NLP is Our FORTRAN
A person writing a math equation on a whiteboard.
Remember the movie Hidden Figures?
The whole movie is phenomenal. One part in particular sticks with me. Dorothy Vaughan (played by Octavia Spencer) watches the state-of-the-art, room-sized IBM mainframe being delivered to her workplace at NASA. Dorothy, who supervises of a team of human computers (as they're called in the '60s), instantly realizes their world will change.
She starts to learn how to program the mainframe in FORTRAN. Not only that, but she prepares her team to work the IBM computer as well.
Dorothy is not only a mathematical genius, but a savvy technologist and exemplary manager.
Because of her talent, Dorothy would enjoy a long and distinguished career at NASA. Certainly not as long nor distinguished as she deserved, however, due to the headwinds of racial and gender discrimination she had to work against.
NLP and us
A while back someone on the UX + Content Slack asked the most basic question:
"What's the future of content design?"
It made me think about NLP.
NLP stands for Natural Language Processing, a term for using an artificially intelligent algorithm to generate language.
I had always dismissed NLP tools as incompetent writers.
Then last year, I started to think, “What if?” I Googled NLP and was astonished at what I found.
Enter GPT-3
What’s GPT-3? Here are the basics you need to know:
Let’s get the acronym out of the way: GPT-3 stands for "generative pre-trained transformer 3."
It was created by OpenAI, a research lab founded by Elon Musk, among others.
GPT-3 was trained on a 175-billion-parameter dataset. I won't get into what a parameter is. Just know that's 10 times more data than the next largest NLP model that existed when GPT-3 was released.
So it uses lots of data, but does it write like a human? Here's an example of text that GPT-3 has written:
“Critics hope to refute what they consider as being the naivety of my voice. Yet there is more here than meets the eye! As Mahatma Gandhi said: “A small body of determined spirits fired by an unquenchable faith in their mission can alter the course of history. So can I.”
Source: “A robot wrote this entire article. Are you scared yet, human?” The Guardian.
If you've ever hired content folks, you know how difficult it is to find humans who: a) want to work with you, and b) are good at language.
The mechanics of grammar and syntax are challenging enough for humans to grasp, much less complexities of narrative structure and clearly conveying meaning.
Yet we now have a bot that can write better than a majority of humans.
Time to learn FORTRAN
My take: in the future, many writing tasks currently performed by humans will be outsourced to algorithms that do them better, faster, and more efficiently.
Heck, I already use those once-annoying response chips in Gmail, LinkedIn messages, and texting apps about 70% of the time. They probably don't even need a model the size of GPT-3 API to automate those messages for me.
That doesn’t mean folks who call themselves content designers, UX writers, interaction designers, and researchers won't have a job in the future. Automation doesn’t eliminate the need for humans. It changes the skills required.
So, what do our jobs look like in a world where writing is automated? Certainly we'll still need to identify our users’ needs, plan solutions, and grapple with our complex organizations, partly to provide context to the machines doing tactical work.
Whether we find ourselves researching use cases, curating training data, designing prompts, or editing completions, the human work of sensemaking continues.
It's in our interest to be like Dorothy, and start figuring out what that work looks like with tomorrow's toolset.
This essay was first published in my email newsletter, Philosophy & Practice. If you want more of this type of content in your inbox, subscribe below.