What's new

Artificial Intelligence as an assistant composer?

Thanks for this post. Definitely an interesting discussion.

Some random thoughts:

The music demos sound, to put it bluntly, like a composer who is just starting out. Which I guess is what the AI is anyway.

Slight more perplexing for me: Is it really worth all the investment and man hours to tackle a market (royalty free music) which is already saturated, where there is no shortage of composers, and where the prices are already driven down? It kind of seems like a solution to a problem that doesn't need solving?
I assume the attraction would be to "edit" and "direct" the AI composition. Maybe that's the key.

Strikes me that if this sort of AI takes off, then as composers we'll have to keep one step ahead and write better, complex music.
 
Last edited:
I own Synfire Pro. I really love how it sparks my creativity.

However, Hexachords Orb Composer's stated functionality goes way beyond Synfire Pro. Take a look and let me know what you think.

Well, talk is cheap ... I suppose.
I think we can't really compare until we actually have Orb Composer. Do you have a copy?
Synfire is here now, and a few years back it had a lot of features that were unusable, too ... and it cannot replace anything in the composers workflow. It just makes it easier to get from A to B ... like changing a chord or the whole progression in certain places ... etc. It's just a tool ... and there is very little AI there. (even though there is some in the analysis of MIDI)
 
Is it really worth all the investment and man hours to tackle a market (royalty free music) which is already saturated, where there is no shortage of composers, and where the prices are already driven down? It kind of seems like a solution to a problem that doesn't need solving?
Yeah, but it's kinda fun and interesting (to some people), and if it improves to the point that it becomes commercially viable, you'll be years ahead of everybody else.

I mean, why did I make a female virtual death metal vocalist?
Trust me, if you read around their web sites, the subtext is clear. They want to eventually replace composers.

They simply don't want to tick off composers right out of the gate. Someone has to be around to push the button ;)
Well, I suppose "we want to replace all humans, and we're starting with composers because they're the easiest" might rub some people the wrong way...
 
Trust me, if you read around their web sites, the subtext is clear. They want to eventually replace composers.

They simply don't want to tick off composers right out of the gate. Someone has to be around to push the button ;)
Well, composers can also become "filmmakers" with all the AI gadgets. The visual technologies are way more developed (imo) and there is more interest in them. The mathematical side of music (at least 12et, but you can find interesting articles on non-Western music at JSTOR etc) is explored a long time ago.
It's a matter of time DAWs to start impementing auto-composing and harmonizing (probably based on styles), melody writing (vectors?), orchestration (ircam already has something in this vein) tools.
 
I don't think there's any need to worry about AI replacing composers. Because by the time AI could actually truly write monumental symphonies or real art music (can it be called real art if it was done by an AI?) AI would have already replaced every other jobs of other people in every professions. And agreed with @Polkasound artificial is still artificial. No matter how good/cool a machine could write a piece of music, I wouldn't call that piece "art". Art works are humanly creations
I would not be so confident there, AI is much farther ahead than people think if you have been reading into it. These things are kept under wraps, until they are ready to be unveiled...and they have been working on this stuff for I would say at least 10 years now. Why do you think all of a sudden there is a surge of robots and intelligence in the market.

All electronic devices are connected, so you are looking at AI already having machine learning for something like 4 billions peoples, likes dislikes, preferences, facial expressions, locations, languages etc...people forget data is mined by the second.
 
It's certainly an interesting idea. While I don't think we'll be seeing computers suddenly start writing symphonic masterpieces, I can definitely imagine a system where you say I need 30 seconds of music that evokes emotion X, with edit points at 15" and 21". That could then generate you 5 pieces to choose from, pick one and it generates 5 variations on that version, etc, etc. In fact I could probably program that right now given a copy of Omnisphere and a scriptable DAW!
It's not going to put Hollywood film composers out of a job, but it might become a viable option for TV on a tight schedule, or lower budget games.
 
I heard of a book that was written about the future of automation and its effect on societies. I can't remember the author or the book title but he basically said there will be a generation of "useless people" coming soon. People that would have had lower skilled jobs that will be replaced by automation. Perhaps these people will be given a yearly wage by the government to just sustain themselves. I think this type of life would be maddening for most people. Enough money to "survive" but note enough money or incentive to "thrive". You don't even have to get out of bed to get down to the fast food restaurant to flip burgers. It's a frightening outlook.

As far as AI and composing go no robot is going to f*ck up an Am chord like I can so bring it on Terminator. I'm the king of "artificial intelligence" ;)
 
This is a really interesting discussion.
There are many jobs already affected by mechanisation and as AI develops further it will affect more and more people. I'm just as interested in finding out what everyone will do when there are limited opportunities for employment available. That could be a great opportunity.

edit - You beat me to it @dpasdernick . This is precisely what I was getting at. How will we engage our minds in a future with limited employment? I'm sure I heard a radio interview with an author of a book just as you describe.

I'm looking for a copy of Warnings by Richard A Clarke & R.P. Eddy right now. I'm not sure if this sort of discussion would fit into that book but, it might. Especially if someone is looking at potential flow-on effects, and I'm sure someone is.
 
Last edited:
I don't think the AI tech is as significant as the cultural shift toward "music" as a commodity, its use in the service of selling other commodities, and the ever more sophisticated exploitation of scientific understanding of how tone and rhythm affects the brain so that the music becomes nothing more than a tool to be deployed as a means to achieve another end.

A car commercial track on TV sounds, well, like a car commercial, because the "artist" doesn't have anything to say other than "buy this car". Whether this is generated by a human being or AI is of no concern to me.

