Why A Writer Built An A.I Writing Robot
It was June, 2019. Nearly twelve years had passed since the day I told the nice lady on the phone that I was a Copywriter with loads of talent, and looking for a job. I was a freshly-minted college graduate struggling to make ends meet and drowning in debt. And I had no idea what copywriting involved. But I called the office almost every day for two weeks until she relented and booked me for an interview. I got the job.
My three-month probation whizzed by, and over the next year I devoured every book on Marketing and Advertising I could find. I practiced my writing at nights, and gradually my confidence grew as my words began to appear in print, radio, and television beside the logos of some of the top brands in the market. I have the late great Ad Man Adrian Robinson to thank for that.
For a young man trying to make his way in a tough industry, it meant the world, knowing you had somebody who could guide you, talk you through it, and cut you a little slack when you were young and inexperienced. And as my responsibilities grew, he noticed my knack for thinking ahead and entrusted me to lead his agency’s Digital Marketing Unit. Writing was at the heart of everything we did, and I was in my element.
In many ways, it was even more challenging than writing for traditional formats because you had to be able to write for different platforms — websites, blogs, social media, email marketing, and so on — and you had to be able to do it quickly, to keep up with the fast-paced world of online marketing.
I loved the challenge of coming up with ideas that would make people buy things they didn’t need. But the hours became longer, the demands became greater, and I was working all the time. The stress started to take its toll. I was constantly exhausted and on the brink of burnout. Then it hit me that even though I enjoyed what I was doing, there simply weren’t enough hours in the day.
There had to be a better way.
The Writer’s Algorithm
And so it was, in March 2019, transitioning from agency to startup world, I stumbled across a string of research papers including a two-year old publication by Ashish Vaswani and company entitled “Attention is All You Need”. They proposed an invention called the Transformer. It was a simple, cheap, and efficient way to train a machine to complete various language tasks such as translation. It claimed to be better than any method before it, and its promise was to rain down success on businesses everywhere by automating their most tedious and laborious tasks, from customer service to transcription.
Transformers are a type of machine learning algorithm composed of a large number of interconnected processing nodes, which can recognize and mimic patterns in data. These models are trained using massive globs of text and images from the Internet. And can be used to generate original and coherent content from a simple text input-output interface.
I was fascinated.
It was the first time I had seen anything like this, and it blew my mind. I could suddenly see the potential for how AI could change the field of writing. I read and re-read the paper, and more like it, then I started to dream. I dreamt of a world where machines could not only generate text, but could also understand the human context in which it was being used. A world where AI could become a true creative partner in the writing process.
2019 turned out to be a pivotal year as several AI prototypes and apps were released into the wild. A year later, researchers at Open AI released the third iteration of their transformer- the Generative Pre-owned Transformer (GPT- 3). At the time, it was the largest and most efficient model of its kind. The Guardian Newspaper even published an article titled: A robot wrote this entire article. Are you scared yet, human? as this writing machine took on a hype of its own.
A Moral Dilemma — Playing ‘Fast & Loose’ with the Future?
Naturally, with any new technology comes a set of moral dilemmas. As humans, we are used to the privilege of being in control and making our own creative decisions. But what happens when machines begin to take on some of these roles?
I had mixed feelings about all of this. On the one hand, I was genuinely excited about the possibilities that AI could bring. I believed it could help us to create better, more efficient, and more insightful content. Yet on the other hand, I was worried about the impact that it might have on professionals who had spent their entire lives mastering the craft.
I couldn’t shake the sense of excitement and fear that coursed through my veins and I realized that this was all too familiar. The feeling reminded me of the first time I heard about Facebook and how it could potentially render traditional media irrelevant.
I determined that the best way to calm my nerves was to lean in and learn more. So I did the boldest thing I could think of and built an AI Writing App of my own. I called it “Guru”. It had an easy-to-use interface that allowed you to input text — a headline, and the name of your favorite publication. The AI would try to mimic the writing style of the publication. It could generate text on any topic you could throw at it. Sometimes it was weirdly accurate, and at other times hilariously ridiculous.
It was a way for writers to generate ideas more efficiently without having to fiddle around for hours trying to get started. I shared the app with colleagues in my circle, and most flat-out refused to entertain the idea. They thought I was playing fast and loose with the future of our industry. But I pushed back.
See, I know all too well the arduous hours of study and practice it takes to become any good. Nothing compares to the divine magic of finding the big idea. And nothing can truly substitute for the human touch. But I also remember when my best efforts were undone by clients who just didn’t get it, or by teams that changed their minds at the last minute.
How many hours would I have saved if I had a machine to help me out?
How many other professionals could this technology possibly help?
Birth of a Writing Machine
It was at this point that I felt compelled to take the initiative and get out of my own circle. I spoke with PR folks who were facing burnout due to the heavy demands on their time. I spoke with Communication Heads who were desperate to find ways to automate repetitive, time-consuming writing tasks. Even the talented content writers with whom I spoke agreed that the demand is insatiable.
I grew more and more excited — not only because I heard other people say the same thing across professions, but because I no longer felt alone. I realized that the key to the whole thing was to see AI as a tool to help with the creative process, not as a replacement for the human touch.
In the past year, a number of startups have entered the AI-writing space — clearly, the need is there. As I continued along the path, I founded a startup called HelloScribe. We are building AI-powered writing tools for communicators in the PR & Marketing industries.
In the next article, I’ll share more on the use cases for AI writing and some thoughts on the ethical considerations around building and using such tools.
Stay tuned.
