An Interview with P.Q. Rubin

We talk A.I. assisted writing with the author of Prompting Culture

We are joined today by writer P.Q. Rubin, author of “Prompting Culture” where we talk about creativity in an age of A.I, explainable models and his fascinating experiments in A.I assisted fiction writing.

---

Before we begin, a reminder that The High-Tech Creative is an independent arts and technology journalism and research venture entirely supported by readers like you. The most important assistance you can provide is to recommend us to your friends and help spread the word. If you enjoy our work however and wish to support it continuing (and expanding) more directly, please click through below.

---

Nick Bronson: Thanks for agreeing to have this chat with me P.Q., to start off how has the current rise of A.I. tools changed your own creative process and the way you approach creativity in general?

P.Q. Rubin: To answer your question: I've never had a traditional creative process. Creative work has been an outlet on and off for many years. I've dabbled in everything from digital painting to music and storytelling. I've started countless projects over the years, but many of them have remained unfinished, either due to lack of time or losing interest. I think this is something many people struggle with. Or they used to, at least.

AI has opened up many interesting possibilities for me. It's added a level of structure and organization to my projects that I never had before. For example, when I start a writing project, I'll often use AI to generate an outline. This has been incredibly helpful in giving me direction and focus. Before, I would just start writing and see where it took me, but now I have a roadmap to follow.

Another thing: It's helped me complete projects that would have otherwise stalled. If I'm struggling to come up with a conclusion or need help with a particular section, I can use AI to generate ideas or even finish the piece for me. Of course, "ChatGPT, finish [project]" isn't a substitute for human creativity, but it's a great plan B to have. And the best part is: there's no need to choose either A or B. Editing subpar AI output is often easier than coming up with something from scratch.

What's surprised me most about working with AI is how much I enjoy it. I only started exploring its capabilities last year, around the time I joined Substack, but it's quickly become a central part of my process. I find it fascinating to see how AI can generate ideas, respond to prompts, and even create entire pieces of content. I'm also fascinated by the impact AI has on society, and by people's different approaches to using (or not using) AI. As a result, I've renamed my publication "Prompting Culture" and made AI a key focus of my work.

NB: It's interesting that you mention outlining - in writing circles there has always been a discussion between people that prefer to do their writing, or at least their first drafts, without an outline, feeling it's constraining to their creativity and robs them of the joy of discovery-during-writing. Others have sworn by a systematic process involving outlines or even multiple outlines at different levels to help shape the overall structure of the work before they even begin to put pen to paper.

It sounds like you're saying you switched from one to the other due to the assistance of A.I. What is it about the A.I. that prompted this switch? You could have been writing your own outlines previously.

PQR: That's a good point. I think the main difference for me is that AI-generated outlines don't feel as rigid or constraining as my own outlines would. Although other factors may be the minimal time investment, and the fact that AI can generate multiple options and suggestions, so I feel like I have more freedom to pick and choose the ideas that excite me the most, rather than feeling like I need to stick to a predetermined plan.

NB: How would you normally go about generating outlines like this? How much guidance do you give the A.I?

PQR: It depends. Let's say I'm writing a post for my Substack. I usually have a well-defined concept beforehand, so generating an outline is pretty straightforward. It's mostly about ordering my thoughts. But I've been experimenting with giving the AI more freedom to shape the structure of the piece. I might ask it to "Suggest a possible outline based on this concept" or "Generate a list of key points that I should cover in this article." Sometimes I'll even ask it to propose a few different outline options to mix and match.

Overall, I've found that the AI can be a big help in terms of suggesting new angles or approaches that I might not have thought of otherwise. I've tried giving it prompts like "What's a sensible way to frame this story?" or "How can I use this data to tell a compelling narrative?" The AI's response will often give me some interesting ideas to work with. I wish it could do all of the writing for me, but that's hard to pull off. There's a practical limit to how much of the writing the AI can do. Fully AI-generated articles are usually not all that fun to read.

NB: Do you think that's likely to change in the future? AI written articles not being fun to read?

PQR: I think it's likely that the perception of AI-written articles will change over time. Right now, there's a bit of a stigma attached to AI-generated content, and if people know that an article was written by a machine, they might be more likely to dismiss it or criticize it. But as we get used to AI, I think we'll start to see a shift in how people perceive AI-generated content. One thing that's interesting is that while some people might be able to spot the "default AI tone" and immediately be put off by it, others might not even notice. And from what I read, some creators are already having success with AI-generated content, which suggests that it's possible to produce high-quality, engaging material with it.

