Automated Insights: Computers May Not Replace Human Writers, As Long As Perspectives ExistBy: Chris Crum - May 13, 2012
There are companies currently providing machine-generated content to various sites and businesses – articles created by machines rather than humans. Some wonder if the future of journalism is auto-generated content, and if journalists will be out of jobs because of it.
We’ve written about Narrative Science, one such company, several times. This week, we spoke with Robbie Allen, CEO of Automated Insights, another such company. Based on our conversation, it seems like humans will still have plenty of writing jobs for the foreseeable future.
One great point Allen made is that people will always have different perspectives on stories. It’s easy to say, “Uh oh, the robots are going to take our jobs,” but really, why should it be any different than if the AP or the New York Times or the Wall Street Journal puts out an article? You’re still going to get numerous other sites, blogs and social outlets with other people sharing their own thoughts and crafting their own stories. In fact, that’s what has made the Blogosphere such an interesting evolution of news.
Allen shared some similar thoughts in a conversation at a recent O’Reilly conference:
Of couse more content (auto-generated or not) means more competition, so we may still be competing with auto-content for readers. For that matter, the auto-generated content may be competing with other auto-generated content.
At this stage, there is clearly still a long way to go as far as auto-generated content being able to deliver on certain types of analysis and opinions. That’s why it tends to do better with data-driven topics, like sports and finance (though, from the sound of it, the technology is getting better).
Allen thinks in 50 years, we’ll look back and think it was funny that we ever had humans doing quantitative analysis. Computers are just better at number crunching, he says, but humans are better at qualitative analysis – another reason why human writers shouldn’t fear for their jobs too much just yet.
Still, as far a humans competing with computers goes, computers can get better, faster than humans can. Technology evolves much more quickly.
Automated Insights currently provides most of its work to clients on a B2B basis, and is not really focused so much on providing consumer-facing content. They do continue to operate the Statsheet Network, which Allen tells us, kind of serves as a showcase for what Automated Insights is capable of. People are often skeptical of how good auto-generated content can really be, so they provide this content where people can get a better idea and judge for themselves. You can see some of the work at CarolinaUpdate.com. Everything there is auto-generated. Here’s an example of how a piece of content from that site reads:
Top-seeded North Carolina advances past the second round with a 77-58 win over 16th-seeded Vermont in Greensboro. North Carolina opened up strong with a 37-25 halftime lead and maintained that pace to seal the win.
The scoring lead for the Tar Heels was shared by two players, including Tyler Zeller with 17 points while recording a double-double with 15 rebounds. The other scoring leader was James Michael McAdoo, who also had 17 points in 23 minutes. Additional effort came from Harrison Barnes who had 14 points and Kendall Marshall who had 11.
North Carolina found an edge in the haphazard play on the other side, claiming 20 takeaways.
Vermont finishes the season with a 24-12 overall record as well as a 13-3 record and second place regular season finish in the America East. The Catamounts made the NCAA Tournament by earning an automatic bid as America East champions, defeating Stony Brook, 51-43, in the conference tournament.
Vermont got its best effort from Sandro Carissimo, who registered 11 points. Other scoring efforts included 10 from Matt Glass, eight from Pat Bergmann, and seven from Clancy Rugg.
North Carolina will now square off against eighth-seeded Creighton in the third round on Sunday, March 18th.
The text is accompanied by various graphs and charts, as well as a list of game notes. Not bad for auto-generated.
To be accurate, though, there are humans involved in the process. Allen has a team, of course, and programmers and writers actually work together to make the output stronger.
While Narrative Science and Automated Insights may be early to this auto-generated content space, it would be somewhat surprising if we don’t see more entrants, though Allen implies that it’s a challenging space to enter, particularly because writers and programmers have to work together, which can be challenging.
While the company may be more focused on providing content on a B2B basis, there are both B2B and Consumer verticals they are currently tackling. These include: Life Sciences, Healthcare, Business Intelligence, Sales Productivity, Monitoring Solutions, Sports, Finance, Weather, Real Estate and Local Interest.
Local Interest, which I find to be a particularly interesting vertical for this kind of thing, could include things like real estate (down-to-the-neighborhood reports), and music (band info, songs, venues, etc.).
It will be interesting to see what other verticals Automated Insights and any other companies eventually move into with auto-generated content. I’m particularly interested to see how this type of content infiltrates the daily lives of consumers, as well as how search engines treat it.
Narrative Science thinks this kind of technology can solve the “filter bubble” problem. That’s another area of great interest that the news and media industry will continue to follow.