A.I. "Jazz"

[Docbop]

51 minutes ago, ElmerJFudd said:

Ya graphic art, animation and writing, music as well (especially for the non discerning consumer).  Definitely a tool for creative people.  Replacement for human creators?  Perhaps around the corner… maybe closer than we think.  

I don't see it so much as a replacement but another tool for the human creator.   That DScript when I saw it demo by a creator he was still in control of the script being feed to it and making changes on what it did.    Now the video generated that could replace a lot of making B-Roll for video saving time to focus on the main video content.    So just another tool.  

[AROIOS]

jazzpiano88 and ReezeKeys both know what's up.
 

Back in 2015, the copywriters and graphic designers around me had very similar responses to the self-assured brothers above, when I told them it's just a matter of time Deep Learning is gonna replace them professionally.
 

"A.I. will 'NEVER' challenge our creativity." is what they said.
 

It only took 8 years for Large Language Models and Generative AI to seriously threaten their revenue streams. 😃

[niacin]

as soon as AI works out that becoming a jazz musician is hardly a sure road to financial success I’m sure it’ll move onto more lucrative tasks

[niacin]

On a more serious note I just can't see the point.  Jazz is at its heart a form of live improvisation.

[jazzpiano88]

I could easily see AI doing some jazz tasks successfully.   Like “reharmonize Somewhere Over the Rainbow”.  It generates the notes.  

[Josh Paxton]

In the early '90s, when I was first studying both jazz and computers, I wondered if any given musician's improvisational style could be boiled down to what note they were likely to play next given the notes they'd just played, the harmonic structure of the tune, tempo, length of the improvisation so far, etc. I guess we're all about to find out. Honestly I don't think it will be very long at all until we can tell a chatbot "Create a 4.6-minute video of the Oscar Peterson trio with Ray Brown and Ed Thigpen in 1964, playing a swing version of the Prince song 'Kiss' at 200bpm. The environment is a crowded jazz club in Copenhagen. Black and white film footage shot from four cameras", and the result will be effectively indistinguishable to anyone apart from serious, hardcore Peterson scholars, and serious, hardcore film scholars.

[ProfD]

3 hours ago, Josh Paxton said:

Honestly I don't think it will be very long at all until we can tell a chatbot "Create a 4.6-minute video of the Oscar Peterson trio with Ray Brown and Ed Thigpen in 1964, playing a swing version of the Prince song 'Kiss' at 200bpm. The environment is a crowded jazz club in Copenhagen. Black and white film footage shot from four cameras", and the result will be effectively indistinguishable to anyone apart from serious, hardcore Peterson scholars, and serious, hardcore film scholars.

Given the parameters of the inquiry, I think we can rest assured there won't be much an audience for such a thing in Jazz or any genre/style of music.🤣

 

However, that approach is exactly how AI will benefit music producers in TV and film and library music.😎

[ElmerJFudd]

5 hours ago, jazzpiano88 said:

I could easily see AI doing some jazz tasks successfully.   Like “reharmonize Somewhere Over the Rainbow”.  It generates the notes.  

It has to be taught that by being fed the right information.  It makes perfect sense that it would be able to do this if it was consuming midi arrangements that exhibit the principals of arranging.  Give a thousand melodies each with 16 possible reharms.  It has to learn something from that.  

 

But throwing it Bill Evans recordings, I don’t know what it thinks it’s making out of chopping up audio and piecing it back together in a different order.  It sounds bizarre and lacking in any understanding of what it’s trying to achieve.  

[stoken6]

1 hour ago, ElmerJFudd said:

chopping up audio and piecing it back together in a different order

That's exactly what it's doing, and it demonstrates a complete lack of appreciation of "what is music" - not only on the part of the AI itself (which cannot demonstrate appreciation for anything - it's just "predicting the next sample" 44100 times per second), but also on the part of the creators, who think that a LAM (Large Audio Model?) is the right tool. 

 

Cheers, Mike.

[ElmerJFudd]

Just now, stoken6 said:

That's exactly what it's doing, and it demonstrates a complete lack of appreciation of "what is music" - not only on the part of the AI itself (which cannot demonstrate appreciation for anything - it's just "predicting the next sample" 44100 times per second), but also on the part of the creators, who think that a LAM (Large Audio Model?) is the right tool. 

 

Cheers, Mike.

That’s exactly what I believe is happening.   

[ElmerJFudd]

What did Adobe build here?    It sounds sh!t.  

 

 

 

[ProfD]

There will always be a demand for incidental music.  Composers will use AI to cut down their production times drastically. 

 

Composer tells AI to create 15 second bumper for a rainbow trout segment on the Fishing channel. Done in the time it takes Keurig to produce a cup of coffee.😎

[jazzpiano88]

Or a new TV network needs a catchy theme:

 

 

 

[CHarrell]

1 hour ago, ProfD said:

There will always be a demand for incidental music.  Composers will use AI to cut down their production times drastically. 

