Can you automate insight? That depends on how well you can teach robots to deal with human meaning.
Last week, I briefly touched on Automated Insights and their role in automating the Associated Press newsroom. Crack human reporter Alexander Jackson takes a deeper dive. Here’s Alexander:
Across the internet, articles plum full of useful data have begun to sprout up with the simple tagline: “This story was generated by Automated Insights.” Google brings back over 200,000 results from a search for that exact quote.
Notice that these stories are ‘generated,’ not written. That’s how a robot produces text.
That’s right. The robots over at Automated Insights (AI) are generating financial reports, fantasy football recaps, business intelligence analysis, and other documents that (at least in the writing) inspire mind-numbing boredom in humans.
AI uses a platform called Wordsmith, which is based on Natural Language Generation—shortened to NLG for us non-robots. As Automated Insights explains it on its blog, NLG users have to outline the format of their end product and feed “structured data” to the software. Then they can relax as it “automatically turns data into human-friendly prose.”
Let’s examine our language here. Potted meat products exist and are somewhat nourishing and better for your digestive system than a paper clip. That makes it human-friendly food. Organic, grass-fed beef is more human-friendly, but at least potted meat product won’t puncture your intestines.
This loose definition of “human-friendly” lets us attach the adjective to what the AI at Automated Insights churns out. Hence, AI’s automated articles are the potted meats of the journalism world.
What does all this mean for writers?
That may change very quickly though, so let’s look down the road. Are the digital children of these robots a threat to human writers? Ironically, you can find the answer to that question in the company’s name: “Automated Insights.”
Insight is the power to look into something and understand its nature. Useful insights produce new meaning by extracting the right data from one context and then fitting it into a new context.
For now at least, insight is a human property.
Can you automate insight? That depends on how well you can teach robots to deal with human meaning. Are mechanical factors, like profit and temperature, the only arenas people consider important? No. Human meaning wafts out of a complex potpourri of emotional responses, rational exploration, quests for risk or safety, irritation with the weather, thirst for praise, and so many other factors.
Until a robot can understand love (not just sex), stability (not just risk management), or enjoyment (not just distraction from pain and boredom), a robot cannot understand what humans find meaningful.
So Wordsmith does not offer insight because it cannot create new meaning. For now at least, insight is a human property.
However, while Wordsmith’s written production may be akin to a toddler dropping square blocks into square holes, at least these robots can turn data into tinny, information-heavy sentences. You didn’t want to do that anyways.
Now that your time has been freed from the drudgery of robotic work, go make insights! Look at the world long and hard. Tell us what you see and why it matters. After all, that’s what humans do best.