What's new

Aiva - Artificial Intelligence Composition: beta starting today

Imagine a world where you had to "prove" your composition was original enough to receive copyright by submitting it to publisher's databases. If your work were found infringing, you've be required to assign a proportional amount of your rights (and profits) to the copyright holder.

That seems to be the general direction we're heading in. :eek:

I'd love to have a thing that let's me conveniently check what I might be unintentionally infringing on and protect me from potential future lawsuits if it doesn't cost too much. The alternative is a pure gamble and never being sure you're not gonna get sued because you happened to find the same melody as someone else.
 
I'd love to have a thing that let's me conveniently check what I might be unintentionally infringing on and protect me from potential future lawsuits if it doesn't cost too much.
There's no guarantee against lawsuits, because there's no minimum number of notes, and copyrighted elements of a musical composition can include melody, chord progression, rhythm, and lyrics. Even a chord progression can be copyrighted, assuming it were "creative" enough.

So if a publisher has enough material, everything new is infringing.

The tech companies eventually built large portfolios of basic patents that created a sort of mutually assured destruction standoff against patent battles.

Similarly, a searchable database of public domain songs might be useful to protect against lawsuits where a trivial part of a composition was claimed to be infringing.

Just some more (dystopian) thinking.
 
Back in 1984-85, a college classmate of mine was a computer scientist and musician, and we were both in NYC for the summer. He was working for Laurie Spiegel at the time, and was writing a simple program that would compose music algorithmically as part of his dissertation. We had the dystopian idea to get his program running on the college's mainframe, churning out every possible combination of notes and chords, and uploading them constantly to the copyright office (although no such capability existed back then).

Then, once our body of copyrighted works reached critical mass, we would establish "The Ministry of Music" and claim ownership of every possible combination of rhythm, melody, and harmony, requiring the entire human race to pay us for the rights to engage in any musical activity.

Sounds like that might be about to come true....
 
I think it's much more likely the we'll slowly, over time, become so accustomed to AI music that we will start to prefer it to human created music.

Call it the "boiling the frog" hypothosis of musical evolution (or devolution - you pick the noun)...

It will start slowly with catchy "can't get it out of your head" jingles in commercials, on websites, and in video games; places where you're not focused on the music but it's there subliminally chipping away... then, gradually, in bits-and-pieces, being transitioned into more mainstream entertainment vehicles like radio, TV, movies, Indie venues like Youtube, Twitch, etc... then before you know it you'll be humming an AI song all day long that you can't get out of your head... or, then again, maybe not... :sneaky: ;) :cool:
 
Last edited:
Ugh these conversations about AI are so boring. Music has been shitty, uninspired and full of basic plagiarism long before AI came along. People need to stop acting like humanity is some well of creative perfection. Out of all the music in the world only some of it is truly great. The rest is people trying to copy that greatness, or purposefully being shitty in opposition to greatness.

As a performer and songwriter, I know AI will be coming for my job. Tough shit. Every industry faces this and the smart people learn to adjust.
 
Hello. My name is Jean John and I am a New York based classical and jazz composer. I signed up for this forum solely because I have been in shock ever since I received an email from AIVA and was vigorously searching through internet to find if this very very serious matter is being discussed anywhere. Luckily I found you guys.

To say that I am personally terrified is an understatement. I have contacted many composing groups already and warned them to not share any information until many more regulations are being presented. Music is human and should stay human. Please do not share any information about your creative process with AIVA, believe me, you worked too hard to get where you are, sharing your creative process for artificial intelligence purposes is like shooting yourself not in the foot, but in your heart. This is the email I got:

I am ***** from AIVA, AI composition assistant. I recently came across your profile and was wondering how you come up with new ideas when creating music.

I am asking this because we are a team of composers and engineers that created AIVA, and we are currently looking for people to try AIVA for free and give us feedback to make it a useful creative assistant for composers like you. (etc.)
To me this is a bit like: “please tell us about your relationship with your girl friend. We are about to create a puppy that can take the place as a life partner for other people. We want to know as much as possible about relationships to make the experience for our customers as pleasing and useful as possible.” For those that are happy with puppies maybe interesting. Will it be the end of sex life and relationships as we know it? I doubt it. Real people will stay far more interesting for most of us. No matter how smart and superhuman our puppies will be. And so is music as a language from humans to humans for me.
 