 

Composer tells AI to create 15 second bumper for a rainbow trout segment on the Fishing channel. Done in the time it takes Keurig to produce a cup of coffee.😎

 

In composer production studios, I can absolutely see that being a more common procedure, and in our economic system, a way to stifle work and payments for the "underlings"--that's one of the main reasons many are scared of these technological developments, because our material livelihood and ability to secure the basics is dependent on our ability to produce. 

[Docbop]

12 minutes ago, CHarrell said:

 

In composer production studios, I can absolutely see that being a more common procedure, and in our economic system, a way to stifle work and payments for the "underlings"--that's one of the main reasons many are scared of these technological developments, because our material livelihood and ability to secure the basics is dependent on our ability to produce. 

 

AI will put a lot of ghost writers out of work who made their living writing incidental music.  

[CHarrell]

1 minute ago, Docbop said:

 

AI will put a lot of ghost writers out of work who made their living writing incidental music.  

 

A lot of them have been screwed out of royalties for ages already, especially with the increasing pressures and time crunches that have led to more and more of these studio teams. Sad stories abound of those team head honchos stamping their names on cue lists regardless of whether they actually did any work.

 

On a related tangent: I did a movie about a year ago that played some festivals in Greece. I told my Mom, "Now your son's name will be glazed over and ignored by people across the world!"

She responded, "Well I pay attention."

"Who's the composer for the last movie you saw, Mom?"

"I don't remember."

"I rest my case."  

 

With more behind the scenes roles that take a certain anonymity, those are gonna be the ones most easy to replace and have to be the most diligently protected: I believe in the 60s there was a law or some kind of order passed that mandated that every television show had to have at least a certain percentage of original music as opposed to library music. Like the recent WGA and actor's guild movements, it seems that now would be a very opportune time to revisit and update those kinds of protections.

[AROIOS]

10 hours ago, CHarrell said:

...like the recent WGA and actor's guild movements, it seems that now would be a very opportune time to revisit and update those kinds of protections

This reminds me of another prediction I made back in 2015: Universal Basic Income (UBI).

It's inevitable that we as a species will become economically "useless", facing challenges from A.I. and robots.

Oppressive regimes like North Korea can simply "solve the problem" by starving their "excess" population to death. Luckily, we own guns here, and the AI/Robot corporations and their government cronies will have to throw us a bone reluctantly.

Eventually, we'll exist as code and merge with A.I. Let's just keep our fingers crossed SKYNET doesn't wipe us out before then. 😃

[ProfD]

9 hours ago, AROIOS said:

This reminds me of another prediction I made back in 2015: Universal Basic Income (UBI).

It's inevitable that we as a species will become economically "useless", facing challenges from A.I. and robots.

Agreed.

 

Especially for people incapable of producing tangible goods or services with their own hands.

 

I'm inclined to believe America wants to move towards a society where very few people actually *work* in the truest sense of the word.

 

A UBI will provide just enough income for people to pay for essentials without having to take up their guns.

 

Otherwise, technology could force a labor revolution that inspires people to do real work. Fewer folks sitting in corner offices and twirling around in expensive chairs. 

 

A visionary skilled person or entrepreneur will always be able to earn a living through a combination of labor and technology. 😎

[JoJoB3]

Oh. No no, first will be 'the great battle'.  Cameron was a prophet, Terminator is coming.
In the meantime please enjoy some General MIDI music.

 

[AROIOS]

12 hours ago, ProfD said:

Agreed.

 

Especially for people incapable of producing tangible goods or services with their own hands.

 

I'm inclined to believe America wants to move towards a society where very few people actually *work* in the truest sense of the word.

 

A UBI will provide just enough income for people to pay for essentials without having to take up their guns.

 

Otherwise, technology could force a labor revolution that inspires people to do real work. Fewer folks sitting in corner offices and twirling around in expensive chairs. 

 

A visionary skilled person or entrepreneur will always be able to earn a living through a combination of labor and technology. 😎

Yup, for now, I'd choose living within a self-sustaining "Intentional Community" over being a farm animal in those "15-minute cities", hooked on UBI and VR/Metaverse goodies.

But I'm not delusional about how weak our species' will power is. The gaming industry has already been bigger than movie and music combined for a few years. And humans are only gonna get increasingly addicted as virtual experiences improve.

 

Maybe all it takes for me to give in, is a VR trip to VR Caribbean with a VR girlfriend, on my VR yacht. 😆

[AROIOS]

12 hours ago, JoJoB3 said:

Oh. No no, first will be 'the great battle'.  Cameron was a prophet, Terminator is coming.
In the meantime please enjoy some General MIDI music.

 

Judging by the term "General MIDI", I highly suspect you're one of them T-1000s who just got wormholed here from 1990.

[JoJoB3]

7 hours ago, AROIOS said:

Judging by the term "General MIDI", I highly suspect you're one of them T-1000s who just got wormholed here from 1990.

80's!! 🤘

(kind of thinking that old GM is still better than all the Ai I've heard).
 

 

[AROIOS]

20 hours ago, JoJoB3 said:

80's!! 🤘

(kind of thinking that old GM is still better than all the Ai I've heard).

