This video about AI music education at Berklee was quite interesting. It signals a deeper issue that I wanted to probe today.

Berklee College of Music now teaches classes in AI songwriting, and that’s a really dumb idea.

Source: They Teach AI Music at Music School Now…

Something Adam pointed out here that stuck with me and I spent hours mulling about it. The CEO of Suno says this about making music:

It’s not really enjoyable to make music now. It takes a lot of time. It takes a lot of practice. You need to get really good at an instrument or really good at a piece of production software. I think the majority of people don’t enjoy the majority of the time they spend making music.

– Source OG Source

I don’t want to pick on this guy specifically, because at least from a commercial perspective, he seems to have walked that back).

However, in it is a nugget. This view makes sense if you don’t value the process.

For craft oriented people, process isn’t an obstacle. That’s the raison d’être. This is why I’ve opined that people find AI compelling when it’s an output outside your own craft. I’ve called it the AI Gell-Mann Amnesia, the Knoll’s law of AI etc.

In other words, the more someone’s trained in a craft, the more they’ve developed their taste through the labor, the more likely they are noticing the averages, the flattening, and the “expected” in AI output. I can say that about writing, developers whom I deeply respect say that about coding, Adam Neely says that about AI music.

Let’s take the example of a home garden versus factory farming: Home gardens are craft centered. In your home garden, you are growing produce to feed yourself and your family. You will carefully choose what you want to grow from, the soil, the water, the seeds. You will put in the hard labor needed to tend, to weed out, you will check-in every day until you bear the fruits of your labor (sometimes literally).

Factory farming is output centered: You are focused on generating the maximum amount of crop for the minimum amount of cost. The crop is not the output. The output is a sustainable farming business.

In a home garden, the labor can be frustrating. Gardeners complain. Similarly, musicians get irritated. However, the frustration is meaningful. It is ingrained in the reward structure of the activity.

If you consider the descriptors used, it becomes clearer: In a home garden we talk about care, taste, attachment, ritual. In a factory farm, we talk about throughput, yield, standardization, cost.

As a result, AI works best where the output from AI is good enough to serve an über goal. Where they care more about the downstream impact rather than the artifact itself. Craft centered versus Output centered. If you consider the “skill process” as inefficiency, then it totally makes sense why you think generating that away is meaningful.

I worry this explains why people with higher organizational distance from work are framing how AI is a “replacement.” For example, most executives sit where that process, the craft becomes numbers. They talk in the the same set of terms that we used for factory farming. From that PoV, a tool that promised output without skilled labor is of course magical.

Similarly, in today’s AI-writing use cases: marketing copy, internal comms, status updates, LinkedIn peacocking: literary quality isn’t the goal. It’s a means to an end.

A looming contradiction: AI might be useful because it removes labor. However, that raises the importance of taste. Yet, it’s the labor, the process that trains the taste. What happens in a world where we don’t have enough people building their taste?