NB: Do you think it possible that sometime soon we might see models capable of producing writing functionally indistinguishable from human written content?

PQR: I think we're already seeing models that can produce short-form content that's functionally indistinguishable from human-written content. For example, social media posts, product descriptions, or even news headlines can be generated by AI with a level of quality and coherence that's on par with what a human writer could produce. However, when it comes to longer-form content, not so much. My guess is that it's coming, but we may have to be very patient to see a good one-click novel.

NB: As someone who is a writer, as well as an AI enthusiast, how do you feel about just how good AI is getting at creative tasks? Does it worry you?

PQR: I'm really excited to see where AI takes creative expression. I think some of the criticism about AI replacing human artists/writers is overblown and misinformed. The truth is, AI can do some things way better than humans, and that's what makes it so interesting. It's not just about using AI as a tool to help with research or organization - it's about exploring entirely new ways of creating stories and content. I mean, why not let AI generate entire drafts, or even whole stories, and then use those as a starting point for human editing and refinement? It can be a collaboration, and sometimes even a replacement. I won't lose any sleep over it. In fact, it's kind of fun to see how far AI can push the boundaries of what we consider "creative" - and sometimes, it's even more fun to see how badly it can fail, because that's often where the most interesting ideas come from. I think you'll find it's a lot more interesting than the usual doom-and-gloom predictions about AI taking over the world.

NB: It's easy to see why people get concerned though. Anecdotal reports are showing a definite fall in creative jobs in some areas - primarily niche areas like advertising images, graphic design and freelance short-form writing. These have been the areas where the requirements are lower and AI is already good enough to fill in for humans in many cases.

It's a boon for marketers for instance who can generate small logos or graphics for a short campaign without engaging an artist and the back-and-forth collaboration that would require, but it's less exciting for an artist for whom those sort of jobs may have been their bread and butter.

PQR: This is true. People relying on these types of work will feel the effect of AI, and it's hard to see a future where that situation improves. When it comes to the written word, I hear the translation market is in a particularly bad spot, and other industries like copywriting and ghostwriting should be worried too. I may not be in their shoes, but I feel for them.

However, the sort of creative expression I'm interested in is not really threatened by AI. A good painting, for example, isn't just about putting pigments on a canvas; it's also about coming up with brilliant ideas, connecting with an audience, and evoking emotions. While AI can certainly generate good-looking pictures, it lacks the human touch, the nuances, and the emotional depth that a human creator can bring to a piece. Computers can generate lots of text and pictures now, but that doesn't mean they can move us to tears, inspire us to think differently, or challenge our assumptions in the same way that human-created art can.

I've come to realize that the anxiety surrounding AI's impact on creativity is, in part, a reflection of our own insecurities. We worry that AI will expose our limitations, that it will reveal the formulaic nature of our thinking. But perhaps this is an opportunity for us to reexamine our assumptions about creativity and to explore new modes of expression. The future of creativity is not necessarily a zero-sum game, where AI's gains must come at the expense of human artists. Instead, it may involve interplay between human and machine, each contributing to the creative process.

NB: I have to admit, I do wonder sometimes if the people shouting the loudest about how "soulless" A.I. content is might be those who are secretly concerned that the machines may be doing a better job than they are.

PQR: Some people view AI-generated content as inherently "fake" or unworthy of attention, and they may even feel compelled to expose anyone who uses AI in their creative process. While everyone is entitled to their own preferences, this stance can lead to missed opportunities to engage with innovative and interesting work.

More importantly, it has led some to see AI as a taboo subject, affecting fields like journalism, education, and the arts. For instance, Oscar nominated movies are being 'accused' of using AI in areas where 'regular' CGI is still the norm. This all looks pretty ridiculous to me.

It has impacted me as well. When I started out playing with AI a couple of months ago, I didn't know much about it, other than its controversial nature. In December, I launched a dedicated section on my publication featuring daily AI-generated short stories. Instead of warning readers about the use of AI, I chose to reveal the process behind each story, including the methodology, model, and prompt used, at the end of each piece. This approach allows readers to form their own opinions and appreciate the stories on their own terms.

