Do not miss this and save in your calendar:




Ross Goodwin was a ghostwriter for the White House and was on the 2008 Obama campaign team before making a U-turn and devoting his creativity to computational creative writing. Goodwin created the world’s first film from an AI-written screenplay, and in his practice, he pushes the boundaries of poetry on the blockchain. In conversation with Anika Meier, the artist speaks about his career and influences, training algorithms, and good writing and bias in AI.

Anika Meier: You worked as a ghostwriter on the 2008 Obama campaign, you held a presidential writer position at the White House where you wrote nearly 100 presidential proclamations, and later you created the world’s first film from an AI-written screenplay. How did you get from working at the White House as a writer to letting AI write a screenplay and then produce it?

Ross Goodwin: If I'm honest, I burned out on politics after the time I spent serving in the Obama administration. It was an intense period that lasted years, and in the final few years of it, I began to automate briefing binder processes at the U.S. Treasury Department for then-Secretary Timothy Geithner and his Deputy, Neal Wolin. That's when I truly began learning to code, and as a writer, I immediately gravitated toward natural language processing (NLP) and natural language generation (NLG) applications. My first projects were Markov poetry generators with various constraints, and I suppose I was trying to do something that felt like the opposite of politics—something with minimal decorum, and poetry fit that role perfectly at the time for me. Only when I became a better programmer and my bots began to perform better was my work labeled as "AI". Initially, I was skeptical of that label and rejected it outright. But eventually, I had no choice but to accept the inevitability of AI as a term in the non-technical culture for the type of deep artificial neural networks I started working with in late 2015.

AM: "Watching it for the first time, I almost couldn’t believe what I was seeing—actors taking something without any objective meaning, and breathing semantic life into it with their emotion, inflection, and movement," you wrote about the film titled SUNSPRING. Didn’t the AI do a good job as a screenwriter?

RG: Yes, it did. But the actors did a better job as interpretive agents, in my opinion.

Scene from SUNSPRING, 2016, directed by Oscar Sharp, starring Thomas Middleditch: "He pulls his eye from his mouth."

AM: You’ve developed BENJAMIN, the first AI software that writes screenplays. How did you learn to train algorithms?

RG: Trial and error. I got my head around deep learning models in early 2014, before I entered graduate school at NYU ITP. The crucial factor was a friend from my undergraduate years at MIT (David Dopson, currently a rockstar at Google) who understood the technology better, along with my experience with regression analysis from my undergraduate degree in economics. A deep neural network can be understood as essentially (but not exactly) a very large, complex econometric regression equation. That basic understanding gave me the foothold I needed to go further. Like regression analysis, training models involves a lot of little tricks that you learn the more you do them.

AM: The media was quite irritated by the film and criticized the lack of storytelling and coherence. What are the criteria for good AI art for you?

RG: Concept and execution. There are a lot of boring concepts in AI that have already been done to death. If you want to make art professionally and not simply as a hobby, I believe you need to remain highly original, conceptually speaking. There's also a lot of craft to executing a good concept within AI, such as training your own models, and I believe at times a good execution can save a mediocre concept and vice versa.

A lot of people criticized SUNSPRING's screenplay, which I expected. But most critics also praised the overall concept and experimental nature of the film, which I found very reassuring.

AM: Is being the first important when it comes to working with new technologies? I ask this question because I assume it’s often hard for artists to decide whether to experiment and be fast or to wait and learn from other artists experiments.

RG: No. Being first doesn't matter in the arts, at least in my opinion. It's being the best—or rather, making the best art possible—that matters. But that's subjective, etc. physical lexograph, 2015, mirroring a polaroid photograph by A.J. LeVine at NYU ITP.

AM: How do you explain generative text to people who are not familiar with what you do?

RG: It's like I'm using a really fancy typewriter of my own design.
Or perhaps it's like the invention of the camera. (Regardless, has been a consistently reliable entry point to my body of work in that sense.)

AM: You call yourself a data poet. Why not just poet or AI poet? Are labels or categories important when positioning oneself in a new field?

RG: Yes. I still don't like the term AI. And I feel the need to be more specific than simply "poet" to explain my work more effectively.

I used to use the phrase "not a poet", mostly as a joke, but people would get confused, and so I switched it to "data poet" because of the slant rhyme it creates with that original branding. Cohesiveness has always been an important component of my practice, as I believe good writing requires it regardless of how it is performed.

AM: How do you explain to yourself the sudden interest in poetry on the blockchain? I remember when you released your projects A CLOCK and A WATCH on fx(hash), and I made sure to be there the second the mint started to not miss the opportunity. At the university in Heidelberg, where I studied German literature, the poetry classes were always the least popular ones. They were my favorite thing, though.

RG: I had been casually observing blockchain activities for many years before I dove in. (I actually had 1M+ dogecoin in early 2014 but failed to hodl. Whoops.) I first considered making art on the blockchain as early as 2016. However, I was more concerned about its environmental impact at that point.

