In case you’re wondering why I use the term “Stochastic Worship” to describe what I’m doing, here are a few thoughts.
I use large language models as an aid to composing and virtually performing contemporary Christian worship music across a range of styles and musical genres. I use the term ‘stochastic worship’ to describe this activity.
My use of the word ‘stochastic’ is to emphasise the fact that the outcomes – what comes out of a large language model in response to a given prompt – are essentially unpredictable, but they can be statistically analysed.
We use stochastic processes all the time without realising it. Your local weather forecast is an obvious example. We can never perfectly predict the weather. It’s a chaotic system. But we can make models which behave in a similar way, based on our statistical observations.
It’s a case of using a tool to do a job that I find I can do better one way than I could using more conventional methods. In the same way that we don’t insist that hymn writers only use quill pens and parchment or compose only for harpsichord and lute.
The important thing is to recognise the limitations of the tools you use, and act accordingly. In the same way that we still pack an umbrella to a picnic, despite the weather forecast and tell our children to wash their hands before dinner after they have been playing outside.
