Re: Anybody else playing with GPT4 and Lilypond?

[Saul Tobin]

I think you may have that impression based on GPT3.5. GPT4 is already being used to generate working non-trivial computer programs based only on a brief text description.

On Wed, Mar 29, 2023 at 3:58 PM Alexandre Loomis <alexandreloomis@gmail.com> wrote:

> given some of the other impressive things it can do

I think that's been exaggerated. It's very good at generating plausible-sounding text responses to prompts, everything else looks cherry-picked.

On Wed, Mar 29, 2023 at 3:54 PM Nate <therealnathanstewart@gmail.com> wrote:

Hah yes. It once said \begn{music} and i said "are you making this up?" "I'm sorry, you're correct. The start tag should be \begin{lilypond}.

Its super handy but you have to watch it. It can be a pathological liar. I asked it how to do something on the Akai Mini Play and it said to use this button  On the upper left corner. when i asked for clarification instead of admitting it was mistaken it said it was white and next to another button. Twice it doubled down before admitting it was wrong.

On Wed, Mar 29, 2023, 6:44 PM Saul Tobin <saul.james.tobin@gmail.com> wrote:

I've seen some examples of other people succeeding in getting ChatGPT with GPT4 to compose simple music in other text based music formats. I've had limited success getting it to output Lilypond code. It is able to correctly structure the code with a score block, nested contexts, and appropriately named variables, and bar checks at the end of each measure. It seems to struggle to create rhythms that fit within the time signature beyond extremely simple cases. It also seems to struggle a lot to understand what octave pitches will be in when using relative mode.

It also seems to have a lot of trouble keeping track of the relationship between notes entered in different simultaneous expressions. Just asking it to repeat back which notes appear in each voice on each beat, GPT4 frequently gives stubbornly incorrect answers about the music it generated. This makes it very difficult to improve its output by giving feedback.

I'm curious whether anybody else has tried playing with this. I have to imagine that GPT4 has the potential to produce higher quality Lilypond output, given some of the other impressive things it can do. Perhaps it needs to be provided with a large volume of musical repertoire in Lilypond format.