My Take On AI
Where I Stand on AI-generated Content For This Newsletter, and AI-Generated Music
AI and This Newsletter
First, I am a real guy, and I’m writing this by typing with my own, personal fingers. Up to this point, nothing in this newsletter has been generated by AI. Second, it’s probably worth noting that I’ve worked in tech as my main source of income for since the 90’s, and I currently use AI quite heavily for work, and I’m quite happy to do so.
With those bits out of the way, I do not have any plans to turn this newsletter over to some kind of bot and become completely uninvolved in this newsletter. I have the knowledge and understanding of the technology to do this, but I won’t, because what I’m writing about here is genuinely a passion of mine, and I really enjoy sharing my enthusiasm and excitement toward playing guitar, and everything that goes with that.
I haven’t, but might someday use AI to help with:
Editing. AI seems genuinely good at editing in a way that existing automated grammar checkers are not good at. I have actually worked as an editor of a newspaper and two magazines as past side hustles, but editing one’s own work is harder and requires more discipline, and AI can be genuinely useful there.
Ideation. It is way, way easier to keep up a daily or even weekly newsletter when you have a very strong backlog of ideas. Being that I have a day job, I can’t literally sit around ideating about guitar-related topics all day. AI is great at ideation in all contexts I’ve thrown at it, so I might try throwing a prompt at an AI like “Think of 50 topics related to playing guitar in some way that I could write about in a newsletter”. Then, from those 50 topics, I might choose 10 that hit me like “Oh, yeah, I could totally write about that!”
Research. I did not use AI to write the article about strings. I did a lot of Googling, sifting random (or not-so-random) sites, validating, verifying, cross-checking, etc. I referenced probably 10% of the actual content I looked at. The other 90% of the content I looked at was either stuff I couldn’t verify, stuff that didn’t matter, stuff that didn’t have enough detail to be useful, etc. I have used AI for research at work, and I know it would’ve made that article take 1/10th the time I spent on it. I could easily prompt AI to go do the research and cite its sources, and quote the relevant bits for me, and then the ones that look like they have particularly interesting or relevant quotes, I can follow the link to the site and validate that the quote actually exists on that site, and that it makes sense in context.
So, with writing, it’s not the actual act of writing that takes all of the time. It’s all of the tasks and little bits of work that happen around that process that take most of the time. A lot of that time is ideation, research, and editing. I’ll do the writing myself.
I know some writers flip that script: they just edit what is written by the AI. I’m not a fan, because the writers voice is completely replaced with some generic obviously-AI voice. Even when you tell an AI “write using x tone, as someone passionate about y”, etc., I can always tell it’s AI. For factual information content where the focus is being accurate, I’m not sure I care as long as it’s fact-checked. But for passion projects like this, I think the author’s voice is really important, and so I’ll do the writing myself.
AI-Generated Music
Practically speaking, I have heard AI-generated music that is kind of impressive in a “I can’t believe AI can do this” kind of way. But, I’ve never been impressed in a “this song is going to be in regular rotation on my playlist” kind of way. I’ve never downloaded an AI-generated song to my collection. My interest in it, or maybe tolerance for it, is purely academic. It’s a curiosity.
Conceptually speaking, I’m not really in favor of AI generating music in a context that takes jobs away from actual living, breathing artists. So, I’m not going to go to a movie with an AI-generated soundtrack, for example. I’m also not going to watch a concert where a singer is on stage singing over an AI-generated backing ‘band’. Not because I don’t think AI either is, or will be, generating good-enough sounding music, but rather on the principle that artists have a hard enough time making a living and we shouldn’t be making it harder.
And that’s basically that. I don’t have a big long diatribe about the quality of the current generation of AI-generated music, because it’s not super relevant: it’s going to get 10x better over the coming 12 months anyway. It can create pretty good lyrics from what I’ve seen, but again, I think I’d be more likely to use it as a means of ideation (“show me 50 titles in genre x that touch on topic y”), and then write the songs myself. If you were someone who wanted to practice cowriting with a lyricist, maybe that’s good practice. If you are a songwriter but need a better quality demo produced, maybe you could architect the song & have AI produce the demo.
“Wait, isn’t that taking jobs from people?” Not in my case. I’m a songwriter. I currently don’t demo my work because I wouldn’t be able to justify the cost of producing one, so I’m not someone who was ever a candidate to give anyone work, and I never have given anyone work in that context. But I see the point. If I can do create my demo with AI, lots of people who can justify the cost can also do it. It’s a quandary. So far, I have not considered using AI for this, btw. I’ve just collected all of the instruments & equipment needed to do a demo & I expect to do that myself if I ever get around to it.