I mentioned this on our About page, but thought I’d go into a bit more detail. I love AI, but like others, I have some very strong opinions on what it does and how it should be used. Many of the controversies over this new technology are misconceptions about how it’s trained or it’s resource use, or the fault of bad actor users who abuse the technology and no fault of the technology itself.
I put up a policy and philosophy page to clarify much of this, but thought it would be a good idea to preemptively answer a few questions specific to VISIØN and my use of AI.
FAQ (or questions I imagine may be asked frequently:
Q: Does the AI write the songs for you?
A: No. Or at least not always. I write the lyrics and direct the song flow. I choose which instruments to use, choose what key the songs are in, what chords to use and where to use them, I determine the overall structure of the song. Due to the nature of how generative AI works though, all this work ultimately functions as a very firm suggestion to the AI. Most of the time I get what I ask for, and a little extra that I didn’t. Sometimes this extra is good, sometimes not. Sometimes I don’t get what I ask for and need to ask again a little differently. It’s a lot like working with another musician you hand your music to and who sometimes insists on “doing it their way” until I can convince them otherwise. Ultimately, I make the decision what to keep and how to use it. It’s a lot like throwing a bucket of paint at the wall and wanting it to splash in a specific pattern. Sometimes I get lucky, sometimes I have to throw a lot of paint, but every time I get better at throwing paint and can more reliably end up with what I have in my mind appearing on the canvas.
Q: If you are using AI, where is the human element?
A: Well, I’m right here, being human. I do the writing, the direction, make the decisions. The AI parts of my process generate the sound of the instruments, but I choose the sound of the instruments. The AI, at least the way I do things, is just a really advanced synthesizer.
Q: How much control to you really have if the AI is doing all the work?
A: Well, the AI isn’t doing all the work. It’s doing very little of it actually. I have… let’s say nearly complete control. When I write a song and give the AI the instructions it needs to produce it, or just the part of it I’m asking for, it’s like asking an extremely talented toddler to do a trick. A lot of time that toddler is really excited and does the trick eagerly and with skill, making a few mistakes but still giving me what I wanted. Sometimes that toddler decides to show off and do something else… improvise, or even do nothing. I accept responsibility for this undesired output though, as perhaps I didn’t phrase my request well enough or accidentally used the wrong term in my instructions. In the end, if the AI didn’t do exactly what I wanted, I can disregard, rephrase, and ask again. And again. Until what I asked for, what I want and have in my mind, is there to then put in my DAW and use how I want to use it. Sometimes what I have in mind just isn’t as easy to describe and I get something close enough, and I take that to my DAW and change it manually. In every case though, I do the work.
Q: Why disclose this, aren’t you afraid of being accused of making AI slop?
A: To tell the truth, I am afraid of that harsh accusation. But, since I am using AI in at least part of the process, that accusation is bound to happen at some point. Since this isn’t slop, it’s actual human-led music, I’d rather be fully transparent and open about my use of AI and the ways in which I use it. I don’t think the use of AI in art is a bad thing. It can certainly be abused and used in unethical ways, but that’s not me, and I want to not only clarify that, but be able to prove it if challenged.