An AI dilemma (in Podcasting and Beyond)
An AI dilemma (in Podcasting and Beyond)
December 14, 2023, 10:51 pm
Filed under: art, podcasting, technology, writing | Tags: AI, algorithms, bots, congratulations, Goodpods, grant writing, podcasting, promotional text, robots, technology, writing
When I heard (in an Audio Drama group), about the AI descriptions taking over podcasts on Goodpods (a podcast platform), I headed straight over to see if my audio drama (The Dragoning) had been subjected to this treatment. It was not, so I moved on with my week, not thinking much of it. Then another audio drama group began to talk about how outrageous and wrong these descriptions were and how they were pulling their shows from the platform. So I went back to check and, still, The Dragoning was unaffected. But this time, I thought to check my other podcast, the audio version of this blog, and lo and behold, there was a whole bunch of text I’d never seen before.
There was a description of my podcast, a paragraph about who should listen to it and a summary of three episodes. And unlike the descriptions of my colleagues’ podcasts, it was pretty accurate. Actually, it sounded like a PR person got ahold of my work and went to town. It sounds like a pretty nice review. If a person had written it, I would be flattered. But a person didn’t write it so it just made me confused. Does the AI who created this text really think I’m insightful? No. It doesn’t think. Period. It doesn’t even know what insightful means. It’s likely just using predictive text to write a reasonable sounding bit of copy.
A lot of my podcasting colleagues were pretty upset about having their podcasts saddled with text generated by AI that they didn’t ask for or choose, and understandably so. I find myself more flummoxed by it.
The AI “wrote” these things in a very particular style. It says many things that I would never say about myself that I would, nonetheless, like to be said about me. It feels like Goodpods got me a publicist who makes me feel a little uncomfortable but who I wouldn’t complain about, if they got me more visibility. And I already have three times the listeners on their platform than I did before they added this made-up description. So…I’m very torn about this new development. When I read it, it feels as though this AI has made some interpretative choices, that it listened to my podcast and summarized what it heard. But, of course, that is not what happened. I don’t know how it came up with what it came up with but there was no interpretive work done by the robot. And if I understand the purpose of this text, it is to draw the attention of the algorithms that suggest podcasts to people. That is, the AI chose words that would get the attention of some other AI and theoretically put my podcast in front of more people. It’s the robots talking to each other. If it gets me some clicks, is it okay to let it do that?
The thing that becomes so plain here is how promotional language is so formulaic and meaningless that an AI can convincingly do it. This is probably one of the most productive ways AI can be used. It can handle anything that is, essentially, a formula without too much originality required. It reveals the simplicity and emptiness of advertising copy. Lord knows I don’t want to write things like this. If AI can do it for me, why shouldn’t I let it? There are a lot of bullshit things I have to write -promotional text, artist statements, grant applications. I mean — for sure an AI could do as good a job of writing the bullshit for a grant application as I do. In fact, it’s likely to do a better job, as it’s not afraid to use self-aggrandizing words and ideas and could be programmed to tell grant readers exactly what they want to hear. If there’s one thing I am not good at, it is talking myself up. And, it would seem, if there’s one thing AI can be good at, it is talking things up. And, in fact, they could use an AI to read all these bullshit grants, too. One AI could talk up the artist and another AI could search for the words it wants to see, i.e. “read” it. Artists need not set down a single word.
Is it more important that I write a grant application in my actual voice or that I use the phrases that are likely to yield good results? A good grantwriter knows what those phrases are and you can bet they make sure to include them in their grant applications. Is it possible this whole exercise of writing grants for artwork might be bullshit? If an AI could do it better than I can?
I’m in the middle of writing a grant now and I am absolutely tying myself into knots trying to invent the right text for it. I could, even now, probably plug some of my usual arts project language in to some large language model and end up with something usable. I won’t do that but I’m not sure I shouldn’t.
It’s clear that the podcast AI has thus far rewarded me for the formulaic language it made up for me. I have three times the listeners that I did. (Maybe. Are those numbers representing real people or bots? Does it matter?) So, if an AI program can get me more podcast listens, who’s to say I shouldn’t use it to get some arts funding, if it saves me pulling a muscle (a muscle made of bullshit!) every time I write one? I am fundamentally opposed to AI doing artist’s jobs. I am incredibly grateful that the Writers and Actors’ unions have fought to keep the robots from taking their jobs. But I’m starting to see the value in AI for the bullshit part of our jobs. If the AI knows which words to use to get other AI to move me up the charts, I’m grateful. There’s no virtue in me trying to guess what those words are to try and game the algorithm.
I think about how writing “Congratulations” in the comments on Facebook puts a post in front of more eyes. If lots of people write congratulations, it increases those eyeballs even more. These descriptions seem to be a version of saying “Congratulations” on Facebook. It’s all just ways to game the robots, I guess.
Do I feel good about living in a world where we have to think about how we manipulate the programs of robots? I do not. But I acknowledge that the likelihood is low of a real human person stumbling upon my work by chance, without the help of some robots. As much as I am a little disturbed by the language a robot invented, I think I might leave it there, just as an experiment. I feel very weird about having an unattributed piece of language on a page about my work but it’s also very hard to argue with, probably because it didn’t come from a person.
To be honest, I haven’t paid much attention to all the hype around AI. As an indie artist in an unprofitable corner, I really couldn’t imagine a scenario in which it would be relevant to my life. But those bots are full of surprises. And AI is so trendy, it’s fixing to show up everywhere. There were rumors that Goodpods was readying itself to get investors and that AI is a thing big companies use to get the money people interested. Say you’re going to use AI and you get more cash! We’re probably all going to be wrestling with AI here and there now. Would I let an AI write a blog for me? Hell to the no. That’s for a person to do. And that person is me. But would I let it write my grant application? I might. (I mean, I won’t. Please don’t reject my grant applications because I said I might let a bot do it. I’m doing it, okay?) In a way, this all just makes clear what is bullshit writing, for which we can accept the work of a bot, and what is writing we need a human for, a face for, a voice for. Maybe a bot could do this. But it would be meaningless without a person behind the words.
And in the latest development over on Goodpods, my moral dilemma has taken care of itself. I guess they had enough complaints about their AI that they just removed it all. Now, where there was once a puff piece, there are the three questions that the AI previously answered. Now it wants me to answer them? I am unlikely to. Can I just type in what the AI wrote for me? I don’t think I have the self-aggrandizing stomach for it. Also — I looked at my nine new listeners and I’m pretty sure they are all bots. Their user names are things like GuestUser947 and such. So… nothing’s really been ventured here. And nine bot listeners are all I’ve gained.
Want to see what a bot wrote?
This post was brought to you by my patrons on Patreon.
They also bring you the podcast version of the blog.
It’s also called Songs for the Struggling Artist
You can find the podcast on Apple Podcasts, Spotify or wherever you get your podcasts.
Every podcast features a song at the end. Some of those songs are on Spotify, Apple Music, my website, ReverbNation, Deezer, Bandcamp and Amazon Music.
*
Want to help me not turn to bots?
Become my patron on Patreon.
Click HERE to Check out my Patreon Page
Or you can subscribe to my Substack
*
If you liked the blog and would like to give a dollar (or more!) put it in the PayPal digital hat. https://www.paypal.me/strugglingartist
To help me pay off my trip to Crete, donate on Kofi — ko-fi.com/emilyrainbowdavis
Leave a Comment so far
Leave a comment
Originally published at http://artiststruggle.wordpress.com on December 15, 2023.
