Google's advanced music generation model and two new AI experiments • Hacker News Board

Re: Google's advanced music generation model and two new AI experiments

#41

Post by halflings » Thu, Nov 16, 2023, 9:57 AM UTC

bambax wrote: post #36

Some examples are interesting (the sax, the string ensemble); the "realistic vocal choir" is anything but. I tried riffusion and sang 10 seconds of "the house of the rising sun" [0] It produced... this... https://www.riffusion.com/riffs/d0464fb4-a53a-4ad8-a765-0a64... which, by some measure, is impressive (perfect recognition of lyrics, for one), but is also... a musical abomination? So IDK. Let's see where this goes…

Growing pains. Remember how this all started (Google Magenta, circa 2019): https://www.youtube.com/watch?v=ZRnbbtqxBEc

Re: Google's advanced music generation model and two new AI experiments

#42

Post by raincole » Thu, Nov 16, 2023, 9:58 AM UTC

vouaobrasil wrote: post #11

From the article: > Each participating artist has partnered with us and will have a hand in helping us test and learn to shape the future of AI in music. Let me correct that: each participating artist has been given a wad of dirty cash in return for their help in training replacements for mass music. Notice that I said "mass music" -- people will always be able to create by hand, but it's clear that the ultimate goal…

> It creates a new standard where music is no longer about human expression but about creating a basic layer of entertainment to go along with advertising.

The most music we (humans collectively, not you and me specifically) hear are made for one single purpose: profit.

It's not an AI problem. It's nothing new. Actually it's older than you and me.

Re: Google's advanced music generation model and two new AI experiments

#43

Post by earthnail » Thu, Nov 16, 2023, 9:58 AM UTC

bambax wrote: post #36

Some examples are interesting (the sax, the string ensemble); the "realistic vocal choir" is anything but. I tried riffusion and sang 10 seconds of "the house of the rising sun" [0] It produced... this... https://www.riffusion.com/riffs/d0464fb4-a53a-4ad8-a765-0a64... which, by some measure, is impressive (perfect recognition of lyrics, for one), but is also... a musical abomination? So IDK. Let's see where this goes…

When you sing into Riffusion, it just puts time constraints on the lyrics. Example:

"I've never seen a diamond in the flesh" from Lorde, sung into Riffusion: "I've:0.72-0.96 never:0.96-1.20 seen:1.20-1.48 a:1.48-1.64 diamond:1.64-2.32 in:2.32-2.76 the:2.76-3.02 flesh:3.02-3.86"

With the timestamps being second-annotations. The Google model takes the whole melody and phrasing as input as well.

Re: Google's advanced music generation model and two new AI experiments

#44

Post by barrkel » Thu, Nov 16, 2023, 9:58 AM UTC

nervousvarun wrote: post #26

gardenhedge wrote: post #6

Wow, AI really will democratize music creating. This is an awesome way for people to translate their ideas to music without needing to be privileged enough to get violin or guitar lessons. I see a future where we all make our own music and share it freely with family and friends. This means less listening to manufactured pop songs written by suits with the sole aim of making a profit. This will still happen of course…

What “privilege” is required to learn to play guitar? Was Jimmy Hendrix privileged? Robert Johnson?

Hand and finger privilege. It's much harder to play if you don't have hands or fingers!

Re: Google's advanced music generation model and two new AI experiments

#45

Post by soulofmischief » Thu, Nov 16, 2023, 9:58 AM UTC

gardenhedge wrote: post #6

Wow, AI really will democratize music creating. This is an awesome way for people to translate their ideas to music without needing to be privileged enough to get violin or guitar lessons. I see a future where we all make our own music and share it freely with family and friends. This means less listening to manufactured pop songs written by suits with the sole aim of making a profit. This will still happen of course…

> I see a future where we all make our own music

What will it mean to "make music" when the amount of necessary skill, experience and insight approaches zero?

> This means less listening to manufactured pop songs written by suits with the sole aim of making a profit

Unfortunately, unless we replace capitalism, this is exactly what you can expect. The veil of false corporate authenticity grows every day, you can see it in the opening sentence of this press release:

"From jazz to heavy metal, techno to opera, music is a much loved form of creative expression."

The statement, while true on the surface, takes on a different tone when you remember who the speaker is, who the audience is, and what their ultimate goal is.

I am excited as a musician to explore this coming age, and I am also deeply saddened and fearful of how things will shake out.

Re: Google's advanced music generation model and two new AI experiments

#46

Post by Juliate » Thu, Nov 16, 2023, 9:58 AM UTC

Mashimo wrote: post #16

iamsaitam wrote: post #13

Earlier quoted context omitted.

