ChatGPT, my drunk friend

In spite of, or because I’m a Xennial who has been faffing around with this kind of thing since the 90s, I view new technology with a certain level of skepticism. I’m all for innovation and futurism, but I also don’t make it a habit to hang out on the bleeding edge unless absolutely necessary. The platform can be unstable, the documentation often sucks, and the community can be polarised to the point where the signal-to-noise ratio is unproductive. But I’m also a curious optimist and will happily sniff the Kool-Aid from time to time to get a feel for where we’re at.

With all the buzz around AI at the moment, one of the things that has been coming up a lot as a creator is making use of AI tools to help generate SEO-friendly content to supplement the primary podcast/YouTube episodes. There are several services out there, including ones specifically for podcasters, but I’ve had mostly mediocre results. I’ll write that exploration up later, but the outcome so far is that I feel the raw material that gets generated for my content is either so off base that it is unusable, or would take more time to finesse than to create from scratch.

ChatGPT though, that’s interesting, and kind of familiar. I’ve hung out with ELIZA before and she was great fun. I’d like to see where we’re at now with actual machine learning in the mix.

Clickbait me

I’ve spoken previously about posting reels/shorts of my episodes on social media,¹ you can see an example of this on my Instagram for @steampoweredshow.

What I’ve started to do to speed up my copywriting workflow is to take the transcripts of these clips, so about 60 seconds of text, and ask ChatGPT for a punchy heading and a one-sentence summary. I’ll mash the generate button a few times until I hit gold, or there are enough bits to pull together into something I like myself.

I’m not naturally predisposed to writing clickbait, so asking for something in that tone can return something cringy along the lines of what you might see on the cover of The Fail, but I can walk it back.

It also isn’t terrible feedback when it misses the point entirely either, because I can review whether that particular clip really sounds that way and go edit it, or pick something else to highlight that might be of higher value.

Go home, ChatGPT, you’re drunk

This one is a bit more fun (to me).

I’m participating in a content cohort with the delightful Matty Staudt (@mattystaudt) of Jam Street Media for leveling up my skills and content for STEAM Powered and The Rejection Economy. He has a wonderful sense of humour, and is so generous with his knowledge and experience. Honestly, it makes me want to be able to hang out in LA again.

One of the things Matty has suggested we work on is the inclusion of additional content segments separate to, in my case, the primary conversation, that aside from being a nice little break for a longer chat, can be used as standalone content.

For The Rejection Economy, I was looking for famous or historic rejections, because fun facts are fun, history can repeat itself, and all that inspirational goodness. But research for this was while entertaining, incredibly inefficient. There are too many rabbit holes here, and this won’t be sustainable to do for every episode. Google was also surfacing tonnes of lists of famous rejections in tech and otherwise, but all were rather repetitive, and none I felt were contextually appropriate to the guest whose episode it would be for.

Given everyone is talking about how ChatGPT will change your life, I gave it a go:

Which tech entrepreneurs were rejected but later became successful?

It gave me a summary of the rejections of Steve Jobs, Jack Ma, Arianna Huffington, Brian Chesky and Joe Gebbia, and Jan Koum.

Most of which were on those aforementioned lists from Google. Yawn.

But hang on:

Arianna Huffington: Arianna Huffington, the co-founder of the Huffington Post, faced numerous rejections when she first started pitching her idea for a news and blog site. One investor even told her, "What you are proposing simply can't be done." Despite the initial setbacks, Huffington went on to build the Huffington Post into one of the most popular news sites on the internet.

Intriguing. She was rejected for pitching a news and blog site? Really? This piqued my interest because it’s contextually relevant to my guest, but I don’t remember hearing about this.

Google turns up nothing on the quote "What you are proposing simply can't be done.", and I can’t find anything else on that rejection. Rejections relating to getting a book published, yes, but nothing relating specifically to founding HuffPost or any other news and blog site.

Can you cite a source for the one about Arianna Huffington?

I didn’t think this would really work.

Yes, Arianna Huffington's story of facing multiple rejections when pitching her idea for the Huffington Post is well documented. Here's a source for further reading:

Huffington, Arianna. "How I Was Rejected by 36 Publishers." The Guardian, Guardian News and Media, 23 Feb. 2011, https://www.theguardian.com/media/2011/feb/23/arianna-huffington-rejected-publishers.

The link 404s, and if you search that title, sure stuff comes up, but again, it’s about the book. Wayback Machine also has nothing on that link, although that’s not guaranteed proof.

That isn't a real citation.

Yeah, I’m not sure if I should be anthropomorphising it, but I apologise to my Roomba for getting in the way, so…

ChatGPT apologised for the error and proceeded to generate another invalid source.

Can I have another 5 different tech entrepreneurs who were rejected and later became successful?

This time I got Walt Disney, Mark Cuban, Kevin Systrom, Reid Hoffman, and Vera Wang. Not all tech and not all accurate.

The Disney one is interesting though, and it occurred to me that my guest and I are both comic book nerds. So…

Are there any rejection stories for the founders of Marvel Comics?

Here I’m told that Stan Lee and Jack Kirby founded Marvel and that Stan Lee created Captain America but the concept was originally rejected. False. Cap was created by Joe Simon and Jack Kirby and the character idea was not rejected.

I’m also told that Jack Kirby was once fired from a comic book company for drawing characters that were “too weird” and was rejected by DC where he hoped to work on Superman. The former I have found no reference to, and the latter is also pretty much false. He left Marvel for DC and did end up working on Superman, but there’s an interesting story here about Kirby’s work on Superman’s face.

A functional drunk, maybe

I’m still getting a feel for what I can and can’t use ChatGPT for.

So far, it’s great for summaries of short content and experimental word-smithing. There’s a character limit for long-form content and I’m dubious about how good it will be at dissecting that in any case, but if you have a short section of work you’re not vibing and want some ideas of how to re-word it, it’s pretty great. I had excellent success at rewriting my intro for The Rejection Economy and got a few extra bits of content out of it that I’ll be tinkering with later.

For fishing for content like I did for the podcast segment, not so much. I feel it got me there faster than googling my queries because the fact-checking invariably served up the truths it might have derived its fabricated information from, which in turn inspired another route to explore. When I googled the old-fashioned way, I kept getting served the same lists over and over from different outlets (a different problem, I suppose) which created dead ends.

Ultimately, my fishing expedition was successful because looking up whether Captain America was actually rejected pointed me to something usable about Spider-Man. But whether this will be a time-saver going forward? Who knows.

If you’ve used ChatGPT for any of your podcasting or non-podcasting workflows, how’s that been going for you? If not, give it a go. If only for funsies.

Stay curious,

— Michele

---

Footnotes

1. As a small update on that front, my metrics are still inconsistent, but I am persevering. One week I’ll think I’ve started to get the hang of it, and then the next, no, no I don’t. I’ve stopped number-chasing as a result, but if something does perform better than normal, I’ll do a quick debrief to see what might have prompted it to determine if it’s something I can actively replicate. For the most part, the answer is currently ‘not really’. WIP. ↩