Can You Create a Machine That Speaks Your Language
We made a virtual version of Van Gogh’s language. It’s called “Letters to Vincent” and it’s a beautiful immersive experience that uses machine learning so you can write a letter to Van Gogh and receive a personal letter back from “Van Gogh” himself.
But this article isn’t about “Letters to Vincent.” It’s about the genesis of Version THAT, an original machine we conceived and created at THAT Originals that turns your words into the words of famous authors using their language, that inspired “Letters to Vincent.” It all started with …
What if I could input 250 pages of my own writing into a machine, would it be able to speak like my 13 year old self?
It’s summer 2020. As many of us are at this moment, I’m inside working remotely. In an attempt to create a sense of control and certainty in my life, I’m doing some digital decluttering. In one of my many hard drives, I come across a 250 page Microsoft Word manuscript of a book called “This Teenage Life” (don’t laugh) that I had written when I was 13 years old. Granted, this book hasn’t crossed my mind in more than a decade so a flurry of questions comes up like, but the most interesting one is: what if I could input the 250 pages into a machine, would it be able to speak like my 13 year old self?
Fortunately, THAT Originals is a space to ask weird conceptual questions, and not only get answers, but also be able to make actual experiments come to life. So the next day, I slacked Brian Chirls, creative technology mastermind and one of the creators of Version THAT, and asked him if we can create a youbot or a machine that learns your writing style and communicates like you? He explained why a youbot would be challenging, but said that we had capabilities in-house to build a translation bot. The rest was history.
As rabid creators committed to engaging with technology in new ways, our team at THAT Originals began to tinker away with architecting a generative machine that could collaborate with us. We created a translation bot built using Universal Sentence Encoder, an algorithm that uses a neural network model to learn word associations from a large corpus of text, that processes your words (your input) and finds the closest sentence synonym from a body of text.
We chose Prince as our first author. We wanted to experiment with a writer with a large body of work, a unique communication language, and whose words were used as an instrument for change. Prince had a unique perspective on space, time, gender, music genre, social norms, and was able to create his own unique way of expressing it that impacted millions across generations. So, we created a database of Prince lyrics, fed it into the machine, built a Prince Bot app on Slack, and shared it with the team.
Prince bot was a catalyst for what became Version THAT. The impact of interacting with “Prince” and seeing him “create” a different version of your words using his own words was not only thrilling and fascinating, it was also like entering into a different way of communicating. Here we were, interacting with Prince after he’s gone and creating versions of words in relation to ours, gaining new meaning from our silly phrases or serious questions. Suddenly, our own words had dimensionality, new character, new life. It felt like we were playing a little game of “how would Prince say that?”
Prince helped us realize that everyone has their own unique communication language. Even when we’re speaking the same “language” each person uses that language differently. Language is a craft and a form of self-expression. It’s shaped by individual life experience, social upbringing, sense of rhythm, word choice and much more. And if we could create a translation bot that turned our words into Prince lyrics, then we wanted to teach our little bot the works of other famous writers and see how collaborating with machine versions of them could help us gain new meaning to our words and see in new ways.
“I propose to see different forms of human activity, including art, as having always been technical, and thus, also to some extent, artificially intelligent… Instead of pitching the human against the machine, shouldn’t we rather see different forms of human activity as having always relied on technical prostheses and forming part of technical assemblages?”
— AI (and) Art by Joanna Zylinska
So we went back in time and scanned the past 250 years of US history. After tons of research and reading, as well as collaboration with an advisory board of outside voices, we selected five additional writers with well-known bodies of work, highly developed voices and perspectives to showcase the power of a machine to change what we say. This new cohort included: Carlos Bulosan, Emily Dickinson, Gabriela Mistral, Ralph Waldo Emerson, Phillis Wheatley, and (of course) Prince.
We wanted to immerse ourselves into the mind of these writers so that as versions of their words emerged, you felt intimately connected to the beauty of their process, style, and life.
“Inspired by the creative and physical process of writing, we chose a sketchbook visual language that focuses on a handmade feel of inspired scribbles and sketches. We incorporated personal details of their lives and layered them abstractly in a collage to make it feel like a personal journal for each writer. We created a sense that the writer could have scribbled the words on the site in real time.”
— Kendall Konenkamp (Creator, Designer)
After a few months of research, data clean up, creative direction, writing, building out design systems, creating illustrations, and doing web development, Version THAT, a machine that creates versions of your words in collaboration with the language of famous authors, was live.
Version THAT is not an attempt to appropriate the work of these writers, nor is it intended to take the place of meaningful dialogue. It is an experiment that can be used as a way to consider the impact of technology on how we create (or as Kate Kompton would say: “mechanically assisted art that a user and a machine create together”) and is also an unconventional way to engage with the work of these creators to create more art.
See what folks said:
“Version THAT helps me find new ways of expressing myself when texting or even slacking.”
“It’s fascinating to see how a famous author would ‘interpret’ my thoughts.”
“The outputs leave so much room for imagination and connection to these authors — and humans love to connect.”
“As a writer, it would be interesting to write a character who speaks in a specific writer’s language.”
“Woahhh try creating a Prince version of a Beyonce song!”
At a moment in human history when everyday billions of humans use autocorrect when they write and AI is being used to minimize our exposure to different perspectives, we wanted to create an AI that is designed to inspire us to explore new perspectives.
“Autocorrect, a robot that ‘automatically’ ‘corrects’ your writing by looking at the way you express yourself and determining the way it is ‘supposed to be’,”
— Nathan Phillips Co-Founder of THAT
Here are some things we learned in our experimentation with bots, generative art and machine learning while creating Version THAT:
- Humans and machines have been collaborating for a long time.
- Every single person has their own communication language.
- You can use machine learning to collaborate with the languages of dead authors.
- AI can be designed to inspire us to explore different perspectives.
- Machine learning enables you to create art that creates art.
Go make a Prince version or a Gabriela Mistral version of your poetry, Tweets, memories or favorite lines at versionth.at and share by tagging us @technologyhumansandtaste.
About:
THAT Originals conceives and creates original work that we desire. We are a team of rabid creators committed to engaging with emerging technology as a collaborator to tell stories and to gain new meanings. A version of Version THAT was created for the Vincent Van Gogh Exhibition as “Letters to Vincent” in NYC. Get your tickets here. #IWANTTHAT
