An AI walks into a bar...
We speak to writer and performer Madeleine Brettingham about the viability of a robot Charlie Chaplin and why she is trying to kick AI's ass
British writer and performer Madeleine Brettingham has pondered what becomes of her craft in a world where our robot overlords can write a poem about Starbucks in the style of Rimbaud almost instantly. She kindly agreed to an “old-school,” in-person interaction to discuss, which I (Eleanor) recorded with software that transcribed our conversation in real time, perhaps to be recycled into fuel for another android anthem…
Our conversation, condensed and edited for clarity, is below.
🔑 Key takeaways from our conversation
AI will raise the bar for quality and make some humor less valuable. AI generates the obvious and the derivative, thereby highlighting exactly where the unexplored creative territory lies. The kinds of jokes that machines can create easily, like parody or dad jokes, will lose their social signaling value.
Comedy needs a body and a soul at risk: AI can't replicate the physical connection a comedian has to their material, such as the "mouth feel" when a joke doesn't quite fit, or the vulnerability of someone doing standup. The risk, embarrassment potential, and effort are part of what makes something funny (and why we probably wouldn’t laugh at a robot Charlie Chaplin).
To survive, double down on being alive: The competitive advantage lies in a unique perspective, original insights, life experience, and a voice that feels authentically human. These exist outside the AI dataset.
EW: You recently published an essay about your experiments with AI, which I loved reading and which brought me to you. Tell me about it.
MB: The first thing I want to emphasize is that I do not use [AI] at all for writing. The essay that I wrote was in the spirit of Googling an up-and-coming rival. Does it have the chops? Where can it replace me? Where am I going to have to get sharper?
My knee-jerk reaction is to say that it will never replace the magic of the human soul. But the people who pay me to write aren't going to be doing that; they're going to be very interested in what this does. So my attitude is: I need to get interested too, and I will feel calmer once I've had a proper look.
EW: How do you feel people in your industry are reacting to developments in AI?
MB: People are joking about it a lot, which is to be expected, because that's how comedy writers process anxiety. I've heard everything from, ‘I think this is going to replace me in five years, and I need to retrain’ to ‘It's not a threat. It can't be funny.’
It's at a stage, like when Ozempic hit the market, where there are probably quite a few people who are on Ozempic, but nobody's admitting to it. It's kind of like the Ozempic of the mind: who knows, really, who's taking this stuff? Until the norms are clearer around when it's acceptable to use it and how it's acceptable to use it, we won’t know.
EW: Talk me through the process of comedy writing. What different kinds of writing do you do? What types of jobs would you be paid to do?
MB: There is a whole range of things that come under the umbrella of comedy writing. I've written short fiction, plays, narrative comedy, pilots for sitcoms, jokes, and sketches. I also did voiceovers and stuff like that when I started out.
There are so many different types of jobs. You could be doing a punch up of somebody else's work. Or there's a live piece and you might get paid to throw in some ideas about things that the comic could do or different scenes that they could play out. Generally, it falls into categories of pitching, brainstorming, drafting, and revising, which is true of a lot of writing. But it would be impossible to give a typical day, which I think is one of the reasons why [AI] doesn't feel hugely useful to me.
EW: What is your creative process like?
MB: I feel like the best ideas are something you have a connection to, personally, even if you're writing for something else, which is why I think it's a little bit difficult for an AI to replace [a writer] or to build on an idea that an AI has given you, because an LLM can't replace your connection to the words and the idea.
For example, sometimes in stand up, somebody will suggest a topper joke to me. Even if it's a good idea and it technically works, it will have an odd mouth feel to me. Therefore, I'll have trouble delivering it, because I don't feel connected to the words. It's almost like my body rejects it. It's sometimes hard to graft on ideas from external sources.
EW: You talked about feeling the joke with your body, and I guess AI doesn’t have senses. It has ingested our words but not our physical sensations.
MB: It’s good at cringe, quips, and doom. Instagram caption, self-help things, quippy comedy, and apocalyptic, concept-driven sci-fi. The reason I think it is good at all those things is that they’re all about rubbing pre-existing ideas together. They’re quite abstract and conceptual forms of writing.
When you take it into other areas, like, for example, dialogue writing, it particularly struggles. ChatGPT won’t give you broadcast-quality dialogue, but it will give you something that approaches it more than it could have a couple of years ago.
The big problem with [AI] is that a lot of what a scene needs to work is going on beneath the surface. It’s about subtext. It’s about changing relationships between two characters. It’s about the emotional connection between them. It’s about what they’re realizing, or what the scene tells you about human nature. I don’t mean that in a nebulous, sentimental way; those are all real things that the scene does.
I think it’s very hard for an AI to have a fine-tuned enough sense of those things to deliver them in a convincing way, and also do it in a way that feels truthful, surprising and doesn’t feel like something you’ve seen a billion times before.
EW: Because the AI is going to the mid, to the average. But what elevates great writing is something that makes a connection you would have never thought of.
MB: AI will draw our attention very rapidly to all the things that other people could have thought of, so it makes it clearer where the virgin territory is.
