Generative AI tools like ChatGPT have only existed for a few months and they’ve already spawned a bajiliion videos where creators show how they’re using them — to spark new ideas, as writing partners, to do the grunt work of social media posts and ad copy, even to write scripts.
How to create faster, better, more creatively with AI
Get our Creator's Guide to AI and you’ll get:
Our report on how creators are using AI tools
Tips for using AI tools to bolster your creativity
A look at how successful creators are using generative AI in their workflows
Our guidance for getting started with AI, if you're one of the holdouts
So I thought I’d try using it to come up with ideas for a new podcast. Now, I’m no podcast expert — although I've worked on several podcasts and video series, I've never produced my own show. And to be honest, I didn’t plan to. I thought using ChatGPT in this way would just be a fun experiment I could write about.
That is, until it came up with an idea that I loved too much to pass up. Here’s how I got ChatGPT to give me what might become the topic of my very first podcast.
My first prompt asked ChatGPT whether it could help me figure out some new ideas for a podcast. Obvious enough. It suggested six different show styles that I could consider producing.
The suggestions included a mix of formats and topics including an interview show, a roundtable discussion, a niche-specific show, a solo show, an audio drama, and a how-to show. Some of the suggestions overlap — for example, a how-to show could take the shape of a solo show or an interview show — but still, I hadn't actually considered any of these formats.
When I replied that I was interested in a niche-specific show, it immediately came back with some suggestions for how to get started, including choosing a niche, determining the format, and researching competitor shows — all good advice for a beginner like me.
Using my personal style to decide on a format
So far, the advice had been useful, if relatively generic — the kind of thing you would find in any number of blog posts or online articles (like the one on this very blog). But I wanted ChatGPT to give me more customized advice for my personal style. To do that, I first asked it to describe my style using the script of a talk I gave that captured my voice and the tone that I envisioned for the show. Since it was too long for a prompt, I split it into two sections.
Well on my way to asking the right kinds of questions about developing a new show, I prompted ChatGPT for a recommendation for a show format using its description of my voice and tone.
It suggested a solo show would most match the tone and style of my talk. I'd previously suggested I wanted to try a niche-specific show so I asked ChatGPT to tell me what the differences are.
Its description of the differences convinced me that I'd made the wrong choice initially, and that a solo format was indeed better for my show. Not bad! I felt like I had actually gotten somewhere and that my new show was starting to take shape.
It gave me good advice on how to get started without prompting
It encouraged me to think about aspects of the show I otherwise wouldn't have
It knew and suggested best practices that I would have skipped like researching what shows already exist in my niche
It talked me out of a niche-specific show format
Most of the information I could have gotten elsewhere
Only one suggestion was specific to my own situation, while the rest were generic suggestions
As ChatGPT told me: "Ultimately, the best topics for a solo show are those that align with the host's passions and areas of expertise, and that they are knowledgeable and enthusiastic about." Alright, ChatGPT, let's see if we can find some.
Finding a topic
At this point I decided to be more direct about the kinds of topics I was interested in so I could get specific suggestions. I prompted it with a broad idea (the history of technology and how it influences human art and culture) and started to narrow it down from there by asking for lesser-known technologies and movements that had outsized impacts. I also made sure to ask how the theme of the show could be incorporated into each episode idea.
I hadn’t been specific about the topics I was most interested in, so ChatGPT gave me suggestions that were all over the map. Roughly half of its suggestions were art movements including punk rock, silent film, and dadaism and half were technology trends including the telegraph, the microchip, the laser, and satellite technology. But when I asked it to suggest how technology influenced the artistic trends it mentioned, several of the suggestions started to get interesting.
From here, I could have taken the easy way out and just chosen the single idea I liked the most to develop. But, instead, I decided to test a few ideas and see how well I could develop them using some back-and-forth prompting with ChatGPT. And to test it at an even higher bar: Could ChatGPT convince me of an idea that I didn't initially like?
Idea 1: The technological influence of silent films
Verdict: ChatGPT talked me into an idea
I've seen a few clips of silent films but I'm overall pretty meh on the whole topic, so I decided to pick this one to see if it could convince me there was something interesting enough in the topic to pursue.
ChatGPT's initial outline was standard, but overall pretty good and I felt it gave me a way to organize the content for an episode.
The suggestions it gave about how silent films established the cinema genre sounded okay but frankly, it didn't pass the "so what?" test for me. I challenged it to generate ideas and flesh out topics, but none of them seemed all that relevant. Overall, I was disappointed and ready to give up.
That is, until I got a few interesting ideas reading its response about the ways that the influence of silent films are relevant to current video trends. They were generic, but I realized that several touched on current trends that I'd seen and been curious about such as the universal popularity of Khaby Lame and the #gentleminions trend.
Because the topic was so specific, I had a feeling that these connections would have been hard to make if I had just read other commentary about silent films online.
To test this, I did a search for "why do silent films still matter" and read the first few pages that came up.
None of them had any hints for the connections that I made with ChatGPT — without it, I probably wouldn’t have hit upon this idea (not to mention that searching and reading all those pages took a lot longer than my conversation with ChatGPT).
ChatGPT was starting out strong.
Idea 2: Satellite communication and art
Verdict: ChatGPT couldn't quite get it right
I know a little about satellite communication because my husband works in the field. As a result, I was pretty sure ChatGPT wasn't going to be able to convince me there was a viable story in here. Sure enough, it fessed up right away that there wasn't much there there. It's good to know that there are times ChatGPT won't try to push forward with an idea — there must be at least some kind of quality control going on in that machine brain. I dropped it and moved on.
