What's new

Aiva - Artificial Intelligence Composition: beta starting today

By the way, my story was called The Robot Evasion - because it's just that, an evasion of the real problems we're facing today.

Listen, you people who are justifying this disgusting excuse for a program are working way too hard.

Using algorithms to come up with interesting lines is one thing; using algorithms to generate crap music is offensive.
 
Maybe you can't replace the actual feeling, but can anybody but the creator tell the difference? Would anybody other than the creator care? Even if others did, would directors and producers care? They already don't!

Yes, everyone including the creator can sense the difference.

You may have noticed that narrowcasting (i.e. streaming, as opposed to broadcasting) is here and we're in the golden age of television. I'm not naive enough to believe that shite TV is going away, but only shite directors and producers don't care about quality.

That aside... we'll all be dead and gone in a few decades. May as well satisfy ourselves while we're here.

By the way, fuck that program.
 
I can't wait for the first all-AI bands to hit stadiums and theaters.

The skynettes
The rap-licants
Andy Roid and Max Headroom

Something to look forward to. :laugh:
 
I would love to get a poll of how many people, mainly professionals actually spend their time doing original work. How many commercial jobs actually let you just go at it with no reference and no "please give us a song that sounds like this person's song"? I still see AI taking that work. The jobs that require mimicking something else.

Will AI someday create new unique masterpieces with no references and just its knowledge of theory and what programmers have told it about human emotion? Maybe...but if it does, it's not like it's taken someone's real job. Its creations will still have to pass the test of consumption. Even if it makes great music, people will always have preferences.

Maybe I'm wildly optimistic about tech, and way too pessimistic about humans. Still, in my experience as a professional, I often find people making uninspired redundant crap on the commercial level. The best music I hear these days is still the stuff by people who don't care if it sells, and who just want to make what's in their hearts. For them, AI can't replace them any more than a pop powerhouse like Ed Sheeran or Taylor Swift could.

Be unique, and you shouldn't have to worry about this on a creative level. As professionals pushing out regurgitated crap for audiences who don't care about quality, or for companies who selected their musical requests based on marketing data? Yeah, we might lose that job...
 
Do Androids Dream of electronic A.I. Music?
Of course!
And when future film composers knock it out of the park with their A.I. generated cues, they'll get this from the director.

addtext_com_MjMzODM1MjY1Mw.png
 
How many commercial jobs actually let you just go at it with no reference and no "please give us a song that sounds like this person's song"? I still see AI taking that work. The jobs that require mimicking something else.

Ah, but good musicians write *good* soundalikes.

The obvious example is John Williams out-Wagnering Wagner (and others).

No I won't post it :) but even I've written at least one blatantly paraphrased cue that I like better than the original!
 
In its current form, it's shaping up to be a useful tool in the next few years for people with large (commercial, mostly) work volumes or library composers. Eventually, the features and interface might make it possible for directors and marketing people to get hands-on with it, at least those that are somewhat music-savvy. I would not be surprised if the vast majority of commercial, trailer, and library music is 90% A.I. composed within a decade. No, they won't be hiring AIVA for star wars episode 16 most likely, but just as samples made it cheaper for low budget films to afford a good score, AI composition will make it even more so, probably at the expense of some jobs. I'd rather adapt to use the tech as a tool and maybe eventually an assistant than try to explain to a director why the b-flat I wrote in measure 4 has more soul in it than the one the robot wrote.

I mean, imagine churning out a career's worth of library music in a year or less. Even if only a few of those tracks are successful, that's enough to justify using the software. For simple, upbeat background music or generic epic action tracks...I have no doubt this will be a viable option. Those who adopt and take advantage of this stuff earlier will reap more benefits.
 
Ah, but good musicians write *good* soundalikes.

The obvious example is John Williams out-Wagnering Wagner (and others).

No I won't post it :) but even I've written at least one blatantly paraphrased cue that I like better than the original!

