It might be more precise to say "learning about patterns". There are AI models that have been around for a few years that are designed to generate distributions (patterns) that are not seen in the training dataset. Generative Adversarial Networks (GANs) are very popular, powerful, but are fiendishly difficult to train, for instance.... and "learning" the patterns
I'm not familiar with Band in a Box (although, I just watched a quick video).So can I imagine this a little bit like "Band in a Box", just much more advanced and for orchestral JW-style instead jazz and pop styles?
I think you'd enjoy playing a game I played recently; The Talos Principle.TBH I was a bit worried when I first saw this, but I do strongly believe that Art is all about how you can express your emotions that no matter how we try to put in fixed words (such as sorrow, fear, joy etc...) would still be experimented differently and uniquely by each individual- hence interpreted and produced in a unique way belonging to each artist. This could be a useful tool, based on A"Intelligence" but not consciousness. Thankfully, human audience (receptors of our music) interpret & taste our art the same way as we do produce it - based on their own conciousness and emotions to a great degree. This is a complexity that goes beyond the capability of any AI I believe. Humans with artistic sense(that don't need to be artists themselves) will always be able to differentiate an original "Monalisa" and prefer it over a printed one.
I'm not familiar with Band in a Box (although, I just watched a quick video).
From the looks of things, they're quite different animals. Aiva is capable of composing music in a variety of (mostly orchestral) styles with near infinite permutations of melody and progression, rather than being a list of mix'n'match presets.
I gotta say, as a human who composes music, this doesn't sound good...
It will not take long for AI programmers to beat the four chord epic guys. We are there within a year maybe. A robot will throw hz out like a wet towel. After two years the AI is making horizontal development. To get to JW level, maybe three to four years from now.
”Written by a human” will maybe be something to add. But then we also get where anything you write will be suspected as done by AI. You make a masterpiece and critic say it was done by AI.
In conclusion. We need to publish our masterpieces within next three years, otherwise you will be suspected as having used AI.
Time to work harder.
Weird what is generated through the AI and what still requires human labor in that video. A lot of the most mechanical tasks—copying music, re-entering notes, etc.—are being done manually but a lot of the editorial work required to make it sound like passable music as well. Basically, in this scenario the human is serving as the assistant to the AI (and saving the AI's ass!). What a world we are building!I only watched this one video, so I don't know what kind of orchestrator AIVA is, but this sure shows a lot of the human hand.
"And I imported the MIDI into Sibelius... and I cleaned it up a bit just for the sake of this presentation. At this point you can do whatever you deem necessary. If you want to make adjustments to the orchestration, the production, etc, it really becomes up to you. ... I'm doing a bit of arranging... I'm doing some notational things just for my own sanity...
It's more practical at this point to think of AI as tools rather than people. We are far form an all-encompassing AI that can handle a bunch of complex individual tasks and then combine those tasks to achieve a goal. This AI generates MIDI data from other music, and does so in a way that is musically "passable" to serve as a palette of colors for a composer or orchestrator to use as they will.Weird what is generated through the AI and what still requires human labor in that video. A lot of the most mechanical tasks—copying music, re-entering notes, etc.—are being done manually but a lot of the editorial work required to make it sound like passable music as well. Basically, in this scenario the human is serving as the assistant to the AI (and saving the AI's ass!). What a world we are building!
I have listened to a couple of orchestral versions what that AI created and while I think it is mediocre stuff on the one side I am impressed still that that pogram whatever wrote that and actually I think the quality of the compositions are for most time on par with a lot of music out there done even by some working composers.
It will get better, but understand that this thing just listens to a song and spits out midi that sounds stylistically like it. It hardly "composes" yet, and you can still hear the original in the music if you knew it before hand.It will only get better, don't worry. This is just the beginning...and it is unsettling. It doesn't matter if it's going to be better - it will be cheaper and that's all it matters. Ironically, it will probably erase those sites such as Audiojungle first, which are largely responsible for 'race to the bottom', since amateur filmmakers and people who don't need top-end music for their projects will have access to unlimited versions of likely even cheaper music.
Here's an interesting website: https://willrobotstakemyjob.com/
It predicts 2% chance for AI to replace the composers, however, I wouldn't be so sure about that percentage.
Given the amount of editing required, does it actually speed time to finished piece? Is it relieving composers/orchestrators of drudge work? From the video, I don't see that it does. I mean, it's a fascinating technology to be sure, and it may or may not get better in ways that matter.It's more practical at this point to think of AI as tools rather than people. We are far form an all-encompassing AI that can handle a bunch of complex individual tasks and then combine those tasks to achieve a goal. This AI generates MIDI data from other music, and does so in a way that is musically "passable" to serve as a palette of colors for a composer or orchestrator to use as they will.