Idea 3: Technology’s importance to punk rock
Verdict: ChatGPT generated a good idea but couldn't make it great
The initial suggestion made by ChatGPT was that technology made it cheaper and easier for punk musicians to bypass the traditional music establishment and still get their music heard. Fair enough.
Its suggested outline was pretty generic, but I wanted to see if I could get a more interesting angle on the story so I continued to prompt ChatGPT. It suggested the obvious: that musicians like "bedroom pop artists" (a term I'd never heard but turned out to be common when I researched it) could still use technology to bypass the traditional music industry via streaming services and social media. When I protested that the politics and culture around these types of musicians was different than that of punk rock bands, ChatGPT never could give me a satisfactory answer.
Although it didn't tie the story up in a neat little bow, I felt it ended up reasonably strong as a story, if pretty obvious.
Idea 4: AI and the randomness of dadaism
Verdict: ChatGPT sparked a great idea that I might even pursue
Now for my favorite idea: Dadaism. I was pleasantly surprised that it showed up on the initial suggestion list because it was an idea I had previously looked into on my own. A few years ago, an AI transcription tool mistranscribed "data science" to "dada songs" and the unexpectedness of the little glitch intrigued me. After all, dadaism was an art movement characterized by nonsense and irrationality, so the fact that this AI tool landed on a nonsense word all about nonsense was pretty on the nose.
As ChatGPT told me earlier, a solo show focuses on "the host's personal perspective, experiences and opinions on the subject." Since I had experiences and opinions on this particular subject, the idea was fairly strong, even though it gave me something I already knew.
ChatGPT’s initial outline was pretty generic. But it did suggest a few angles that I could run with and weave into the episode because of my many opinions about it.
I prompted it to flesh out those areas, and there were several that I thought were good enough to include.
From there, I used ChatGPT to further explore some of the ideas that I felt were interesting, like how randomness and glitches in AI tools could be a way to inspire new ideas.
Its answers helped me weave the ideas into the outline, which at this point was okay — but not great yet.
Since it had been several years since I had my original idea to interpret the history of dadaism through the lens of new technology, the topic was a little stale. Technology moves fast, and a lot has happened in those short years. So I asked ChatGPT for some more recent ideas. I ended up feeling pretty good about its suggestions.
I also shared some of my own ideas with ChatGPT, and I thought it did a good job summarizing and extending them.
By the end of the brainstorm session, I felt that it had really helped me come up with some new and interesting ideas and to refine ideas that were fairly raw. It did a good enough job that I started really considering how the show might shake out for real: I started thinking about who I might invite as guests and what personal stories I might like to tell.
Although this was designed to be an experiment, my experience with ChatGPT made me consider actually making a show — one that would explore the intersection of modern technology and concepts of dadaism. The chatbot convinced me that this topic is only becoming more relevant, especially the way dadaists used randomness and nonsense as commentary on society. It also helped me think of many present-day examples I could draw from and to link it back to past examples I didn't know much about.
And the best part: it gave me lots of ideas of how I could weave those concepts together. From here, it's human powered — I'll have to figure out what I really want to say on these topics and get started mapping out the story that I want to tell.
Overall, I had a great experience with ChatGPT. I ended the conversation excited about the ideas it helped me combine and build. The AI’s suggestion of Dadaism resonated with me the moment I saw it, and after spending a few hours going back and forth with the AI, I ended up even more excited about it. The session sparked a number of ideas that I will seriously consider working on.
It was able to weave together several ideas into a coherent whole that I wouldn't have thought of.
It brought new excitement to an idea I'd shelved a couple years ago and helped me flesh it out with new angles on the idea that I hadn't thought of.
There were a few times where it added some context that saved me from having to happen upon topics through research (Although, because ChatGPT isn't always truthful, I'll have to do research before I start writing the content though)
The interactive experience made it feel much easier than researching the topic through other methods
It was fun! It didn't feel tedious like reading and researching can sometimes be.
The ideas it generated on its own tended to be pretty obvious, and it was only after I dug in that I ended up getting some sparks of good ideas
The outlines were also pretty generic and not very creative.
Its suggestions were hit-and-miss and I kept having to bring ChatGPT back to my point-of-view
I will still need to research the topic to get real sources and to validate the information it provided
ChatGPT's early suggestions often come across as generic, but my example shows that it’s worthwhile to keep trying. I was glad that I wasn't too fast to discount an idea that didn't initially resonate with me and used iterative prompting to challenge my own thinking on the topic.
But, as with all technology tools, the experience revealed that nothing can replace your point of view and your conviction about an idea. AI can breathe new life into old ideas, challenge you to make the ideas better, and consider angles you hadn't thought of. It can also save time on research (although it's still important to do research since ChatGPT and other large language models can tell you false things with conviction). Working with it can guide your own thinking on the topic and help you come up with ideas you might not have had otherwise.
Today, we’re rolling out a new set of Descript subscription plans with new pricing tiers. And we’re thrilled to announce the long-awaited public launch of Overdub, a state-of-the-art voice synthesis feature that replicates your own voice, included in the new Descript Pro plan. Of course, Overdub isn’t the only thing that makes Pro our most powerful offering yet — check out the full feature set in this video.