I think it’s similar to the dawn of the Internet. Social media roughly coincided with me starting out, so I’ve never worked in the pre-social media landscape. But when writing jokes for satirical shows, for example, you’ve got to be conscious to say something that hasn’t already been said by millions of people on X or Bluesky or Instagram. The bar for originality and authenticity has been raised for everything. So I think LLMs will lower the bar for entry and raise the bar for quality.
EW: When you were playing around with these tools, did you feel any magic at all?
MB: I played around with an earlier version of ChatGPT a couple of years ago, and it left me feeling very smart. I was like, ‘This guy's not gonna make it in this business.’ It was very wooden in an obvious way. It loved cracker jokes and puns, like dad jokes. The whole feeling of it was like, if you put a dog in a funny hat, the dog doesn't know why it's funny. It just knows that people are laughing. The obliviousness of the machine was part of what was funny about it.
I started playing around with it again, and I did have a couple of experiences that made me say, ‘Oh!’ This has been pointed out by loads of people, but parody is something that it can do well because it plays to the chatbot’s strength: combine this dataset with that dataset. Combine everything I know about the poet Rimbaud with everything I know about Starbucks.
Jokes are a way of signaling social status. So I think an interesting thing to think about is how that's going to affect the value of certain kinds of jokes. Because things that are abundant can become low value. So I think certain kinds of parody will probably become extremely tired because an AI can do it.
Also, I think with certain kinds of humor, part of what's funny is the effort that somebody's gone to to render this utterly ludicrous comic product. That gives it a sort of charm and a silliness. You appreciate the human folly of all this frivolity. But if you've done it through an LLM, that charm is gone.
EW: Like famous people taking ice baths. Bringing people together to do something crazy that would usually never happen.
MB: Similarly, with stand-up and other forms of comedy, the risk and the fact that someone is putting themselves on the line is part of what’s funny.
There are two slightly different things there. One is that things that the LLM can do easily all become tired and low-value. Even playing around with it had that effect on me. It’s a bit like a party trick: ‘Okay, I get it.’ The second point, a bit different from the effort and risk in comedy being part of the comedy itself, is: would a robot Charlie Chaplin be funny? Can you have a slapstick robot? Because the robot doesn’t care.
EW: It’s not going to get embarrassed. ChatGPT doing standup wouldn’t be funny because ChatGPT itself isn’t funny, even if it could write you a funny joke.
MB: People wouldn’t watch the robot Olympics, for example. People don’t watch the chess computer world championships. They watch the chess world championships because they’re interested in other people going through the journey of life and wrestling with the same things they’re wrestling with.
EW: Your recommendation to survive in an AI world is to be alive.
MB: I’m a competitive person, and I don’t like to admit defeat. For me, it’s about working out what the AI can’t do and trying to kick its ass. It basically comes down to doubling down on the things that have always made good writing: unique perspective, original insights, ideally a bit of life experience, a voice that feels alive and can’t be replicated, the sense that the person behind that voice is struggling with some of the same things we’re all struggling with. These are all things that exist outside of the AI dataset.
EW: I think that we as creatives should be able to go out and say, “I’ve used these tools and this is what it’s good at, this is what it’s not good at.” What else might AI be “good at”?
MB: I was speaking with a friend about people trying to compose orchestral pieces through synthesizers. That didn’t really work because it’s a different tool, and synthesizers took off when producers used that kind of techie, alienated quality to create a distinct aesthetic, which my friend described as the aesthetic of alienation.
I got interested in how that might work for writing. It’s tough to think about it in terms of comedy, because comedy is more warm-toned and human on the whole, but how might people start to take the “sound” of AI, if you like, and use it in a creative way? At the Edinburgh Festival, I saw one troupe do a sketch they claimed was written by AI. Part of the joke was the uncanny, repetitive, inhuman, slightly disconnected, fragmented logic. So I think you’ll see a bit of that.
Other voices, other rooms
⛔️ A thought on comedy and humanity from me (Kenn). George Carlin once mused about a classically atrocious joke: “You get to play with people’s little danger-zones.” The perfect encapsulation of the provocation that great artists can do and machiness cannot.
🤝 Last week, I spoke to a large, global consultancy about GenAI, which some saw as a threat to its entire business model. I told them to reframe what they’re selling, from “a strategy to enact” to “a relationship of trust” (similar to what Ann told us two weeks ago): AI provides answers; more valuable is devising the right questions. And since most people get their impresion of a brand not from direct experience with the product/service but through a screen, corporate tone-of-voice is even more essential.
✨ Madeleine noted: “LLMs will lower the bar for entry and raise the bar for quality.” In a professional services context, it can’t just be up to employees to ensure the quality bar is met. Leaders with the taste, ambition and care must also demand a higher standard. (Hence, we need CWOs!) In creative contexts, a new generation of curators might also become important signallers of quality.
Thank you for reading! Talk to us via chiefwordofficer@substack.com, and share this newsletter with anyone you think would be interested. Upcoming guests include a world-famous neuroscientist 🧠 and a preacher ⛪️…
Last week’s missive used AI for text transcription. This week Kenn added two deliberate typos. If you were forwarded this, subscribe below.