Is this what will literally completely devalue song-writing, engineering, mixing, mastering?

[BWC]

It would! ...and I think it's just a matter of time. I just don't know how much time we have...

[amanise]

How much time we have for that depends on us finding out what's so special about where we live first, and giving up on the idea of finding an alternative somewhere out there in space. If we fix what we've broken with Earth, and then look after it thereafter, we may have the time. At a guess I'd say we have less than 100 years to turn that around now, and then however long the collapse takes after that if we fail.

Edited to add: I should also say that other opinions are available, of course.

[BigRedX]

The most important bit is still the user input. Without that there is no music output.

[BWC]

Edited to add: I should also say that other opinions are available, of course.

Mine being one of them.

How much time we have for that depends on us finding out what's so special about where we live first, and giving up on the idea of finding an alternative somewhere out there in space.

We've already found out, not so special as we once thought. Special, yes, but not singularly so. ...and the stability we've enjoyed is time-limited (regardless of our actions, or even our presence). We will need "eggs in other baskets" sooner or later.

If we fix what we've broken with Earth, and then look after it thereafter, we may have the time.

...until a threat from somewhere out there in space comes along. What we've broken is far from the only threat to our survival.

At a guess I'd say we have less than 100 years to turn that around now, and then however long the collapse takes after that if we fail.

When we fail. I think the idea of "fixing what we've broken" is misguided. Our nice cushy, life-supporting environment will not last forever, and can't be preserved forever. It's a resource. What we can control is how quickly we use it up. Entropy can't be reversed, but we might could slow it down a good bit. I'm in favor of more efficiency, but against actively trying to "fix" anything. ...especially when we're still not totally sure how the machine works (or is supposed to work).

[Albatross]

...I'm in favor of more efficiency, but against actively trying to "fix" anything. ...especially when we're still not totally sure how the machine works (or is supposed to work).

You know that's something I never heard before and is a very thought provoking point.

It may be that we are playing our part just perfectly in earth's evolution with all our chopping and burning and plastic and chemical dumping and throwing our crap around everywhere.

Perhaps that is just the way god planned it.

[BJG145]

It doesn't really affect me because AIs don't yet trawl eBay looking for vintage MIDI controllers and taking them to open-mic nights. And if I play a gig with them they never buy a round.

I think it's an interesting tool, and I don't really mind it too much, but the one-click-song is boring. It would be far more useful if you could feed it a track and ask it to write a sax solo, then ask for specific variations of that; which might come. Having said that, I'd still prefer an imperfect human one. And random computer-generated songs don't appeal to me much beyond their gimmick value. I'm impressed, but I'm not moved.

(I can see that it's annoying for people releasing music, but that's never been an income stream for me.)

[alexis]

... It would be far more useful if you could feed it a track and ask it to write a sax solo, then ask for specific variations of that; which might come. ...

That's what I'm really looking forward to!

Tunetracks has a technology that from the user point of view is somewhat similar: feed it your song, and it will spit out 10-20 drum tracks from its library it sees as a good match. Results are meh, but for non-drummers that can save hours/ days to get a quick, easy, acceptable-ish sound. Great for the home hobbyist who just wants something to play in the car, "I'll go back and fix that later".

They also have that for bass, keys, "mixing", and maybe others.

Now if AI could do that for a sax solo, or strings, complete with imperfections to make it sound less robotic, that would be great!

I wouldn't be hiring a sax player anyway, most likely, so as long as there is a reimbursement for the scraped players built in to the model, I think ethically it should be fine (happy to hear counterarguments!).

[Drew Stephenson]

If the machine is just pseudorandomly extrapolating, then we're into "room full of monkeys eventually producing the works of Shakespeare" territory.

This is the level I am expecting in the near future and machine learning is already doing this in other fields.

It needs a level of understanding that we still find hard to define, even though we're capable of achieving it.

I wonder about this, but I don't have a good enough music theory knowledge to go any deeper really. I am assuming that all the current 'rules' around music theory are already contained within the 'box' of the learned data, but plenty of great new content is created every day within that same box so, on further thought, is the box / extrapolation even relevant?

[BWC]

It may be that we are playing our part just perfectly in earth's evolution with all our chopping and burning and plastic and chemical dumping and throwing our crap around everywhere.

Perhaps that is just the way god planned it.

The answer to our age‐​old philosophical question?
https://www.youtube.com/watch?v=6vGE73tTSVU

While searching up that link, saw this: https://www.cato.org/blog/revisiting-ge ... t-plastics So maybe it's just that God, or the Earth, or whatever, felt that the mealworms were underfed.

