SocraticGadfly: Are journalists about to meet their machine overlords?

March 17, 2015

Are journalists about to meet their machine overlords?



According to the New York Times, it's possible.

Here's the key graf:
Kristian Hammond, Narrative Science’s co-founder, estimates that 90 percent of news could be algorithmically generated by the mid-2020s, much of it without human intervention. If this projection is anywhere near accurate, we’re on a slippery slope.
And, if a bot can already write the lede graf in a pro sports story well enough to fool most humans, we're past the days of police blotter and quarterly financials. And, yes, there's an example of a bot doing just that on a sports story, about eight grafs about that pull quote. (And if that one example isn't enough, there's a quiz at the bottom of that link. I got five of eight correct, double crossing myself on one, but getting a gimme on another, per discussion in the story.)

Now, there will still be room for some time for sports columnists, but beat writers? You could be in trouble. Ditto for news writers.

Because of cost, this is less likely to trickle down to the community level by 2025. Hyperlocal bots specialized to write about the police blotter and real estate deeds are one thing; general purpose bots writing about a contentious city council meeting are another.

Among other things, the bots have to be fed information, or, shown where to feed themselves in a controlled environment, if the blotter and real estate records are online. If a city secretary is slow in posting meeting minutes, at least the version she transcribes off audiotape, a human's going to knock that out. Ditto if a high school coach is slow on posting game stats, as far as sports-writing bots at the community newspaper level.

But, community newspaper editors and writers shouldn't sleep too easily, especially in suburban or exurban vs. rural areas.

One editor who knows the lay of the land on the set of suburbs or exurbs, assuming a company has a group of such newspapers, can polish up the bots' writings easily enough, one would think.

And, one would think that one editor would be expected to do so by corporate hierarchy.

Now, Narrative Science has $everal rea$on$ to push its alleged skill level. So, take some of its claims with one or more grains of salt. At the same time, though, if a loose version of Moore's Law applies here, the mid-2020s are a decade off and a lot of things could happen. Five years ago, idea of bots putting a police blotter into coherent form probably would have been considered laughable, let alone them writing a passable sports story or cheap poetry contest submission level items, per that quiz.

If not by the mid-2020s, then by the early 2030s, computers will be doing ever more of this. And other things. As computers continue to replace paralegals, and even junior lawyers, at larger law firms, with more and more of the discovery process, especially in civil cases, being done by computer, computers will do more and more "investigative journalism." The more information that gets deposited online, the more the journalism about it will be done by computers.

If you want percentages on my bit more skeptical take, let's say 50 percent of news could be bot-generated by mid-2020s and 80 percent by early 2030s.

Combine this with digital ad sales, and, kiddos, don't study journalism in college! That goes for radio, TV and web-based media, too. No, we don't have computers who can read the news on TV or radio, or shoot video yet. But, that's "yet." Why couldn't a Watson of IBM fame be given a humanoid voice for radio, or even a humanoid voice plus skin for  TV?

Hell, we could remove half of such a Watson's "brains," put a blond wig on it, and it would be about right for Faux News.

More seriously, though, the shakeout from bot-writing is likely only going to accelerate, Narrative Science's marketing hype aside.

That said, the PR person who wrote Narrative Science's press release? Don't get all schadenfreudey; if the bot can replace a journalist, it can replace you.

No comments: