Popular YouTube Artist Uses AI To Record New Album (theverge.com)

An anonymous reader shares a report from The Verge of a popular YouTube artist who is using artificial intelligence to produce a LP: If you heard Taryn Southern's new single "Break Free" on the radio, you'd probably just keep driving or grocery shopping, or doing whatever you do in places that still have radios playing. The song is a big, moody ballad -- the kind that might play during the climax of a Steven Spielberg movie. "Break Free" wasn't composed by a John Williams copycat, but by artificial intelligence. The song is not a fluke or a novelty for Southern either; she's using artificial intelligence platforms to create an entire album, called I AM AI. It's the first LP to be entirely composed and produced with AI. Southern used an open source AI platform called Amper Music to create the stems of "Break Free." For each track, she plugs in genre, the instruments she wants to use, and beats per minute. In return, Amper churns out disjointed verses that can be rearranged into a song, and layered beneath Southern's vocals. Southern told The Verge she's toying with four other AI music platforms, but she's not sure which of those will make the final album cut.

Comments

Nice

[TimothyHollins]

For once it might be Beyonce and Justin Bieber worrying about their jobs getting outsourced instead of programmers and service personnel. I like this new plot twist.

Entirely

[arth1]

It's the first LP to be entirely composed and produced with AI. Southern used an open source AI platform called Amper Music to create the stems of "Break Free." For each track, she plugs in genre, the instruments she wants to use, and beats per minute. In return, Amper churns out disjointed verses that can be rearranged into a song, and layered beneath Southern's vocals

That's an interesting definition of "entirely".
I would say "partially" is a better word.

Re:Entirely

[PopeRatzo]

That's an interesting definition of "entirely".
I would say "partially" is a better word.

You're correct. In fact, it's only a few steps removed from playing a Casio keyboard with the arpeggiator turned on and randomly pressing the white keys.

I've used Amper Music and it's an interesting platform. Algorithmic and generative music certainly is interesting, but like this song, it's kind of dull.

A more interesting application is having an AI produce a musical score which is then performed by human musicians. Gottfried Koenig does something like this in a more traditional Western Classical tradition rather than pop music. I prefer the live interaction of man and machine to the "throw a bunch of standard chord progressions into some software".

Re:Entirely

[Mal-2]

That's an interesting definition of "entirely".
I would say "partially" is a better word.

I'd be inclined to agree. I'd say this is as "entirely" produced by AI to the same degree as my own work has been for years, and many others as well. It's called Band In A Box, which has been around (though not with all the features it currently has) for almost 30 years.

I also have tracks where the entire composition was done by cgMusic, and I did all the arrangement and production, like this one, or where the computer did 80% of the composition, such as this or this.

To musicians in my position, who create their tracks from the ground up all the way to mastering, none of this is even remotely new. We've been doing it not for years, but for decades. It wasn't called AI, and "expert system" is probably more appropriate, but it doesn't change the way it is used or the results obtained.

Obig Orwell

[seven of five]

"Here were produced rubbishy newspapers containing almost nothing except sport, crime and astrology, sensational five-cent novelettes, films oozing with sex, and sentimental songs which were composed entirely by mechanical means on a special kind of kaleidoscope known as a versificator. "

LP????

[OzPeter]

In what world does a music download equate to an LP??

Yeah get off my lawn.

OTOH This will be the Autotune of the future. I can't say that I am inspired by that song (seemed pretty generic and soulless to me) but I am sure I will be surprised someday soon.

Re:They'll be alright.

[geekmux]

The AI composes the songs - not performs them. And many pop stars don't write their own songs.

Elvis Presley had an awesome career singing other people's songs.

Unlike damn near every entertainer today, Elvis Presley didn't need Autotune.

And the only reason they'll be "alright" is because the music standard has dropped so fucking low that we don't even call them musicians anymore; they're now entertainers. The fans will welcome these AI-writing overlords and not miss a beat.

Re:They'll be alright.

[mark-t]

I think it depends if you define "musician" to be synonymous with singer.

I would not consider a singer to be a musician *unless* they also wrote songs.

Re:Copyright all possible variations of "music"

[Bios_Hakr]

There should be an easier way.

Music is encoded at 44khz * 24-bit samples. That's about 1Mbits per second of audio.

Why not just run through all the permutations of possible 1s and 0s and copyright that? You'd literally have every possible 1-second sound that could exist.

Now, as to the computational power you'd need, well, there's the rub.

My experience

[Tablizer]

I have experimented with computer-composed music, and conclude it's fairly easy to make "pleasant" music: the rules for western ear expectations are fairly straight forward; stay within the rules and it'll sound "normal".

Markov-like chains of "hit tune" chord progressions can be mined, for example, to find decent chords. And there are fairly obvious rules for how the melody moves around a given chord and chord transitions, if you simply study diagrams/plots of several hits. The melody "likes" certain distances and relationships with regard to the active chord, sort of reminiscent of the probabilistic modelling of electron orbits (positions) relative to the center of the the atom. One can fine-tune the extraction of such patterns/formulas using statistics and AI (although I just eye-balled most of it myself and used some trial-and-error).

Now whether anyone finds music generated from such pattern mining highly entertaining and/or moving is another matter. For one, such is subjective: a pattern or technique that one person really likes, another may not. The hard part is not really generating ideas, but culling them. AI may indeed be able to cull based on patterns of existing hits to find a bigger audience, but after a while the easy-to-find patterns will get tired and stale to listeners. The low-hanging fruit will be all picked and people will want something different. It's why fads like disco, auto-tune, emo rock, and techno come and go.

Even if AI-using composers hone their techniques of generating candidate music, the hard part will still be culling against actual listeners: after all, AI does not buy music: only actual human "testers" will suffice. The cost of testing (getting listeners) will be about the same regardless of whether a tune is composed by a human or machine. Thus, you need to make sure the bot's quality is competitive with a human, or you are wasting money. On the web you still have to pay for attention (unless you get really lucky with cute hamsters or the like).

Another thing, my "music machines" tend to sound like my hand-composed music. It's modelling my own head more or less being I tune it based on my preference and by grabbing samples of stuff I like in order to extract patterns from. Thus, the artificial composer is sort of an extension of myself. It's just "implementing" my preferences.

Re:They'll be alright.

[geekmux]

Puhleez, if you honestly though 90% of musicians wrote their own songs you are sadly mistaken.

Music used to be written and created. It held emotion and power.

Today, pop music is manufactured predictable shit. It's all based on a market-validated recipe of some brainless hottie laying oversexualizing Autotuned vocals on top of revenue-generating beats. The shift to letting AI create "music" only validates how truly easy it is to manufacture shit for the masses.

As I said before, the music standard has dropped lower than a dubstep track.

Generative music

[sound vision]

This is called generative music, and it's not new. Since this is Slashdot, you may recall that the music in Spore was entirely machine-generated, in real-time no less. No pre-recording, no overdubs. The man behind the Spore music, Brian Eno, has been experimenting with music based on rules and randomness (but not computers) since 1975, using a deck of special cards, calling it "oblique strategies".
https://en.wikipedia.org/wiki/...

Calling it "AI" instead doesn't make it new. But hey, I guess some random YouTube douche needSUBSCRIBE NOW SUBSCRIBE NOW SUBSCRIBE NOW