Speaking of GM, Passport produced a bunch of excellent GM midi files back in the day. Some of them got shipped with early versions of Windows and even garnered fans decades after those OSes were released.
 

Early GM support and sonic performance were all over the place, but it paved the way for Roland's GS and Yamaha's XG formats to come.
 

Roland hired a Japanese keyboard wiz duo "IDECS" in the early 90's to produce a series of excellent MIDI files for their SC-55 GS synth. It became a line of commercial MIDI packs they sell.
 

Yamaha soon followed suit and commissioned several American and British musicians to produce a vast library of original demo tunes for their XG synths. Like Roland, they also sold lots of XG MIDI packs.
 

To this day, these GS and XG MIDI files remain the best sounding MIDI productions I've heard.
 

 

 

 

[CyberGene]

I hope this doesn’t get interpreted as disrespectful but IMO jazz is the easiest task for a proper generative model. Jazz is so formalized with all possible devices of improvisation, (re)harmonization, voicing and rhythm structures described and taught to death in any possible online or academia courses, that it’s actually a wonder to me why it hasn’t been done earlier. 
 

As to Bill Evans, I love the guy and his approach (maybe exactly because of what I’m going to say) but his playing is extremely predictable and formalized to the point of being the easiest to use for ML, not sure it even needs a lot of training.

 

While being the next big genius such as Herbie, Chick, Bill Evans, Lyle Mays, etc. (or Bach, Beethoven, Mozart, Chopin, Debussy, etc. for that matter) is a difficult task for either a mortal or AI (because we’re still in the early stages of AI where the models can only learn and replicate/imitate but not innovate), making an acceptable jazz improvisation is a matter of practicing enough, it’s not a divine stroke. Perfect task for an automat. 

[stoken6]

18 minutes ago, CyberGene said:

I hope this doesn’t get interpreted as disrespectful but IMO jazz is the easiest task for a proper generative model. Jazz is so formalized with all possible devices of improvisation, (re)harmonization, voicing and rhythm structures described and taught to death in any possible online or academia courses, that it’s actually a wonder to me why it hasn’t been done earlier. 

I see a gulf between "traditional musicians" and AI specialists working on AI music. It seems to me that the AI types haven't got their head round what makes music what it is. I'll give a simple example: Avid (for some reason) introduced AI chord recognition in a recent version of Sibelius. It looks 60 bars back for context. Why 60? (64 would make more sense). 

 

What's needed to achieve what CyberGene suggested is appropriate convolution/feature engineering that's aligned to musical (say Jazz) features: phrases, 2-5-1, tritone subs, bracketing/enclosures, voicings, choice of extensions, repetition a semitone away etc. An AI should be able to learn that a dominant chord built on a degree other than V often has a #11 - but it has to be presented with data in that form, that makes those features visible. 

 

Next step after that is for the AI to learn these features itself (deep learning), given just the raw data in say MIDI format. This would be equivalent (in a very abstract way) to some of the convolutional "pre-processing" that's used in front of neural networks. 

 

The recent example on AI "Jazz" had none of this "feature awareness". I would bet three digits of currency that the AI was trained on 100msec snippets of audio, and it's just shuffling those around, like a LLM (ScatGPT?)

 

Cheers, Mike.

[nursers]

[Verse] Come on down to The Keyboard Corner Where the ivories play all day and night (ooh-yeah) A forum for players, beginners and pros We gather 'round, sharing our delight

 

[Chorus] At The Keyboard Corner, we find our melody From bebop to swing, improvising in harmony Together we create, a magical sound At The Keyboard Corner, where jazz is found (ooh-yeah)

[JoJoB3]

5 hours ago, CyberGene said:

I hope this doesn’t get interpreted as disrespectful but IMO jazz is the easiest task for a proper generative model. Jazz is so formalized with all possible devices of improvisation, (re)harmonization, voicing and rhythm structures described and taught to death in any possible online or academia courses, that it’s actually a wonder to me why it hasn’t been done earlier. 
 

As to Bill Evans, I love the guy and his approach (maybe exactly because of what I’m going to say) but his playing is extremely predictable and formalized to the point of being the easiest to use for ML, not sure it even needs a lot of training.

 

While being the next big genius such as Herbie, Chick, Bill Evans, Lyle Mays, etc. (or Bach, Beethoven, Mozart, Chopin, Debussy, etc. for that matter) is a difficult task for either a mortal or AI (because we’re still in the early stages of AI where the models can only learn and replicate), making an acceptable jazz improvisation is a matter of practicing enough, it’s not a divine stroke. Perfect task for an automat. 

Nope. Too much wrong here. 
Stamped: disagree, and may gawd have mercy on your soul.

[CyberGene]

One of democracy’s greatest achievements is being able to disagree with people without having to kill them 😀

[ElmerJFudd]

1 hour ago, CyberGene said:

One of democracy’s greatest achievements is being able to disagree with people without having to kill them 😀

Civil debate is in fact wonderful.   Even better is arriving at compromise.  

[Threadslayer]

What planet are you guys calling in from? It certainly sounds nice.😉