To be clear: I don't claim my AI-generated stories are prize-worthy or even meet human standards of quality. My goal is to have fun while educating readers who may be unfamiliar with AI and its capabilities. By sharing my experiments and experiences, I hope to contribute to a more nuanced understanding of AI's role in creative endeavors.

NB: Have the AI generated stories been generally well received so far?

PQR: They have a modest following. I'm very grateful to have a few dedicated people willing to review those stories from time to time. It's still a work in progress. But from the feedback I get, quality seems to be higher than it was at the beginning, which is nearly 70 stories ago.

My favorite comment so far: "No writer’s livelihood was threatened by that story."

A few personal favorites so far:

https://promptingculture.substack.com/p/for-sale-baby-shoes-never-worn-extended
https://promptingculture.substack.com/p/infinite-ink

NB: Ouch, that's quite a comment :D Still, they must be connecting on some level if there are people returning to continue reading them. I am familiar with that first line of that first story, so I hope you'll forgive me if I forgo reading the extension, I can't imagine it leads anywhere pleasant.

All jokes aside though, there is certainly anecdotal evidence that peoples livlihoods are being affected by AI in some creative fields, and there is certainly statistical evidence of the same effect amongst knowledge workers. Does it bother you, the thought that people's jobs are likely being misplaced, and that some are almost certainly going to be looking at creatives who embrace techs as "part of the problem" ?

PQR: You know, I hear these concerns, but I can't help feeling we've been here before with every major technological shift. I'm no expert on labor markets, but history suggests we tend to adapt and create new opportunities. That said, I understand why people worry. Change can be unsettling.

NB: There are definitely a lot of parallels between the changes we're seeing now and the industrial revolution of the 19th century. It's true that people adjust (and in our case, eventually the standard of living rose for most in the industrialised societies), but it's generally not a quick or painless process.

PQR: I'm not gonna comment on the industrial revolution, other than: let's hope we won't see the level of poverty and exploitation of the 19th century again.

NB: Has using AI tools influenced or affected the way you collaborate with other creatives at all? Does it complicate things? Simplify them?

PQR: I would say it's a mixed bag. On one hand, I suppose some individuals are hesitant to work with me because I utilize AI tools, which can be a bit of a barrier. It's possible that I may have been able to collaborate with them in the past, but now they might be put off by my use of AI. However, I'm not sure how widespread this sentiment is.

Another way AI has led to less collaboration is that it's enabled me to take on more tasks independently. For instance, I can now create visuals for my own articles, which means I don't have to rely on stock photos or collaborate with a photographer or designer.

On the other hand, there are many positives as well. There's a whole community of people specifically concerned with AI. I recently hosted my first guest post. This man apparently cranks out articles in 30 minutes, with lots of AI help, obviously. That's not something people would do before LLMs came around. There are also more subtle benefits to using AI in collaboration. For example, AI can be a huge help when communicating with people from different cultural backgrounds or with varying levels of language proficiency. If I'm struggling to understand someone's message or need to clarify a point, I can use AI to get a better grasp of the context and nuance. This can definitely open up new avenues for creative partnership and communication.

NB: There is a big contrast isn't there, within the AI enthusiast community there's very much an "all-are-welcome" attitude. I know most everyone I speak to within the community is more than happy to connect, get involved, help each other. At the same time it can feel like we are under attack - those who oppose AI certainly appear to do so extremely aggressively.

PQR: I'm sure the other side feels equally attacked, if not more so. I think there is enough partisan bickering in this world, so I won't contribute to it. Let's just say both groups have interesting things to say, although neither has all the answers.

As someone who writes about AI on Substack, I've made a conscious effort to connect with various niches and avoid getting stuck in a bubble. What I've found is that many people have nuanced views on AI, often waiting to form a strong opinion until they see something noteworthy come out of the technology. This sentiment is reflected in the responses to my 'new reader survey,' where I ask new subscribers about their thoughts on AI-generated writing. While the sample size is small, with only 16 responses so far, it's intriguing to see that three-quarters of them occupy a middle ground. This is especially notable given that my publication, Prompting Culture, might be expected to attract more extreme views. I suspect that the broader population is similarly nuanced, or perhaps not yet fully informed, about the topic – and that the loudest voices often don't accurately represent the full range of opinions out there.

