network is currently using a service called Narrative Science, which takes the facts and figures from, in their case, a football game- the New York Times here gives a sample from Wisconsin/UNLV- and selects phrases and word choices from a database appropriate to the circumstances. If you've ever played a sports video game with commentators, Madden for example, it's kind of like what they do there to make the commentary sound realistic, but more sophisticated. Entire game reports can be generated within seconds.
The thing is, though, Narrative Science wants to expand the concept beyond sports. They figure the basic structure can be plugged into any other field heavy on statistics; another company is using the technology for housing markets. And they want to go further than that. Co-creator Kris Hammond predicts that the technology will be able to compete for a Pulitzer Prize within five years.
Which, of course, raises a very real concern for any human journalist. The creators present it as a supplemental tool for existing journalists. However, in a penny-pinched industry, the allure of an attention-grabbing computer that can write just like a human, and for a lot less money, might just cause the media organizations to replace those human journalists instead.
On the other hand, someone has to go obtain the data before it can be crunched and spit out by Narrative Science. And it is still humans that are making news. We still don't have anything that's passed the Turing Test yet. Human journalists will still have to go gather information that isn't being given up willingly. Humans still have to conduct interviews. It will still take a human to go into a warzone or disaster-ravaged region and be able to express what it is they see, smell, or experience. Opinions are still uniquely human, at least until Narrative Science creates something self-aware.
Sportswriting, game recaps especially, are actually pretty easy to automate when you think about it. There are only so many types of things that can happen in any given game. Each game follows a set structure. Numbers are everywhere. Unless you are at the very top of the class, there are only so many ways to describe this limited number of possible events before a lot of sportswriters have to start reusing words and terms. Sooner or later, yeah, it starts to seem conveyor-belty. Same goes for things like Treasury rates, housing markets, and most of the other stuff you see in the financial section.
The more complicated topics, however, the ones that aren't so stat-heavy, that require opinions, experiences, human contact (e.g. interviews, debates) and Freedom of Information Act requests? That's a tall ask. Even in the sports section, while some of the stories are formulaic game reports, others are game preview analysis, off-the-field happenings and good old-fashioned arguing, which is really half the fun of sports when you think about it.
Even if Narrative Science manages to somehow pull it off, though, it's not like it's going to stop people from commenting on the day's events anyway.
It's less a danger to journalism (unless we let it be one), really, than it is a sad commentary on how boring it has become to read recaps of what is, in the end, a form of entertainment.