The Andrew Mayne interview: How to succeed as a polymath

In an age where most careers converge towards specialization, genuine Renaissance-style polymaths have become increasingly difficult to come by — but Andrew Mayne comfortably fits the bill. Bestselling author, science educator, original OpenAI prompt engineer, inventor, founder, coder, and protective-shark-suit designer all appear on his packed resume. The sky, it seems, is Mayne’s limit.

After high school, he achieved his childhood dream of becoming a magician. Then he decided to branch out — and branch out he did. Many of Mayne’s subsequent endeavors emerged from his childhood passion. He developed an educational program for public schools that teaches critical thinking skills through magic tricks, hosted his own reality TV show, Don’t Trust Andrew Mayne, and — walking in the footsteps of daredevil illusionists like Harry Houdini — designed a light-reflecting scuba suit that enabled him to safely swim alongside great white sharks. 

Mayne has taken other risks, too. In his 30s, he decided to try his hand at creative writing, going on to author a series of bestselling thrillers. He also taught himself to code neural networks — a pastime that helped him perceive the potential of OpenAI’s work on ChatGPT long before the rest of the world took notice. When the company approached him to become one of its first prompt engineers, Mayne didn’t need to ask what the position entailed. He agreed, and the rest is history. 

Mayne sat down with Big Think to discuss how he navigated the many twists and turns of his professional life, kept track of the golden thread connecting his various projects, and overcame the fear and doubt that prevent many of us from following our most ambitious dreams.

Big Think: You were one of the first people to work with OpenAI as a prompt engineer for ChatGPT. How did you spot the potential of generative AI before the rest of the world caught on?

Mayne: I’d gotten into AI about seven years ago and taught myself to train small neural networks and image generators. At the time, most applications were terrible. When GPT-2 came out in 2019, I was excited because they said it could make coherent text with a semblance of intelligence. I went to GitHub and poured over every publicly available line of code. I then talked about it on podcasts, and I guess somebody heard that, because when OpenAI was getting ready to release GPT-3 in 2020, they asked me if I’d like to play with it. At that point they just thought I was an author and a TV personality. They didn’t know that I’d spent the past years building my own AI systems. 

Big Think: What are some qualities of AI that people continue to either over- or underestimate?

Mayne: I think there are lots of opportunities to use generative AI in smaller, simpler ways. For example, I have to send a lot of emails for my work — many of them straightforward replies to requests or updates. I would not recommend we use ChatGPT to automatically respond to every email, especially those from high-level customers. It simply doesn’t have all the context necessary to do that without hallucinating. That said, it could be used to auto-generate drafts. Or we could build interfaces that help you build emails step-by-step.

Try Big Think+ for your business

Engaging content on the skills that matter, taught by world-class experts.

Request a Demo

As for what we overestimate, I think that sometimes these new technologies can feel miraculous, and people get into trouble because they perceive them as being all-knowing. Again, we need to understand how much context matters. We expect things from AI that we wouldn’t expect from an intelligent person. “Watch it write this PowerPoint deck based on a single prompt.” Who is that deck for? I can ask ChatGPT to extrapolate sentences, but in order to make a great presentation, I first need to know more about the subject I’m presenting on. Often, people just use generative AI to take a little bit of data and add a lot of fluff, which is useless — and a waste of time. 

Big Think: You’re known to use AI as a tool for your fiction writing. What are some tasks that ChatGPT excels or flops at on this front?

Mayne: AI has become better at writing short stories in recent years, but it still can’t do novels, partially because it doesn’t have a personality. I love Stephen King, because he’s a crazy guy in Maine writing horror stories. Someone might train an AI that can write books as good as, or even better than him, and maybe find an audience — there’s an audience for everything — but it won’t be the same. The real value is in the person, the experience.

Personally, I mostly use AI for research. I grew up in South Florida and if I want to write a novel about, say, a place I’ve been snorkeling, I can ask ChatGPT to help fill in my fuzzy memory. I don’t use it to replace the research, just augment it. I also used AI to build my own, personalized dictation tool, which works great.

Big Think: What’s your take on the art versus AI debate — on the notion that AI represents the death of creativity?

Mayne: As far as the quality of the writing is concerned, prompts go a long way. For example, students can fool their teachers by telling the AI to pretend they’re a Minnesota high schooler, and the word choice will change enough to the point that it’s probably not going to trigger AI detectors.