Back in 1984-85, a college classmate of mine was a computer scientist and musician, and we were both in NYC for the summer. He was working for Laurie Spiegel at the time, and was writing a simple program that would compose music algorithmically as part of his dissertation. We had the dystopian idea to get his program running on the college's mainframe, churning out every possible combination of notes and chords, and uploading them constantly to the copyright office (although no such capability existed back then).

Then, once our body of copyrighted works reached critical mass, we would establish "The Ministry of Music" and claim ownership of every possible combination of rhythm, melody, and harmony, requiring the entire human race to pay us for the rights to engage in any musical activity.

Sounds like that might be about to come true....
Sounds like a horror movie! :laugh:
 
To me this is a bit like: “please tell us about your relationship with your girl friend. We are about to create a puppy that can take the place as a life partner for other people. We want to know as much as possible about relationships to make the experience for our customers as pleasing and useful as possible.” For those that are happy with puppies maybe interesting. Will it be the end of sex life and relationships as we know it? I doubt it. Real people will stay far more interesting for most of us. No matter how smart and superhuman our puppies will be. And so is music as a language from humans to humans for me.
Japan’s kind of doing that though...lol
 
These tools will never be more than an assistant to the process of composing, like Rapid Composer, to automate some of the grunt work, like show me the 24 ways to modulate to X key. I have explored algorithmic tools for a while and the results almost always suck. I'd be better off working out the note sequences myself. Especially in the case of melody, these tools sound completely INHUMAN...

After watching some of the videos this is really a testimony to Brad as a HUMAN for taking a VERY rough idea, like a bad improvisation, and polishing it into marketable music, (except for the mixing mastering in many cases).
 
Last edited:
The assumption here is that it becomes "infringing output" by virtue of using copyrighted input. Not all elements of the input are under copyright, and elements can be reassembled in ways that the output isn't recognizable.

While you're right, and I expect there'll be many outputs that won't resemble anything in particular, we were discussing a scenario in which there'd already been an accusation of infringement. It wouldn't be so much by "virtue of using the copyrighted input" - because you can certainly study copyrighted material without infringing on it - but rather a case in which the AI had already misused the copyrighted input.

To expound: I hear AI-composed music that sounds suspiciously similar to my latest hit - maybe it's the melody, the "feel", etc. I sue. The AI's input is examined, and my music was a source. That sounds more difficult to defend than a human composer in a similar situation.

Of course I'm sure it gets muddied when there are multiple sources, or when the AI starts to become what a jury could call creative. Who knows where that line is. I'm really curious how/if we (and the law) will adapt to AI ingenuity. It sounds like Star Trek, but maybe we'll see it in our lifetimes.
 
Last edited:
To me this is a bit like: “please tell us about your relationship with your girl friend. We are about to create a puppy that can take the place as a life partner for other people. We want to know as much as possible about relationships to make the experience for our customers as pleasing and useful as possible.” For those that are happy with puppies maybe interesting. Will it be the end of sex life and relationships as we know it? I doubt it. Real people will stay far more interesting for most of us. No matter how smart and superhuman our puppies will be. And so is music as a language from humans to humans for me.
If they modelled the puppy on a particular girlfriend I had in the past, the result would be a fire breathing cross between a wolf, Tasmanian devil and a velociraptor.
 
oh wow oh wow. This is quite the topic. Ripe for quite strong opinions - showing that amongst us pro/semi-pro/hobbiest composers there is a vast amount of ground between our views. I'm hoping it doesn't descend into a dichotomy... the grey is far more interesting and potentially important. :)

Rather than going too deep into things all in one go... just a few little observations. I might get round to writing more later.

Regarding the copyright. There is no doubt. The piece linked a few pages back is infringing copyright according to the legal frameworks in Australia. I might even send this link to a musicologist in the UK to get his opinion as well from the UK / EU standpoint. Just because I'm interested.
EDIT: Of course it isn't. I stupidly was comparing that AIVA piece to an earlier version of the same AIVA piece - not Willaims Rey. Label your files folks. Let my embarrassment be a lesson to you!

I'm indeed interested enough to send it to a bunch of solicitors to see if any would like to create a test case. Why? Because the legal framework around these things is important. If the company wants to survive, they need to bend to the will of the law. And the law needs to look at what it thinks is important for society / composers / tech companies etc. I'm not sure the law (at least here) has had a chance to look at something like this.
Its somewhat difficult for the company given that although they operate (and are legally setup) in a particular jurisdiction, their product is available for folk to use all around the world. And different regions have very different legal frameworks around this stuff. Even on the fair use of the material that the AI is being trained on.

