What's new

Aiva - Artificial Intelligence Composition: beta starting today

It can replace me. I’m 65, eminently replaceable. I could give a crap.

As to VI’s, sequencers, etc etc, I totally agree. Tech allowed me to profit and people lost work. They already came for me in other ways, but all I’m saying is I have no desire to “learn to work with (that) tech.” Not appealing to me. I’ll drive Uber til self driving cars replace IT, become a barista before kiosks replace THEM. Are there no workhouses?
 
It can replace me. I’m 65, eminently replaceable. I could give a crap.

As to VI’s, sequencers, etc etc, I totally agree. Tech allowed me to profit and people lost work. They already came for me in other ways, but all I’m saying is I have no desire to “learn to work with (that) tech.” Not appealing to me. I’ll drive Uber til self driving cars replace IT, become a barista before kiosks replace THEM. Are there no workhouses?

Thats fine. I suppose the problem is for everyone else left in the industry. We can all be against the change, but if we intend to stay in the industry we have to adjust...or get out of it completely once we're obsolete. If i had to deal with AI becoming the norm, I'd rather be a part of helping it to grow in ways that make the experience better, rather than just trust it to people who only came into AI for the money.
 
At least Monsanto, now owned by Bayer, haven’t yet invested in music AI and it’s intellectual property . If they did, every piece of music to be written henceforth will be owned by them. If you are caught whistling any melody whatsoever in your backyard, you will be violating their property rights and you will go straight to jail.
 
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.
Thanks, keep us posted!

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 once read "There is no 'AI', only someone else's Data"
 
there is no scenario where we win. For as long as music is a business, automation will win. Honestly anyone using VIs has no place to even talk about how bad it is for computers to take the work of people. Learn to work with the tech, or be replaced by it.
Brutal, but true. I will always write music only I hear and that will be with my own angst, strife, passion and wits, but I am old school. It's a brave new world, where everyone's a composer, or are they?
 

There are two areas of law that may apply. One being Copyright Act 1968 the other being Competition and Consumer Act 2010.

The application of musical influence vis-à-vis the copying of portions or samples was tested in the Federal Court of Australia. It has already been established in law that influence is not the same as copying. Men at Work lost in the case Larrikin Music Publishing Pty Ltd v EMI Songs Australia Pty Limited because the work, "Down Under" contained a copy (regardless of the instrument played or method of production) that was discernible to be infringing upon "Kookaburra Sits on the Old Gum Tree"

However this case was not a case where influence affected the outcome, it was copying a previous musical work and reproducing it in a new one where copyright protection still applied.

Providing a data set of influences to a deep learning algorithm that produces novel and original output would not meet the bar set in Larrikin Music Publishing Pty Ltd v EMI Songs Australia Pty Limited because an influence is not able to be copyrighted. You need a 'work' not an 'idea' copied to prove copyright infringement.

In the case of Consumer Law, unless you are representing that the work is original when it is not, or unless you represent that there are no royalties payable because they are, then this aspect of Australian law would also not affect the use of deep learning to create a novel composition.

The Men at Work case was decided based upon, in part, where a substantial portion of a work was appropriated. Kookaburra is 4 bars, Men at Work used 2 bars, it was substantial. It was also discernible to a non trained listener to be a copy of Kookaburra.

The work in the example provided by AIVA does not meet that threshold.

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.

I am a 50% partner in an Australian business which deals with licensing, royalties and publishing. We have very good arts media lawyers. Although many things about deep learning and deep learning output may not have been tested at law yet, the factor of novelty is weighty when it comes to copyright law in publishing. It is not sufficient to show that "this song reminds me of" to prove copyright infringement.
 
Last edited:
There are two areas of law that may apply. One being Copyright Act 1968 the other being Competition and Consumer Act 2010.

The application of musical influence vis-à-vis the copying of portions or samples was tested in the Federal Court of Australia. It has already been established in law that influence is not the same as copying. Men at Work lost in the case Larrikin Music Publishing Pty Ltd v EMI Songs Australia Pty Limited because the work, "Down Under" contained a copy (regardless of the instrument played or method of production) that was discernible to be infringing upon "Kookaburra Sits on the Old Gum Tree"

However this case was not a case where influence affected the outcome, it was copying a previous musical work and reproducing it in a new one where copyright protection still applied.

Providing a data set of influences to a deep learning algorithm that produces novel and original output would not meet the bar set in Larrikin Music Publishing Pty Ltd v EMI Songs Australia Pty Limited because an influence is not able to be copyrighted. You need a 'work' not an 'idea' copied to prove copyright infringement.

In the case of Consumer Law, unless you are representing that the work is original when it is not, or unless you represent that there are no royalties payable because they are, then this aspect of Australian law would also not affect the use of deep learning to create a novel composition.

The Men at Work case was decided based upon, in part, where a substantial portion of a work was appropriated. Kookaburra is 4 bars, Men at Work used 2 bars, it was substantial. It was also discernible to a non trained listener to be a copy of Kookaburra.