The reason those fast food burger joints remain popular despite being unhealthy is that we now have so much empirically derived evidence about the chemistry behind what people crave: Sugar, fat, salt & starch. Bundle these together in a competently balanced package and millions of people will have "Big Mac attacks" or whatever. It's the same with standardized pop rock I–vi–IV–V et al progressions that are fashioned formulaically for making a quick buck, exploiting the known and predictable reactions people tend to have when their ears are exposed to these patterns. No reason an algorithm can't do that as well or better than a human being.

Music as product can and likely will be supplanted by AI in the Information Age, just as factory workers were replaced by machines in the late Industrial Age.

But music as art? Never.
 
That will be replicated too.

There is already an AI which can paint pictures that absolutely nails the style of revered artists from the renaissance.

One day, when the technology advances, AI would be able to reflect on all its experience and create something "unique" in the same way that all of your influences lead to you creating something different.

Seems many people underestimate human creativity, ideas, and personal style. It is not simply the sum of experiences or a process that can be duplicated. Some days I will pick one horn for a solo, another day a woodwind, another day a synth. It's not the song or purely my past experiences that changes my mind, it's how I'm feeling at that moment and how I want to express that feeling which is effected by a thousand things, includig the random dreams I had last night (which are themselves dependent on thousands of things). So people here are thinking there will be a program that will factor in all those things then add a "style" that will work well? And one that other humans will find compelling? I think that is so difficult for a computer to do that it is far in the future at best. Yes, they can write music, even music that can be sold. But don't assume it's easy to make music compelling, that it would touch us all on an emotional level. A great composer can do this, I have yet to hear a computer generated piece that can. So are you looking to make background tracks that just make money? Or do you want to touch people emotionally? Because there are easier ways to make money if that's all someone wants.

Many brilliant musical humans have studied great composers of the past and every score they've done, yet have not been able to duplicate that great composers success. So you're saying some computer programming expert with less musical knowledge is going to program a computer that will "break that code"? I'm suggesting there is no code. Like with great stories, once you put music into purely a formula, it doesn't work anymore. Witness the long line of bad Hollywood films.
 
Last edited:
Well, talk is cheap ... I suppose.
I think we can't really compare until we actually have Orb Composer. Do you have a copy?
Synfire is here now, and a few years back it had a lot of features that were unusable, too ... and it cannot replace anything in the composers workflow. It just makes it easier to get from A to B ... like changing a chord or the whole progression in certain places ... etc. It's just a tool ... and there is very little AI there. (even though there is some in the analysis of MIDI)

LOL. Agreed and I don't have copy yet. But I am signed up for the Beta out this fall. I will let everyone know once I get my hands on it.

And I do love Synfire as an musical idea development tool.
 
I was actually surprised to hear Hexachord's (broken link removed). If it's 100% AI, then it created some nice melodies here and there. AI music usually sounds like diarrhea to my ears.

I see this appealing to a lot of cheap indie game developers. And by cheap I mean those 14yo buggers who watched a few YT tutorial on Unity and want to pay you in Pokemon cards.
 
Like with great stories, once you put music into purely a formula, it doesn't work anymore. Witness the long line of bad Hollywood films.

Witness the long line of bad Hollywood films that sell like hotcakes.

It's not a matter of making music that's good. It's a matter of making music that is good enough, which is what I'm afraid of.
 
I would be worried about AI personally.

It may not not replace all the big stuff like Hollywood feature films or GoT style TV shows but all the little things like TV, adverts, Trailers, Corporate videos etc etc. Library music stufff essentially.

That could all be replaced IMO with a sophisticated enough algorithm.

After all, computers follow rules and music follows rules.
 
I was actually surprised to hear Hexachord's (broken link removed). If it's 100% AI, then it created some nice melodies here and there. AI music usually sounds like diarrhea to my ears.

I see this appealing to a lot of cheap indie game developers. And by cheap I mean those 14yo buggers who watched a few YT tutorial on Unity and want to pay you in Pokemon cards.

What's really interesting is that they allow the "composer" to import MIDI, modify the structure of the song etc as a starting point and it integrates with DAWS and hence VSTs allowing modern composers to use the dozens of sample libraries they own. I know I don't want to throw out my investments.
 
But a computer isn´t able to understand an "emotional reaction" like humans do imo. Sure he can follow rules what humans programmed him to do or to react somehow, but that doesn´t really mean he understands the drama and feel it, his algorithm programmed by humans tell him what to do. But here comes the problem: While an algorithm can cover several situations maybe, when you just change one variable the computer ends up in an error. It is just impossible to inject this into him, because enjoyment, fear, love etc is something irrational and this exactly is system immanent to the nature of a computer. And yes I did listen to some music generated by AI. And while some of the examples are not bad at all, to put it mildly, I would say: Just change one parameter and the machine doesn´t know how to deal with the different emotional situation whilst a human beeing has the ability to generate an emotional real response to adapt to the new situation. You know guys: That is what humans are all about..our brain is so much more complex than any supercomputer in the world, we are no machines, and we don´t think in zero and ones, our imperfections and the ability of having a real emotional response sets us apart from a machine. And that is the whole point for me.
 
Last edited:
  • Like
Reactions: Fer
Maybe not completely independently. But add enough humanly editable variables to the program and I'm sure you could have an "add cinematic music" function to Final Cut or Premier. So the video editor could just add music him/herself without having to go through libraries or composers.
 
Top Bottom