One could potentially lobby for a legislative framework which assigns composers a right to have their output NOT used to train an AI. Or the opposite could occur. This is not as far fetched as it sounds. There are all sorts of legal opinions / research going on around data usage / massed data usage (second derivative use / big data plays) which people far more intelligent than me are working on right now. There's whole new social sciences being created here in australia (the 3Ai institute at ANU comes to mind as does some very interesting projects from the law dept at Syd Uni) and projects looking at framing data usage and societies attitudes toward it. This may seem a long way from what is happening here, but this software is using data to train the AI. Society is now just starting to come to grips with what data use really means. And legal frameworks are being considered. This is going to have massive, long reaching effects on shaping society. There are huge players involved. Big tech co's are lobbying hard to enable an "anything goes" type situation. Other folk are rebutting that actively.

We will see how this plays out in many of our lifetimes. Some will undoubtably become clearer even in the next 2 or 3 years. GDPR in EU is just a start. Govt regulation is being heavily debated in almost all western regions. Conferences are being held all around the world around AI use, data use etc. Its not all fait accompli. Even if there are many who don't care - who just want to see what happens, there are others who are deeply concerned for what these tech changes mean for society as a whole. There will be the inevitable political differences between the right and the left on this in terms of regulation, but we've already seen both sides coalesce around some of the issues (and essentially place themselves against the position of anarchists and libertarians and the like.)

Edit : One final little thought. The cat is out the bag in regards to AI being useful for many many projects that we might not even have dreamed about 10 or 20 years ago. However, it is far from settled how we (society / law / governments / communities) will interact with it. How much we will allow. Where lines will be drawn. And this is very much happening in our neighbourhoods right now. Get involved if you want. Just look up things like the ODI, or the Open Data Conference, or Data Rights, or ...
 
Last edited:
Regarding the copyright. There is no doubt. The piece linked a few pages back is infringing copyright according to the legal frameworks in Australia. I might even send this link to a musicologist in the UK to get his opinion as well from the UK / EU standpoint. Just because I'm interested.

Great post, but which piece are you referring to?
 
When I think of the time, money and effort we've all spent on our craft and passion, the idea of an app doing a passable job of it, is quite discomforting..
Most 'composers' don't actually put in nearly enough effort! One benefit of AI is that it will raise the bar and should encourage more complex or innovative composition from the humans.
 
I saw one video from Aiva and it does not feel like a very innovative or creative tool in terms of the music. There is probably complex technology working behind the scenes to generate that but the results are not great. And most importantly, it pushes no boundaries.

It seems to me that it has to be commercially driven to satisfy a mass market appeal - an OK sounding, very obvious choice of melodies and arrangement. In it's current form, it is more tedious to work with for musicians who know what they are doing. Hardly anything innovative in terms of the music.

It is faster and better to just put down your own ideas. So, perhaps a 10 second McDonalds advert with lots of voice over may be satisfied by some options from this software, it doesn't seem to do much on a more advanced level.

I am very sure that in the future, such software will absolutely be used by in-house music departments to quickly take care of hundreds of small adverts, YouTube videos and marketing. In fact, I think some are already using it. Good for them. For 30 EUR per month, they can own the copyright forever too :)

This is going to save them a lot of money for sure and that is probably the primary reason it is being built. At least, this software.

Good composers can come up with extremely fast solutions and options on the spot. In fact, media composers who interact with directors and producers routinely have to do this on a daily basis!
 
Oh - and even though it might seem its so far off from what we composers are usually thinking about, it seems like legal minds I've spoken to are VERY interested in what it means to have your music used to train an AI.
Public performance is a very defined thing. Different countries have different legal frameworks around this.
But whether or not it is ok to use the output of spotify / a CD recording etc to train an AI is something that is interesting to these folk. And has ramifications for all sorts of uses of machine learning + AI in the future (and not just music AI).

I'm looking forward to hearing about where some of these conversations head. I'm lucky enough to know a few people who are involved with interrogating these kinds of ideas - and at the very least it will make for some interesting conversations - and who knows where else it may lead.
 
Top Bottom