I love fxhash because it's truly a marketplace for code NFTs rather than pre-rendered image or video files. And I love Tezos because of its commitment to minimize environmental impacts and externalities from the beginning. I think that mentality is important and needs to make its way to the broader blockchain community for the paradigm to ultimately succeed. However, I do see that happening, and I'm hopeful about the future of Web3.

"Narrated Reality" device components, 2016, NVIDIA Jetson TK-1 board, Datamax O'Neil Thermal Printer, on-device artificial neural networks, thermal paper.

AM: With poetry on the blockchain, something deeply emotional is connected to something very rational. How do you manage to combine these two aspects in your own artwork?

RG: Let me answer that with an analogy:

AI : Blockchain :: ✒️ : 📓

AM: Who are some of your influences? Are you mainly influenced by writers and poets?

RG: I have so many influences at this point that it's hard to track them all. I've been consistently inspired by my collaborators, including filmmaker Oscar Sharp, public digital installation artist Es Devlin, the band YACHT, and most recently theVERSEverse collective.

However, my original source of artistic inspiration remains the same. My aunt is Annie Leibovitz, and she gave me my first "real" camera at age 11: a 35mm film Contax S2 SLR model with a 50mm Zeiss lens that she’d used herself in the field. I still use that camera. And to this day, I remain deeply inspired by the example my aunt set in the arts. There’s no question in my mind that her work will live forever, as throughout her storied career, she changed the entire medium of photography from behind the lens of a camera.

AM: NOW, THEREFORE, I brings you back to your beginnings as a ghostwriter at the White House. You have captured a high-resolution digital photograph of Pacific ocean waves lapping up to an actual physical copy of a Presidential Proclamation (WRIGHT BROTHERS DAY, 2009), digitally erased the document, then created 10 unique 1/1 editions, each with its own fictional Proclamation text. To create these fictional Proclamations, you trained bespoke text-generating AI on a corpus of actual Presidential Proclamations you wrote or worked on while serving your country at The White House.

What have you learned at the White House that you feel the urge to share in the form of an artwork with the world?

RG: Power is overrated. And those who seek power rarely deserve it. The best American politicians, in my opinion, arrive in DC almost by accident, or perhaps as a last resort.

Proclamations hold excessive power, as they represent unilateral assertions from the Executive Branch that carry the weight of law. I wanted to deconstruct that dynamic in this latest artwork and suggest that the only true power, at least in this world, lies in nature itself. It's the ocean. The beach is a natural border with something so much deeper and more ancient than any human law, which sailors have feared and respected like a god for millennia. Needless to say, the ocean represents death in many poems, and for good reason., first physical version, 2015, built inside an antique Mamiya C33 TLR medium format film camera.

AM: How important is the visual component for you when creating text-based work with AI?

RG: I consider myself a visual artist. But my aesthetic is deliberately handmade and may appear less polished than other practitioners' at face value. However, I consider that a deliberate choice that I've made in my work. Overall, I'm trying to mirror the "homebrew" aesthetic of projects like the Apple 1 computer, the Wright brothers' airplanes, and other works that are considered fundamental in the history of American engineering.

AM: You trained a text-generating AI to create fictional Proclamation texts. Do you work with or around the limitations of AI?

RG: Yes. But I think most good art grows under constraints. My medium is technology, and so I must work within the constraints of any system I'm using or developing.
I believe a computer can be played like a musical instrument and not simply used as a tool in the traditional sense. I also believe computer code is the most unknown and unexplored creative medium we still have, and that's a principal motivating factor in my artistic practice.

AM: Experts fear that the omnipresence of AI in the near future could lead to "an erosion of trust in media, in government, and in society" as I read in the New York Times again today. Do you share the pessimistic point of view on AI and its impact on society and culture?

RG: I have some concerns, and I think we may need a solution like universal basic income (UBI) to alleviate economic externalities that may result from AI or any new technology. But I think the real concerns around AI are more mundane than many commentators would have us believe. And I think the field is making major headway at addressing real problems like bias in AI, which is evident in software products like ChatGPT.

I believe any sort of government regulation at this point would only serve to benefit larger companies at the expense of smaller ones, including individual practitioners such as myself. That's why I am currently (mostly) against regulating AI, excluding common-sense measures like prohibiting certain systems, especially computer vision, from being used with weapons. But I think it's an ongoing discussion we need to have as a society. Elevating the public discourse on AI remains a major goal of mine, and I think that serves the public better than any form of regulatory intervention would, at least at this point in time.

Ross Goodwin wearing his Narrated Reality prototype, 2016, a wearable device that poetically narrates imagery, location, and time on the user's behalf.

AM: Is there anything we could do to prepare ourselves for living in a world driven by AI?

RG: Play with AI. Play can be a serious, mindful activity. It's about discovering and analyzing design affordances. At this point, AGI is more of a design problem than a computational one. And in the context of AI, I believe some amount of play is required for any practitioner who wants to succeed at any level with this particular stack of technologies.