[Sugar-devel] Inquiry Regarding GSoC Project Opportunity.

[Walter Bender]

On Sun, Mar 17, 2024 at 11:15 AM Ritika Kumari <rkjane333 at gmail.com> wrote:

> Dear Mentors,
>
>
> I hope all is well. My name is Ritika, and I am going to submit a proposal
> to work on the project titled - 'Musical ideation through Generative AI
> <https://github.com/sugarlabs/GSoC/blob/master/Ideas-2024.md#musical-ideation-through-generative-ai>'.
> I am writing this mail to cross check my understanding about the project.
>
>
> The user will start the composition, say A, C, D. And this input will be
> passed to an AI system which will read this and suggest multiple
> combinations of notes. Something like:
>
> option 1: A, C, D, D, E, G
>
> option 2: A, C, D, G, C, A
>
> ....
>
> option n: A, C, D, F, A, G
>
>
> Let me know if this is correct?
>

I think we want to consider multiple levels of abstraction here. What you
are proposing is how to compose a phrase with some suggestions for
"auto-complete" from the AI. How it works will be very much dependent upon
the training data and the "temperature" settings. You may also consider the
rhythmic quality of the sound. Then there is the higher-level structure.
How do multiple phrases fit together? What about transformations of the
phrases, with inversions, transpositions, retrograde, etc.? The AI could
make suggestions at that level too.

>
>
> If this is correct, then few other follow up questions are:
>
>
> 1. Does the user need some minimum number of notes and after that we will
> be asking the AI system to suggest possible notes?
>
>
> 2. What are the metrics on which to evaluate the output from an AI system?
> I can think of adding some safety measures to ensure that the output
>
>  a. from the AI system is not harmful
>

Other than generating terrible music, I am not sure how any harm could be
done.


>  b. is relevant to the query from the user
>

As per my comment above, I think there are more options to consider.

>
> 3. How to grade the output from the AI system? My understanding is that
> the output from the AI system needs to be melodic. Is there a way to check
> this in an automated way on a large scale?
>
>
>
> Warm Regards
>
> Ritika
> _______________________________________________
> Sugar-devel mailing list
> Sugar-devel at lists.sugarlabs.org
> http://lists.sugarlabs.org/listinfo/sugar-devel
>


Hope that is of some help.

-walter

-- 
Walter Bender
Sugar Labs
https://www.sugarlabs.org
<http://www.sugarlabs.org>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sugarlabs.org/archive/sugar-devel/attachments/20240317/e8824e17/attachment.htm>