That’s fine, but that assessment is subjective and still sidesteps the fact that writing a “soundalike” is not the peak of originality or artistic expression. It’s something AI would eventually be able to do as well, and people might end up feeling the way you do...that the soundalike is better than the original. our ability to copy and expand on previously established work is absolutely something AI can and will do. So again, I think completely original works based solely on the desire to express one’s emotions are a place where humans can take be replaced. I don’t know if that’ll pay the bills, but at least humans will be doing true art rather than the business of art.
 
There's something to be said for stopping and trying to understand each others points of view, rather than simply giving our own, hoping that will change someones mind.
I personally feel this strongly in a conversation such as this, where there is tonnes of misinformation, where understanding of the underlying tech is generally thin, where conceptual ideas over artform and purpose are being trivialized, minimized, misunderstood (perhaps rightly?!)

This topic is like the tip of the iceberg of what might be facing society in the next 20 years. Its interesting that it occurs in an area many thought would come much later (creativity!)
Now, substitute the word creativity for work - and conversations change a lot. It just so happens that a lot of folk here WORK in the world of creativity. (What that even means from a philosophical point of view fills books. What that means for markets, power, subjugation, future work, well being, and ultimately humanity in general will need to be grappled with. But should the conversations change a lot? Especially in an age when "work" is so highly prized yet not necessarily understood.

One thought bubble I'd like to throw out.

Why do we see anything as inevitable? Is it acceptance of market forces? Is it fear? There's arguments to be made that technology advances are not unable (deliberate double negative) to be better thought about and administered (regulated) by custodians. New thought frameworks and therefore societal attitudes happen over time all the time. Technology is moving quickly, and perhaps quicker than we see frameworks / societal attitudes move in general. But there are tonnes of people who are not throwing their arms in the air saying "it is what it is" or "what can I possibly do" and actively engaging with ML, AI, Big data (all so related) and looking at how we as citizens can shape it rather than it being prescribed "because its technology and we can't control it"

Companies will not LIKE it when it interferes with their golden egg. But humans and citizens and govts understand the need for humans to play a role in their world that makes for effective communities - works for making humanity as a whole better. Some companies will work with these ideas and others will not. Humanity is more important than business. More important than data. Human connection.

(Insert your own paragraph on power here - especially in regards to government and mega business / monopolies. Or I'll dig out some books...)

So what?

Good question. I have my own personal feelings on where things will head / what it means, but right now I'm a little more interested in hearing as many points of view as possible. I'm well aware that I'm only here for a short part of where these things will eventually lead humanity. I'm not sure I can grasp all that, but hope for my daughter to be able to participate in shaping it for the good of the world, whatever that means.
 
That’s fine, but that assessment is subjective and still sidesteps the fact that writing a “soundalike” is not the peak of originality or artistic expression. It’s something AI would eventually be able to do as well, and people might end up feeling the way you do...that the soundalike is better than the original. our ability to copy and expand on previously established work is absolutely something AI can and will do. So again, I think completely original works based solely on the desire to express one’s emotions are a place where humans can take be replaced. I don’t know if that’ll pay the bills, but at least humans will be doing true art rather than the business of art.

I disagree with pretty much everything there, but other than that I agree. :)

And no, I'm not sidestepping anything. AI sucks whale dingus.

(EDIT: I should say it sucks whale dingus when it's being abused. There are certainly good uses for it!)
 
Last edited:
So here's a thing I briefly touched on a few pages back - but worth perhaps expressing a bit stronger.

Now I'm using info for Australia here, but I'm sure due to relationships thru other FTA's etc, there's something similar in other countries / regions.
Australian law gives copyright owners (be it the copyright in sound recordings or in the music composition) exclusive rights regarding the use of their music.

A business needs to obtain permission from copyright holders of BOTH the recordings and the work in order to use them for any purpose.

(emphasis mine)

Training an AI is a purpose. It is a business activity which has (likely) NOT been granted a license for this use. Feels like an interesting rabbit hole to go down to me...