If the machine is just pseudorandomly extrapolating, then we're into "room full of monkeys eventually producing the works of Shakespeare" territory.

This is the level I am expecting in the near future and machine learning is already doing this in other fields.

Yes, that I could see, but still not very interesting (to me, at least). ...sounds like an editing, or curating, nightmare.

It needs a level of understanding that we still find hard to define, even though we're capable of achieving it.

I wonder about this, but I don't have a good enough music theory knowledge to go any deeper really. I am assuming that all the current 'rules' around music theory are already contained within the 'box' of the learned data, but plenty of great new content is created every day within that same box so, on further thought, is the box / extrapolation even relevant?

An understanding of music theory doesn't hurt, but that's not the capability I'm referring to. For example: When working out an arrangement, how do you decide between two (or more) variations of a particular part, when either one works well enough? The machine would choose the one that most closely matches the parameters given in the input, but I'm betting that's not how you do it (or at least not the only way).

[Drew Stephenson]

Hmm, but that 'matching' would be based on a far wider range of listening than I could ever achieve, and likely over a far wider range of style and genre.
I would expect it to be coming up with something more musically original and varied than I do in pretty short timescales, if it isn't already.
Possibly counter-intuitively I think lyrics may take longer to reach best-of-human level because we break the rules so much more easily and still have multiple meanings be understood.
But I could be wrong about that too.

[BWC]

...I'd still prefer an imperfect human one.

Is "imperfect" the right word?

I'm impressed, but I'm not moved.

At best, yup.

Emotional reactions are varied by past emotional experiences, which plays a part in forming opinions and tastes. I haven't seen any good code for that yet.

[BWC]

Hmm, but that 'matching' would be based on a far wider range of listening than I could ever achieve, and likely over a far wider range of style and genre.

Right, that's what makes it personal, "yours".

I would expect it to be coming up with something more musically original and varied than I do in pretty short timescales, if it isn't already.

Hmm, I think I'd rather listen to "yours".

Possibly counter-intuitively I think lyrics may take longer to reach best-of-human level because we break the rules so much more easily and still have multiple meanings be understood.
But I could be wrong about that too.

I think so. I think the same is true of music, but sticking strictly to "rules / known contexts / understood meanings" is less forgivable / interesting enough when it comes to lyrics, perhaps.

[amanise]

Well. This is all very depressing reading for me. I'm sure I must be in a minority on this one, which depresses me further. It looks to me like the answer to "Is this what will literally completely devalue song-writing, engineering, mixing, mastering?" is a big fat yes. How sad.

[BWC]

Well. This is all very depressing reading for me. I'm sure I must be in a minority on this one, which depresses me further. It looks to me like the answer to "Is this what will literally completely devalue song-writing, engineering, mixing, mastering?" is a big fat yes. How sad.

That's not what I'm getting, or getting at. People are always seeking quicker, easier results, but often still value the slower, more difficult work. Like others have said, I do it because it's fun, and I enjoy it. If I can make money from it, great, but if not, that doesn't reduce my desire to do it. The thing is, the folks always looking for the quick & easy haven't (yet) found the fun in doing the work. and they often get bored just as quickly & easily. I'm not sure if this is any real threat to the value, in the long run.

[RichardT]

Well. This is all very depressing reading for me. I'm sure I must be in a minority on this one, which depresses me further. It looks to me like the answer to "Is this what will literally completely devalue song-writing, engineering, mixing, mastering?" is a big fat yes. How sad.

I think there are reasons to be optimistic. AIs could beat everyone on Mastermind, but people still watch humans competing. The human element is what we relate to most strongly, and I think that will continue.

[tea for two]

My response : Video didn't kill the Radio star.

::

...I write music because I like to and there are few things I enjoy more than laying on the floor in the dark listening to whatever it was I created that day.

With a very few exceptions for special cases I don't write music for other people to enjoy, I write it for me to enjoy and part of that enjoyment is the creative process involved. If anyone else likes it that's a bonus but it's not the point.

Just lately I have felt better than I have felt in ages because I've started being creative with music again. As Eddy says, there is almost nothing better than listening back to something you've made for your own enjoyment. I remember the buzz I got in my younger days when I had recorded a song that I liked - the high lasted for days. I had forgotten just what that feels like until I started this current project with a song writing partner - I don't care what anyone else thinks but, to me, just being able to hear the songs we create is priceless.

Often thymes it takes me longer to finish a piece than it should because I'm enjoying it so much, because it's doing something to me. I listen again and again.
After nearly a life thyme of listening to others music, since from lockdown I Iisten almost exclusively to the music that was made thru me.

