Has anyone invented an AI "uber-brain" plugin yet?

[cron]

edit: will save that one for Off Topic

[Tj Shredder]

Yikes! I'm in danger of being misunderstood here... I probably shouldn't have put AI in the title!
The "amazing" I'm referring to is more to do with cataloguing and navigating my increasingly huge palette of samples, midi files, VST instruments and effects... getting "intelligent" suggestions that speed up and actually enhance my own creative process

Haha!
Well, you will have a lot of fun being misunderstood by an anonymous programmer's machine learning algorithm that will slow down and confuse your creative process by showing useless associations among your disorganised palette and uber-cluttering your brain.

You need this:
viewtopic.php?p=8613488#p8613488

No he needs an AI to make that categorization while he is sleeping… That tool is only useful if you start with your own sounds and do the categorization yourself by hand. He want the AI to do the boring part of that work… That is exactly where those models shine and can help…
Even if the AI doesn‘t categorize all sounds correctly its still thousand times faster and probably fails less than you would do as a human… It might have a bias based on the taste of who ever choose the training data though…

[Michael L]

Even if the AI doesn‘t categorize all sounds correctly its still thousand times faster and probably fails less than you would do as a human… It might have a bias based on the taste of who ever choose the training data though…

but AI actually fails when it:
- "doesn‘t categorize all sounds correctly"
- has a "bias based on the taste of who ever choose the training data"
- is wrong a "thousand times faster"

We have three choices in this use case:
- organise based on our own metadata, which builds familiarity
- adapt to the "wrong faster" organisation of an AI programmer
- stay with our current system and not waste time organising

Each choice serves a different creative process

[Tj Shredder]

Its not wrong in 100% of the time maybe 5%. If I do it I am wrong 50% of the time because I am not finished with it and didn‘t categorize most of the sounds at all…
This isn‘t theory. Atlas does it with drum sounds and my experience is, that I would not use 90% of my sounds at all. I have better things to do than categorizing a huge amount of sounds 100% correct (I would even fail with that…)
Deep neural networks can take away the boring non-creative parts of work. They cannot replace my creative expression, and nobody would want that. They have nothing to say, they have no emotions. Especially in music all is about emotions…
But the tool the op is seeking exists, though I don‘t know how well it works, as I don‘t own it… Neural nets are getting better and better over time…
https://www.sononym.net/

[Michael L]

Its not wrong in 100% of the time maybe 5%. If I do it I am wrong 50% of the time because I am not finished with it and didn‘t categorize most of my sounds… and my experience is, that I would not use 90% of my sounds at all.....

But the tool the op is seeking exists
https://www.sononym.net/

Some people collect more sounds than they could ever categorise or use.

Is that a problem AI can solve?

Sononym has 28 predefined categories, while the AI in AudioFinder has 443. Its target is 80% accuracy, so "Female Vocal Phrases" will sometimes include rain and thunder, birds, waves on a beach, cicadas and crowd sounds.
You need to spend a fair bit of time fixing search algorithms or results in both programs.

Does AI just trade one set of problems for another?

[jules99]

A much more practical application of AI in the short run is training an LLM on a manual and FAQ section and have it act as an interactive support agent. Think clippy but actually smart. So you can "ask" Ableton or Cubase how to do this and that or why this or that issue is happening and the built-in AI will draw up a conversational answer based upon knowledge drawn from the manual, the FAQ section and (if properly set up in terms of user privacy) relevant forum answers.

This would make any onboarding make so much more fluid for beginners. And it's in no way a replacement of making music or learning a tool.