Sure - we may let this happen thru publishing agreements and the like (how many of us have publishing agreements with wording that essentially lets our publishers negotiate these kinds of things without our knowledge / our consultation) but I'm pretty sure these negotiations haven't happened yet. Given these same publishing houses are trying to own the publishing rights to music generated by AI (I know at least three big ones that are right now) - yet they are meant to be looking after the interests of the composers who write for them as well... I'll just leave this one here hey? :)

Also - lets actually talk about the use of our compositions for other businesses gains.
 
Why do we see anything as inevitable? Is it acceptance of market forces? Is it fear? There's arguments to be made that technology advances are not unable (deliberate double negative) to be better thought about and administered (regulated) by custodians. New thought frameworks and therefore societal attitudes happen over time all the time. Technology is moving quickly, and perhaps quicker than we see frameworks / societal attitudes move in general. But there are tonnes of people who are not throwing their arms in the air saying "it is what it is" or "what can I possibly do" and actively engaging with ML, AI, Big data (all so related) and looking at how we as citizens can shape it rather than it being prescribed "because its technology and we can't control it"

I dont mean to imply that inevitable means that we have no impact, say, or influence. I just mean that when a technology comes in that stands to improve quality of life or drive a new financial incentive, society gradually tends to move toward it until future generations accept it as just the norm. There are few cases of society regressing from one superior technology to an inferior one. If anything most times the replacement comes in the form of something different of equal or greater value.

AI isn’t just about music. As its spread out throughout society for a number of incredibly useful purposes, it will eventually find its place in music as well. What if, for example, the portable device you had could determine your mood and a musical AI could create the soundtrack to your life. Every day, a new musical composition for you and you alone. A musical snapshot that instantly would spark memories of moments in your life. People wouldn’t just engage music on the consumer level, they would feel it on a purely emotional level. This isn’t a job that exists, so no one loses from this application. Will some major big money players try and us AI to outsource the creative work we do? probably, and some big companies will likely abuse AI forcing reforms and limitations on its use. As has been presented here, the legal guidelines dont even fully exist to determine how far AI will go.

Still for the lifetime of most of us, there is potential for AI to help us speed through boring tasks. Maybe its creating melodic phrases, or maybe its just interpreting the piece you’ve written and applied real-time performance, or even suggested new alternatives to work you create. Maybe AI notates your midi written songs accurately, and automates various processes leading up to live performances and recording. There are so many places where we can step in and help make AI work for us, rather than against us.
 
AI notates your midi written songs accurately

That's been working pretty well for a good 30 years. You still have to clean it up a little, but every sequencer can put dots on a stave and get close to the rhythms.

I'm not sure whether that qualifies as AI, but it's quite different from writing the music for you.
 
That's been working pretty well for a good 30 years. You still have to clean it up a little, but every sequencer can put dots on a stave and get close to the rhythms.

I'm not sure whether that qualifies as AI, but it's quite different from writing the music for you.

Writing music for you is one application, that might lead to more practical and unique applications. Instant rearrangements for different ensemble sizes. Maybe automated arrangement ideas around a motif and chord structure you’ve set. (Orb composer does that to a degree). Maybe some software will analyze your work and tell you if it’s too similar to already established works. Maybe it takes a body of your own work and analyzes your writing style so it can auto generate quick cues that you can fine tune. There’s many ways it will likely play out because many people have skin in the game. More reason to be involved in the development so it does the things you want it to do. If you don’t want AIVA just writing random crap for you, then how about participating and letting them know the things you would want an AI assistant to do within your workflow that is valuable to you, but doesn’t overstep the boundaries.
 
So here's a thing I briefly touched on a few pages back - but worth perhaps expressing a bit stronger.

Now I'm using info for Australia here, but I'm sure due to relationships thru other FTA's etc, there's something similar in other countries / regions.
Australian law gives copyright owners (be it the copyright in sound recordings or in the music composition) exclusive rights regarding the use of their music.

A business needs to obtain permission from copyright holders of BOTH the recordings and the work in order to use them for any purpose.

As I understand it Aiva was trained on the work of composers long since dead and produced long before the concept of copyright existed, therefore if you take the original work of Bach, Mozart and so on then copyright does not apply on the underlying musical work and therefore there is nobody to make a rights claim.