The same goes for creative writing. An Agatha Christie novel is going to have a smaller vocabulary than Edgar Allen Poe’s work, and you can train models to be more careful with their word choices. I could do that. But then, what am I doing? I’d have a conversation with ChatGPT about my books, but I wouldn’t have it plot a story for me. I don’t want to avoid the work. I love the writing process. It’s like music to me.

Big Think: You wear many different hats — do you have any tips for people who aspire to be multi-tasking polymaths?

Mayne: I was constantly trying out different things when I was a kid, and failed spectacularly at many of them. At 15, I wanted to be a magician, an inventor — Tony Stark, basically — multiple things at the same time. After high school, I decided to focus on becoming a magician. But after several years of touring, I realized I didn’t like having to say the same words on stage every night, or pack up my things and fly to the next destination the following morning. The life just wasn’t as romantic as I thought it would be. Still, the skills I picked up — carpentry, logistics, moment-to-moment problem solving — helped me branch out. I worked for Penn & Teller, and developed an educational program for teaching critical thinking through magic.

The same goes for my work with OpenAI: I did not look at ChatGPT from an engineering point of view, but from a language point of view, informed by my experience as an author. Basically, the roots of each project sprouted from the previous one. I’d also say: don’t be afraid of looking up the frontier, professionally speaking. It’s a great place to be, in part because it’s not crowded. Sure, I was one of OpenAI’s first prompt engineers, but it wasn’t a super-competitive field at the time. 

Big Think: Being a polymath can be a bit of a double-edged sword. On the one hand, you can bring in fresh perspectives from different disciplines. On the other, you risk spreading yourself too thin. How do you balance that?

Mayne: I’d love to say that all my books get turned in on time, that I flip a switch inside my brain when it’s time to write and don’t get stressed out about other things. But I do get stressed out and distracted. I’m not wired differently from anybody else.

I get rid of distractions. The computer I’m using right now has no notifications. I put my phone away, don’t use Facebook or Instagram, and don’t watch TV during the day — except when I’m on the treadmill. Limiting the amount of time the world can take from me gives me more time to do things.

Now that I have my own firm — with OpenAI being one of my biggest clients — I finally have the freedom to experiment. But even then, I still have to practice discipline. I’ll tell myself, “Nope, you can’t play with this or that for the next couple of days; there are other things you need to finish first.” It’s all about delayed gratification, accepting that not everything will happen the way you want, and choosing your priorities carefully.

Big Think: As an avid diver who swam with great white sharks, what’s your relationship to fear?

Mayne: I can be in a swimming pool by myself and jump out when I hear the tiniest splash. I get scared easily and my capacity for terrifying myself is very high. But I’m also very stubborn. After my dive with the great whites, the crew told me they were convinced I wasn’t going to do it. I said I would rather have been eaten by a shark than given up and face that personal humiliation, because it was a decision about who I was as a person. 

It could be regarded as stupidity — which it partially is — but there have been a couple moments in my life where I can hear the weaker version of myself saying, “No, I’m not going to do it,” and the version I want to be responding, “Yes, you can do this.”

By the way, swimming with great whites is dangerous, but not in the same way we may think. We’re not their intended food. They’re not on a mission to eat all humans. They might come up to you to see if you’re edible, but they’re also cautious animals. Often when they bite people, it’s in murky waters, and they’re just trying to see if they can eat you. In other words, the situation I put myself in was somewhat controllable, not as random as you think. But I’d be crazy to think I understood them or that I wasn’t at risk.

Big Think: When it comes to calculated risk, how do you figure out what is and isn’t ultimately worth doing?

Mayne: Before I went to work with OpenAI, I would build a new project every two months, thinking each time it was going to turn into a company. I’d take it all the way, only to realize, “Nah, not for me.” This drove those around me nuts, because I would get them all excited about an idea, then abandon it. 

The problem was that Iwasn’t truly excited about the ideas myself — I needed other people to tell me they were great in order to stay motivated. That’s not a good way to operate. If you’re trying to get other people to sell you on your own idea, it’s probably not the right idea for you. That doesn’t mean it’s a bad idea — just that it’s not your idea.

There are a lot of opportunities I’ve turned down. I could have built a work-cheating app — didn’t want to. I could have built bots that let people talk to AI girlfriends — didn’t want to. I followed Bitcoin early on and could see it would do really well — but I didn’t invest, because it just wasn’t for me. I’d say you have to think of it like a Venn diagram: what you want, what your skills are, and what the world wants. Find the intersection.