90% of stories will be written by artificial intelligence by 2034 — am I a robot, looks at what’s happening now and if we can tell if the stories we read are written by a robot right now | Towards AI

90% of stories will be written by artificial intelligence by 2034 — Automation and storytelling – did a robot write this?

Paul Wilshaw
Apr 30 · 6 min read

Automation and storytelling — 90% of stories will be written by a bot by 2034:

Automated and personalized news Gif (Wilshaw, 2019)

How do you know what’s real, what’s fake and what’s a bot (an endearing generalized term used to describe automated content) or not? Content generating bots are given a set of parameters and data inputs to output interesting and relevant news into the ever-bulging news feed in your favorite information gateway.

90 percent of our news is automated news — at least by 2034 that is. Giant media conglomerates are already automating a lot of data (Levy et al., 2019). Algorithms are already writing thousands of stories, in fact, one news agency published around 40,000 automated news stories last year (Delcker, 2019). In a move to keep relevant traditional media companies are digital transforming their workforce from analog newspapers to cutting edge digital companies. No longer can journalists keep up with the tsunami of “big data” (Mayer-Schönberger and Cukier 2013) generated from a growing digital population.

In the world of data, humans create millions, maybe more, data points every second. Our beloved and addictive technology is capturing our every move and interaction and sending it the manufactures of these devices and applications. Things like the Apple Watch, Fitbit or Amazon Echo know precisely where we are and what we’re doing, probably better than we know ourselves. This ever available data is a minefield for ethics, privacy, and security. On a positive note, it’s overwhelmingly incomprehensible by us mere humans and can, if used correctly, aid our life — from reporting health issues early, motivating us to exercise more, monitoring our heart — helping us to live longer and healthier. This data forms patterns, much of which is difficult to sift through unless we’re metahumans like The Flash. Some data is more accessible to automate than others too. Stories, such as finance, sports, technology releases, and scientific news have easy to read data and therefore, more straightforward to convert into natural language and standardized narrative structures — take an Apple release keynote; all different in some way but they all have common narrative structures (Delcker, 2019). Some senior person stands on the stage, builds anticipation and announces new software features or hardware. Show some sexy images of their new product, proclaim how improved it is over the previous version — audiences clap with rapture and wait with wonderment until they can purchase it. Just like a novel this product release has a start, setting the scene and introducing the protagonists. A middle, introducing excitement and anticipation and an end — the conclusion of what went before.

Artificial intelligence (AI) in journalism isn’t about replacing jobs; it’s more about doing the things we humans are pretty bad at — like processing data from millions of sources and verifying them. Humans get bored, make mistakes and have to eat, sleep and socialize (or play Fortnite with thousands of other online players) — all this makes it arduous for us to process even a tiny amount of data available to create trustworthy journalism. Perhaps it’s better to think of AI journalism like Iron Man’s suit of armor, augmenting storytellers to become data processing superheroes.

Iron Man Gif (Wilshaw, 2019)

There’s a lot of “smoke and mirrors” around the personalization of news stories — most of which is through monitoring users likes, views, and even eye gaze over time — then turning that information into a more tailored news feed — created using something called machine learning (ML) and natural language processing (NLP). We teach a program what we like and don’t like. AI can, according to Narrative Science founder Robbie Allen, allow the user to customize not only their content and localization but how it’s told — writing personalized and relevant stories for a single audience member. Previously this would have been too costly and unfeasible (Levy et al., 2019) and presented as the traditional one-to-many journalism that existed for hundreds of years before the ever-presence always online smartphone.

Automating the basic task of opening and navigating to my specified website (Wilshaw, 2019)

To test the automation of publishing stories to a social media platform, I created an automated program with Blue Prism, an RPA (robotic process automation) client — whom I work for as Head of UX and UI (to be transparent). I aimed to “scrape” stories from one website (in this example https://theconversation.com/uk) and repurpose it in my Twitter feed through the no-code, but complex, interface; I already do this to some extent using IFTTT, https://ifttt.com, a service for creating interactions between websites, devices, and platforms but find it limiting to the pre-propositioned applets published. I wanted to push automation more, by fusing different sources of my choosing to create a “super” tweet published at a time of my choosing. This part of Blue Prism works by “spying” any website or program and can then interact with it either collecting data or by entering data, clicking buttons, navigating to different pages and then re-using that data in a new format, quoting my sources as appropriate.

My experiment was partly a success; I still need to do a bit more configuration, refinement, and testing but it works, and with a little more synthesis I can start adding natural language processing skills and sentiment analysis to my Tweets and then automating engagement and tailoring content to my audiences.

Mapping my automation process, navigating, finding the relevant data and processing it. (Wilshaw, 2019)

Automation is currently in its infancy often with complex interactions and highly skilled development knowledge to create anything of worth and would pass for a skilled human storyteller. It’s gaining momentum, and the entry level is getting lower — so am I a robot?

No!

I sweated, researched and cherished every word I typed, re-typed, deleted and then quickly pressed ⌘z.

Did I have the help of a robot? You bet — call me Iron Man!

Sci-comms bot 2000 (Wilshaw, 2019)

References

Delcker, J. (2019). This story was not written by a robot. [online] POLITICO. Available at: https://www.politico.eu/article/robot-reporters-newsroom-algorithms-artificial-intelligence/ [Accessed 30 Apr. 2019].

Levy, S., Lapowsky, I., Simon, M., Martineau, P., Baker-Whitcomb, A. and Oberhaus, D. (2019). Can an Algorithm Write a Better News Story Than a Human Reporter?. [online] WIRED. Available at: https://www.wired.com/2012/04/can-an-algorithm-write-a-better-news-story-than-a-human-reporter/ [Accessed 30 Apr. 2019].

Mayer-Schönberger, V., Cukier, K. and Iriarte, A. (2013). Big data. Madrid: Turner.

Wilshaw, P. (2019). Blue Prism. Salford: Blue Prism.

Wilshaw, P. (2019). Iron Man Gif. [image] Available at: https://www.paulwilshaw.co.uk [Accessed 30 Apr. 2019].

Wilshaw, P. (2019). Personalized news Gif. [image] Available at: https://www.paulwilshaw.co.uk [Accessed 30 Apr. 2019].

Wilshaw, P. (2019). Sci-Comms Bot Gif. [image] Available at: https://www.paulwilshaw.co.uk [Accessed 30 Apr. 2019].

Towards AI

Towards AI, is the world’s fastest-growing AI community for learning, programming, building and implementing AI.

Paul Wilshaw

Written by

Automation expert. Innovator, Science communicator and future media experimenter. Creator of craziness. Multiple award winner in IA / UX / UI / ID / CX

Towards AI

Towards AI, is the world’s fastest-growing AI community for learning, programming, building and implementing AI.