Ok bot, I will look up what knee-jerk means later.
2 Likes
Using intelligence as an argument is such a non-argument because intelligence is nonquantifiable.
And also when did anyone EVER ask you for your nationality?
Anywho, I am done with this argument.
2 Likes
It was a waste of time.
What would you want AI to do in Renoise?
I would want it to use its AI powers to see itself out of Renoise and never come back.
10 Likes
Iām against embedding/interfacing current āAIā (i.e. LLMs) in Renoise (pointless and useless).
That said:
- lot of good ideas, tools and modules came from ML (Machine Learning) ideas. Like for example the Mutable instruments Grids module ( Index - Mutable Instruments Documentation ) whose pattern generator is the result of a statistically compressed drums patterns big dataset.
- These times I use Claude code to help me code Renoise phrase scripts. I had to feed it the Pattrns API and some examples. It doesnāt do the creative work for me, but helps with the tedious parts. Iām happy with this setup, and would not want it integrated in renoise.
As a software developer, Iām a bit torn on the LLM topic.
Itās over-hyped, over-sold, toxic and LLMs flagships and companies are downright thieves (eat all the internet content without creatorās consent and resell a compressed latent space of it to masses). Itās also downright dangerous to the whole society, and donāt get me started on ecological and social costsā¦
3 Likes
big ?
thereās no machine learning involved here, just randomization among CAREFULLY HUMAN-PICKED numbers to get an humanly intended result.
You recognize how wasteful and toxic it all is, but still will feed it some doc you could read in-extenso in about 1h ? 
At the core of Grids is a map, that has been obtained through machine learning techniques, of the drum patterns used in a large collection of electronic music tracks and drum loops. The skeleton and structure of the rhythm is programmed by selecting its X and Y coordinates on the map - with smooth interpolation/morphing between points.
Ref: Manual - Mutable Instruments Documentation
Itās statistical pre-computation, like I said above.
You recognize how wasteful and toxic it all is, but still will feed it some doc you could read in-extenso in about 1h ?
I have a nuanced position (I know itās not very āinternet-successfulā 
). AI and ML was a significant part of my student time (20 years ago). I currently work in a company that use āAIā (statistics+physics+biology scientific computation, conceptually very unlike LLMs ; more deep research and much less hype-prone).
Unfortunately, āAIā means nothing (āNowadays, anything more complex than a multiplication is called āAIā ā). Today, people conflate it with āLLMsā, I get it. But their mailbox spam filters (bloom filters) and their automatic face recognition in digital cameras (Haar classifier) are AI too, and itās been here since 20+ years.
If I put myself in āword strict modeā, being āanti-AIā would be like saying āIām anti-mathsā.
Despite ābeing nuancedā, Iām very much closer to LLMs-skeptics than LLMs-enthusiasts.
(and I wonāt write much further on the topic, as I think I explained my ethos enough, and ultimately, we agree on āWe donāt want LLMs in renoiseā)
1 Like
only thing I can instrincly think of is not making beatsynced samples (using the perc/texture modes including linear, cubic, sinc interpolation not sound like total cheeks
AI would not realistically solve that issue without doubling Renoise in size and bloat it, or rely on some external services (which would in turn most likely abuse userās author rights).
A much better option on that matter would be to license zplane stretching (like Cockosā elastique, Abletonā complex/pro, Bitwig, ā¦) : https://licensing.zplane.de/
2 Likes