This is sarcasm right?

You don't think this will lower the barrier of entry for music creation?

The entry for music creation is really not high.

It first resides in your own, personal, self appreciation: do you feel yourself legitimate to sing, to play, to say something, for yourself, or to present to others?

That's the only valid question.

The rest, all the rest, comes from that. And it does not require to be complicated or complex or clever music at this point.

Sure, you may start with the practice, with the education, the experimentation, if you have not answered for yourself that first question, you won't create/play music or songs, you'll just apply a rulebook.

Then, another totally different question may be, can you find a way to get a sufficient revenue from that activity. But that's another question still.

Re: Google's advanced music generation model and two new AI experiments

#47

Post by kkzz99 » Thu, Nov 16, 2023, 9:58 AM UTC

vouaobrasil wrote: post #22

sirwitti wrote: post #20

Earlier quoted context omitted.

Quality never was the only or even the most important aspect of popularity in music and probably art in general. So I'm not sure your projection is valid.

His projection is valid because it's not a qualitative problem, it's a quantitative one. Music may already have been turned into a commodity selected by traits other than quality, but tools like this bring the effect up to a whole new level, a whole new order of magnitude.

Every tool that makes it easier to create art is good in my book. In years past you had to actually have the instruments and had to know how to work them. Now you can much easier learn to use simpler tools to get even better results, which just increases the amount of people that can actually realize their creativity.

Re: Google's advanced music generation model and two new AI experiments

#48

Post by vouaobrasil » Thu, Nov 16, 2023, 9:59 AM UTC

taopai wrote: post #37

vouaobrasil wrote: post #11

From the article: > Each participating artist has partnered with us and will have a hand in helping us test and learn to shape the future of AI in music. Let me correct that: each participating artist has been given a wad of dirty cash in return for their help in training replacements for mass music. Notice that I said "mass music" -- people will always be able to create by hand, but it's clear that the ultimate goal…

I guess someone will make it, if not google or others will. This is the next iteration of music. First we had only live concerts, then we had recordings, and finally auto-generated music. I hope some old dudes with long beards will like to hang out and jam with me when I'll be old.

If not google, others will? That is like justifying stealing from an old man on the street by saying that because the street is dangerous, someone else will do it anyway.

This is not the next generation of music. It is one possible future that we can oppose if only we were not so convinced that technology is completely deterministic and can't be opposed.

Re: Google's advanced music generation model and two new AI experiments

#49

Post by amarant » Thu, Nov 16, 2023, 9:59 AM UTC

vouaobrasil wrote: post #11

From the article: > Each participating artist has partnered with us and will have a hand in helping us test and learn to shape the future of AI in music. Let me correct that: each participating artist has been given a wad of dirty cash in return for their help in training replacements for mass music. Notice that I said "mass music" -- people will always be able to create by hand, but it's clear that the ultimate goal…

As a indie game developer, this could plausibly enable me to add decent sounding soundtrack to my games. I couldn't afford to pay a proper artist, and my musical skills are... Limited..

It _is_ just a tool, a new instrument essentially, one that is easier to play than most. People being angry that there is now a new instrument that is much easier to play than the one they spent years to learn come across as entitled brats to me. This will enable me to express myself in a way I couldn't before, how is that not about human expression?

It's a accessibility thing as well. Expressing oneself through music has always been nearly impossible to the tone-deaf among us. Now that may change, and you want to stop that because it might maybe potentially affect how much money you can earn from playing music?

Glory to Google for enabling the musically impaired to express themselves through music!

Re: Google's advanced music generation model and two new AI experiments

#50

Post by taopai » Thu, Nov 16, 2023, 9:59 AM UTC

vjerancrnjak wrote: post #35

vouaobrasil wrote: post #11

From the article: > Each participating artist has partnered with us and will have a hand in helping us test and learn to shape the future of AI in music. Let me correct that: each participating artist has been given a wad of dirty cash in return for their help in training replacements for mass music. Notice that I said "mass music" -- people will always be able to create by hand, but it's clear that the ultimate goal…

But isn't music more than just the music track? People like to listen to the same music, go to concerts, share it with others, share the experience of audience/artist performance. Moments where music is truly important in one's life will never be replaced by meaningless bit fiddling. I cannot imagine people caring about books, music, art created by bit fiddling. They will always care about things created by other peo…

Well, this could have the potential to make infinite covers of your favorite song.

But think about it, people used to watch the same movies in DVD. The movies they loved. Now people watch and infinity of short random videos (tik tok, instagram)...

Technology changes the habits of people, specially in young people. Think about it, one friend sends a song auto-generated from some text, it is "a song made by someone", at least that will be their perception...