In his TED talk, Pierre Barreau says that Aiva was trained on over 30,000 scores of history's greatest composers. Providing copyright has expired and these scores are in the public domain then that's the final chapter and verse of the copyright issue.
 
As I understand it Aiva was trained on the work of composers long since dead and produced long before the concept of copyright existed, therefore if you take the original work of Bach, Mozart and so on then copyright does not apply on the underlying musical work and therefore there is nobody to make a rights claim.

In his TED talk, Pierre Barreau says that Aiva was trained on over 30,000 scores of history's greatest composers. Providing copyright has expired and these scores are in the public domain then that's the final chapter and verse of the copyright issue.

Unfortunately they're not just focused on scores/works that are in the public domain. The earlier piece highlighted used Williams to train the AI... and in usage they allow anyone to upload a midi file. (Perhaps that is putting the responsibility onto the user - but that's not cut and dry either). If they were purely doing pieces inspired by out of copyright music - sure. I get your point.

I also have looked at 3 other such engines in the works (two not public as far as I know) and in one case they're not training it from MIDI but from audio files - of common works... which for me throws up red flags. Even if this is just internal research, it still seems to be breaking copyright law to me.

There's also copyright in whoever made the midi that they've used...if they indeed have had permission to make the midi in the first place. (in the US at least, midi has at times been classified as a digital recording for the purposes of the court.) Now of course if the original piece is in the public domain, great! But I don't think these companies are going to be satisfied with that.
 
What if, for example, the portable device you had could determine your mood and a musical AI could create the soundtrack to your life. Every day, a new musical composition for you and you alone. A musical snapshot that instantly would spark memories of moments in your life. People wouldn’t just engage music on the consumer level, they would feel it on a purely emotional level.
That would probably be used to drive purchasing behaviour by more targeted "pulling on emotional strings", when people are at certain stores or looking at certain websites.

If you don’t want AIVA just writing random crap for you, then how about participating and letting them know the things you would want an AI assistant to do within your workflow that is valuable to you, but doesn’t overstep the boundaries.
I've heard compelling arguments that all the hightech workflows drive the old-school analog guys out of the creative roles in big productions and that is detrimental to the bottom line quality of the product. The argument was made for 3D graphics tools that were introduced in the last 2 decades (zBrush etc.), but in the context of discussing deep learning AI as tools for artists.
 
Unfortunately they're not just focused on scores/works that are in the public domain. The earlier piece highlighted used Williams to train the AI... and in usage they allow anyone to upload a midi file. (Perhaps that is putting the responsibility onto the user - but that's not cut and dry either). If they were purely doing pieces inspired by out of copyright music - sure. I get your point.

Uploading an influence is different from training.

Uploading an influence just asks the already trained AI model to be biased toward the influence uploaded, influence (at least in Australia) is not able to be copyrighted, there's already legal precedent.

It's the input training data that is important here, if that's public domain then the copyright question is closed.
 
That would probably be used to drive purchasing behaviour by more targeted "pulling on emotional strings", when people are at certain stores or looking at certain websites.


I've heard compelling arguments that all the hightech workflows drive the old-school analog guys out of the creative roles in big productions and that is detrimental to the bottom line quality of the product. The argument was made for 3D graphics tools that were introduced in the last 2 decades (zBrush etc.), but in the context of discussing deep learning AI as tools for artists.

I'm curious to know what it is about "high tech" that makes it so prone to claims of diminished quality? where is that cut off in the history of tech? To classic musicians and composers that had to suffer through the invention of radio and the record player, was the quality of that tech superior to going out and seeing a real ensemble play live? Did music become better or worse with the application of EQ and compression? Did Pro Tools kill music?

It's a bit hyperbolic, but every generation seems to fight change, and the next generation defends it until their norm becomes outdated, so I'm trying to understand when music officially was killed by technology. It's a very old regurgitated line. What academic standard sets the quality of music? While we sit here condemning the concept of AI music, we might find that in another 50 or so years, academia might be studying it in depth and teaching about how new complex musical concepts were born from it. Wouldn't be the first time.
 
Top Bottom