Komplete Kontrol has conceptual "flaw"?

This mostly sums it up.

But still, I would not buy any 3rd party plugin if it is not NKS. It's a selling point.

I just recently tried a free synth, and while it sounded pretty decent, preset management was a mess. It had like 20 banks with 50 sounds each. No way to search for certain words, no tags, always clicking for a preset change. Every developer saves it's presets different, even if there is a dedicated VST3 preset folder. KK sums that up to one place.

What should be done:

VST3!!!!, window size, user presets (in Maschine you can at least change the saving folder, but still no tagging while saving and so on)

Other than that it's a decent software (someone said it's already 10 years old, damn I am getting old).

Some of those who convert non-NKS to NKS could tell you, what different problems they face.... It is difficult for experienced humans, forget ML....

This wave of ML/AI is already third or so. Deep learning has been superpopular 30 years ago and seen as thing that can solve many tasks. Then it evaporated on many problems....

The only difference, now, is that computers are roughly 1000 times more powerfull and dedicated chips even more. And theory of ML/AI has also advanced a bit.

Autotagging is pattern recognition, but NKS is not just tagging of samples. NKS is much more anything else than tagging....

ML might be ripe. Or hype, like at least twice in the future. It seemed to be ripe 30 years ago. And developers and users were in the seventh heaven. But....

Deep learning is not anything mysterious and advanced. Just few matrices/vector calculations. Nonlinear mathematical transformation from n-dimensional space to m-dimensional space. It works fine, if gets similar data it has been trained to. And it works terribly and almost unpredictably if unfamiliar data come....

Future will tell if the technology is ripe and usable in consumer´s segment without problems. Or if dispointment and hangover will come....

NI already uses AI technology in modelling the musical gear. So, they investigate ways how to use AI, neural networks and ML.....

It is more than 10 years. There was Neural networks hype in early nineties and it was based on papers from eighties, if not late seventies.

I have been crunching numbers and doing deep learning on powerful 386 using 387 coprocessor and having biiiig RAM (32 MB) and biiiig HDD (20 MB). There was big hype about neural networks, that they are solution to hard tasks.... Well, ......

Now, thanks to Moore´s law. It might happen. Computers or dedicated chips probably deliver power needed. But. To make it work well, one needs representative data to train on. And experts to decide what data and in which structure are needed.

Garbage in, garbage out... One can train NN on any data, but it is advisable to use data without mistakes and covering wide range of possibilities. We may speak about million+ items of data, that must be checked for correcteness and so on.

There is no battle between iOS and android. Android has 70% marketshare.

Chips for AI is something like adding floating point coprocessor to processor 30 years ago. Or extending CPU instruction set for vector calculations.... Or adding audio/video codecs to CPU. Or adding cryptography support to CPU. AI support is just another special purpose calculation unit added to CPU. Like many, many others before.

Honestly, the only reason I use Komplete Kontrol is because Native Instruments has such awful preset browsers in their virtual instruments. Beyond that, it offers little to nothing I don't get natively in my DAW (Quick Controls) or on other MIDI Controllers at the FW Level (Arpeggiator, Scale Mode, Chord Mode, etc.).

It does deliver eye candy, though.