The work in the example provided by AIVA does not meet that threshold.



I am a 50% partner in an Australian business which deals with licensing, royalties and publishing. We have very good arts media lawyers. Although many things about deep learning and deep learning output may not have been tested at law yet, the factor of novelty is weighty when it comes to copyright law in publishing. It is not sufficient to show that "this song reminds me of" to prove copyright infringement.

Growing up in Australia I remember listening to radio broadcasts in infants school of “Kookaburra Sits on the Old Gum Tree".

All school kids used to sing along and the melody and lyrics were as common as apple pie or vegemite sandwiches. It was the musical equivalent of using the term “escalator”, originally a trademark but now in common usage.

Men at Work quoted this melody most probably as a result of a nostalgic memory. As a musician I saw it’s use nothing more than a quote similar to how a jazz musician would quote during an improvised solo.

I think you could have quoted from any Australian nursery rhyme and the song and it’s intent would not have been diminished.

Unfortunately, when you ask a non musical jury whether something is similar or not, they will naturally say that it is similar. If the jury were made up of a broad range of musicians from academics through to practitioners, the result would have been different.

Personally I think the outcome was deplorable and shame on Larrikin for pursuing it.
 
Unfortunately, when you ask a non musical jury whether something is similar or not, they will naturally say that it is similar. If the jury were made up of a broad range of musicians from academics through to practitioners, the result would have been different.

Personally I think the outcome was deplorable and shame on Larrikin for pursuing it.

I agree it was a fairly egregious use of the law to extract a monetary outcome from Men at Work, however the issues you describe of a trained vs untrained listener were tested in the appeal and the court found that copyright infringement occurred in any case.

It wasn't the Australian music industry's best day, however the case has set precedent and also clarified the difference between influence and copying which is probably very applicable to the discussion of the use of AIVA.

AIVA produces novel, unique output. It might remind you of things but it's compositions don't meet the bar for infringement, not in Australia at least.
 
I agree it was a fairly egregious use of the law to extract a monetary outcome from Men at Work, however the issues you describe of a trained vs untrained listener were tested in the appeal and the court found that copyright infringement occurred in any case.

It wasn't the Australian music industry's best day, however the case has set precedent and also clarified the difference between influence and copying which is probably very applicable to the discussion of the use of AIVA.

AIVA produces novel, unique output. It might remind you of things but it's compositions don't meet the bar for infringement, not in Australia at least.
Out of curiosity, if Charlie Parker was alive today (and Australian) and quoted 2 bars of a copyrighted melody during a 64 bar solo on a standard, do you think the same judgement would be reached?
 
Out of curiosity, if Charlie Parker was alive today (and Australian) and quoted 2 bars of a copyrighted melody during a 64 bar solo on a standard, do you think the same judgement would be reached?

No, because the infringement needs to meet the bar of being substantial. There's some good blog articles written about the time of the judgement by the Federal Court, they're worth a read to understand how "substantial" was defined in the case.
 
Tech bros with no soul. Totally offensive.

What kind of garbage person even thinks this up.

Someone once said something similar about Robert Moog.

The advent of commercially available synthesisers led to all sorts of similar reactions.

I really find it strange that on a "virtual instrument" forum people are so against the application of deep learning in music. It's exciting to witness the dawn of the most innovative music technology since the synthesiser.
 
Someone once said something similar about Robert Moog.

The advent of commercially available synthesisers led to all sorts of similar reactions.

Bad analogy!

I really find it strange that on a "virtual instrument" forum people are so against the application of deep learning in music. It's exciting to witness the dawn of the most innovative music technology since the synthesiser.

This is to innovation what the atom bomb is to the hammer and chisel.

Eliminating human beings from the soul of humanity is anything but music technology.
 
Bad analogy!

This is to innovation what the atom bomb is to the hammer and chisel.

Eliminating human beings from the soul of humanity is anything but music technology.

Why do human beings have to be eliminated? Technology is here to serve humanity, not replace it. Only morons want to replace people with technology.

Humans are smarter than AI because they understand what intellect is and know how to apply it. So humans need to use AI to do menial grunt work like creating metadata.

Humans can do all of the stuff that humans like. AI can be used to do all of the stuff that people do not want to do.
 
Humans are smarter than AI because they understand what intellect is and know how to apply it. So humans need to use AI to do menial grunt work like creating metadata.
If you read "thinking, fast and slow" by Daniel Kahneman, you too can learn in how many ways these statements are incorrect.
 
I think it's easy to dismiss the deep learning algorithms, but they are frighteningly good, especially with deep networks. As long as there is some "order to the madness", modern ML algorithms will find and make sense of it. I submit that most music follow some sort of abstract set of rules, and as such that can be found and duplicated. Whether the music it produces is interesting and/or good, is a different issue, but that same argument can be had with different human composers/songwriters.
 
Top Bottom