It's laughable how little theory I have, how poor full errors my playing is if anyone asked me to play.
I legged it ooot of a Guitar shoppee in Brighton last september when the proprietor enquired whether I would like to play an Acoustic I liked.
Yet when making tunes somehow my fingers find the notes, the rhythm, error free. I can’t explain it.
I am literally taken aback : how on earth did that happen thru my crummy self lol.

AI could never replicate this.

[Martin Walker]

I think there are reasons to be optimistic. AIs could beat everyone on Mastermind, but people still watch humans competing. The human element is what we relate to most strongly, and I think that will continue.

I'm hoping the same thing too, and I think this quote from actor Simon Pegg makes perfect sense:

[Drew Stephenson]

I would expect it to be coming up with something more musically original and varied than I do in pretty short timescales, if it isn't already.

Hmm, I think I'd rather listen to "yours".

That's because you haven't listened to my music...

[Drew Stephenson]

I'm hoping the same thing too, and I think this quote from actor Simon Pegg makes perfect sense...

For what it's worth, I'm not sure Mr Pegg will be right for long. ChatGPT was only launched 18 months ago. It is still on a very steep development curve.

But...

Well. This is all very depressing reading for me. I'm sure I must be in a minority on this one, which depresses me further. It looks to me like the answer to "Is this what will literally completely devalue song-writing, engineering, mixing, mastering?" is a big fat yes. How sad.

I would answer this with a question Adrian: how much do you think most people value these things already? I would say that in general most people don't really care about most music. For them there is no devaluation to happen because it's just pretty background noise anyway.
For the people who do currently value it, I think Richard is right and, no matter how good AI gets people will will still want to seek out and listen to human-made music. (BWC's comments above support this despite my joke!)

[BJG145]

AI seems to have come on so fast recently that I wouldn't want to underestimate it. It's tempting to think that it can't be original, or emotionally stimulating, but I think potentially it could be. I agree that the lyrics seem to be the weakest part; it's very difficult to coax any interesting poetry out of ChatGPT in my experience.

A friend of mine recently shared a sketch he did, and the result of feeding it into an AI. As I've said, AI-enhanced is much more interesting to me than AI-generated. It could be empowering when used that way. It could take your MIDI symphony and make it sound like the LSO.

[alexis]

...
how much do you think most people value these things already? I would say that in general most people don't really care about most music. For them there is no devaluation to happen because it's just pretty background noise anyway.
For the people who do currently value it ...no matter how good AI gets people will will still want to seek out and listen to human-made music ...

Gold (both portions quoted above)!

[Sam Spoons]

AI seems to have come on so fast recently that I wouldn't want to underestimate it. It's tempting to think that it can't be original, or emotionally stimulating, but I think potentially it could be. I agree that the lyrics seem to be the weakest part; it's very difficult to coax any interesting poetry out of ChatGPT in my experience.

A friend of mine recently shared a sketch he did, and the result of feeding it into an AI. As I've said, AI-enhanced is much more interesting to me than AI-generated. It could be empowering when used that way. It could take your MIDI symphony and make it sound like the LSO.

IMO the sketch looks much more 'real'...

[BJG145]

...sure, the point wasn't really to compare them, but just to illustrate the potential for music AIs to work with user input rather than creating random ideas of their own...

[RichardT]

AI seems to have come on so fast recently that I wouldn't want to underestimate it. It's tempting to think that it can't be original, or emotionally stimulating, but I think potentially it could be. I agree that the lyrics seem to be the weakest part; it's very difficult to coax any interesting poetry out of ChatGPT in my experience.

A friend of mine recently shared a sketch he did, and the result of feeding it into an AI. As I've said, AI-enhanced is much more interesting to me than AI-generated. It could be empowering when used that way. It could take your MIDI symphony and make it sound like the LSO.

I agree that AI enhancement is interesting.

Transforming a midi symphony is not far off, Ben. Playback from scoring software has come on leaps and bounds, and in a couple of years I think will sound indistinguishable from the real thing. Strings are there already, IMO - at least, I don’t think most people could tell the difference.

I’m hoping the same thing will happen with choir libraries shortly, which are still quite poor and, in spite of claiming to be able to sing words, do it very badly.

[BJG145]

Playback from scoring software has come on leaps and bounds...Strings are there already, IMO - at least, I don’t think most people could tell the difference.

Sounds interesting...I think you might have mentioned something along these lines before. What kind of software are we talking about...? Unless it's a trade secret.