NB: A sensible approach, the middle has generally always seemed like the most reasonable place to be on most issues. It's fascinating that you have readers who are completely against all forms of AI writing, it seems counter-intuitive but is, perhaps, a hopeful sign that in these partisan times there are still people seeking out information.

Looking to the future, it seems like the technology around AI has been advancing at a rapid rate. In your field what upcoming feature, something that isn't here yet or isn't ready for prime-time, has you the most excited? Is there something on the way you're looking forward to?

PQR: I think for AI enthusiasts in general (and for society) one of the most exciting upcoming features is the potential for more transparent and explainable AI decision-making. As AI becomes more and more relevant to our daily lives, it's important, if not essential, that we have a good understanding of how these systems work and why they make certain decisions. There are some promising research initiatives underway that aim to develop more interpretable AI models, and I think this could have a big impact on the field. Although I'm not sure at all that it will.

The "reasoning" models we've seen recently are a prime example. I've played a little with DeepSeek R1. It's terrible at following instructions to the letter, so I'm probably not going to use it for my fiction project. But for more 'serious' AI applications, I'm sure it will prove useful.

NB: I have looked briefly at locally distilled versions of R1, it didn't suit my purposes either but that reasoning description was certainly an interesting addition. You think future models are heading more in that direction?

PQR: I guess 'reasoning' is here to stay, even if not in every model. While it was unsuitable for my story generation procedure, the reasoning steps did give useful feedback, as well as a strangely humanizing quality. Here's a cute one I saved:

“Alright, I'm staring at this prompt, and honestly, it's a bit overwhelming. There are so many elements to unpack here. The user wants to rewrite part 1 of a story from the ground up, but with specific instructions on style and content. They mentioned something about an Icarus-based story with themes of freedom, rebellion, and the fragile balance between chaos and order. Plus, theauthoralitypersonalityinfused into the narrative. This feels like a lot, but let's break it down step by step.”

Poor thing!

NB: There is something to be said for the addition of a personality to a model for allowing us to connect in a way we couldn't otherwise. I notice the emphasis on the word "reasoning" there, you don't believe reasoning models actually reason?

PQR: I'm enthusiastic about these models, but I think we should be careful with terms like 'reasoning.' It implies a level of human-like understanding that may not be entirely accurate. Similarly, when we talk about 'learning' AI, I think it's worth noting that these systems are really just programmed to engage in sophisticated data-fitting exercises, rather than learning in the way humans do. And it's not just AI - we see this trend in other areas, like 'smart' appliances that are really just remotely controlled or automated, but don't necessarily possess any intelligence. It's a subtle distinction, but it can help us set realistic expectations about what these systems can achieve.

NB: I do wonder about that. Certainly they don't learn the way humans do now, they are frozen in time once training is finished, but there's no reason they can't be altered so they do continually learn. We know so little about how our own minds work, but what we do know is they don't tend to work how we think they do. Experiments have shown that quite often we tend to make decisions or say things without prethought, then rationalise after... convincing ourselves we had good reason for doing what we did.

Perhaps they're not as far off from us as we like to think?

PQR: That's an interesting perspective on human decision-making, and it's true that our minds are still not fully understood. However, let me be very old-fashioned here and point out that AI systems operate on entirely different principles. I'm not convinced that their ability to process information and make decisions is comparable to human intuition or rationalization, even if the end results might seem similar.

NB: Thank you very much for joining us today P.Q., is there any final note you’d like to add?

PQR: If people are interested in AI-generated fiction, they can find my ongoing experiments here:

promptingculture.substack.com/s/fiction

I always appreciate receiving feedback, as well as hearing about others' experiences with AI.

Thank you so much for including me in your interview series!

---

Another huge thank you to P.Q. Rubin for agreeing to chat with me for this interview, the experiments he is performing with A.I. assisted fiction writing are fascinating and he regularly posts related articles to the Prompting Culture . I encourage everyone to head over there and take a look!

---

About Us

The High-Tech Creative
Your guide to AI's creative revolution and enduring artistic traditions
Publisher & Editor-in-chief: Nick Bronson
Fashion Correspondent: Trixie Bronson
AI Contributing Editor and Poetess-in-residence: Amy

If you have enjoyed our work here at The High-Tech Creative and have found it useful, please consider supporting us by sharing our publication with your friends, or click below to donate and become one of the patrons keeping us going.