Alison Parshall: This is Scientific American science, quickI’m Alison Parshall.
I’m going to play you a songAnd I’m willing to bet the good money you have I never have I heard something like that before.
[CLIP: Beginning of “Enter Demons & Gods,” by Yaboi Hanoi]
partial: The first time I heard it, I was intrigued and puzzled. My head was fluttering to the familiar beat, Note I didn’t hear anything at all.
Is it a sound that you can play while sitting at the piano?
[CLIP: Notes on piano]
And the melody doesn’t even know what instrument sounds like it.
After all, there is no such instrument. But for those familiar with Thai music and Thailand’s national sport, Muay Thai, his boxing may sound familiar.
[CLIP: Muay Thai boxing background sounds]
Lam Tarn “Hanoi” Han Trakl: For Thai listeners, it’s like being in the uncanny valley of the familiar as well as the alien. If you want to move, we have succeeded in connecting with you.
partial: That is Lamtharn “Hanoi” Hantrakul. He’s a music technologist, the brains behind this whole new breed of music.
He was born and raised in Bangkok. His nickname “Hanoi” is admittedly confusing considering it’s the capital of Vietnam, but his parents love the city so much that they gave their child a name.
Hanoi is Thai at heart. And musically, he’s known as Yavoi Hanoi, but it kind of started as a joke…
Huntracool: This geeky, music techie was like, ‘Yeah, I’m a cool guy. yaboi Hanoi. “
partial: But people really liked it, so it stuck. And he created the piece you heard a minute ago is a non-human assistant, machine learning, a form of artificial intelligence.
See, artificial intelligence algorithms are musicians now.
Welcome to Part 1 of our 3-part “Fascination”. science, quick A miniseries, FYI, about how artificial intelligence is creeping deeper into the world of music.
so about the whole AI has become a musician. Almost kidding. These computer his algorithms, no matter how sophisticated, are really a reflection of our own creativity and ingenuity.
They are given a large amount of data to learn from and can detect subtle patterns that can be used to make useful predictions in many artistic activities, not just music.
And they’re getting really good at it. In fact, these algorithms are advancing very rapidly, felt They seem to do more than just pattern recognition.
ChatGPT, a large-scale language learning model AI, can create poems about inheritance tax and methamphetamine. You can create a recipe for “French-style chicken thigh with carrot and cream”. Alternatively, you can make your job application cover letter less sarcastic.
A deep learning model, Dall-E 2 can create fantastical art from simple verbal prompts such as ‘a bowl of soup, a gateway to another dimension in the style of Basquiat’.
And now, people like Hanoi are using machine learning to challenge the musical chokehold that Western Europe has basically brought to all of us.
This brings us back to “Enter Demons & Gods,” the Machine Learning-powered composition of Hanoi that started in the episode. In 2022, he submitted it to an international competition called the AI Song Contest.
he won
[CLIP: Recording of award ceremony]
announcer: This means that the total voters and judges scored 21.1 points…Yaboi Hanoi is the winner of AI Song Contest 2022!
[CLIP: Applause]
[CLIP: Hantrakul: Yeah, I’m, I’m just completely lost for words. I’m just so excited that the song spoke to so many people both at home in Thailand … and that it also spoke to the jury.]
Huntracool: I think the fact that your ears aren’t accustomed to it really heightens the senses of many Western listeners.
partial: He mentions 12 equal temperament tuning. Even if you’ve never heard its technical name, you’ve definitely heard it in action. Able to play on the piano, he represents the 12 recurring notes and notes that basically underlie all Western pop and classical music. See, you have an octave — high notes are exactly twice the frequency of low notes…
[CLIP: Piano plays an octave of A at 440 and 880 hertz]
Western music divides the octave into 12 notes and spreads them out evenly.
[CLIP: Ascending chromatic A scale on the piano]
Hanoi knows this tuning system well. Like many people around the world, his mother encouraged him to play the piano when he was a child.
Huntracool: My mother once said, “If you understand how to play the piano, you will understand all other instruments.” And she remembers, after she graduated music major, she came back to her and said,
partial: That’s true. His 12 notes on the piano can sound natural and normal if you grew up listening to mostly Western music.
please listen. I’m going to geeky out a little bit. i want you to be with me The sound of a piano called A4 is like this…
[CLIP: 440 Hz tone]
…usually the frequency is set to 440 Hertz. This means that sound waves vibrate 440 times per second.
But that 440 number is nothing special. It was first standardized in the 1930s. Also, some orchestras, such as the New York Philharmonic, tune to 44.2 hertz. This is such a small difference that we can hardly hear it. However, during the Baroque period around the 1600s, the same note was set at 415 Hz.
[CLIP: 415 Hz tone]
All this is to say that pitch is a spectrum and there are literally endless sounds you can tune to play your instrument.no real reason offal Divide the octave into 12 notes instead of 22 or 5.
Nor is there any real reason why these notes should be evenly spaced across the octave. Many musicians claim that the 12 notes of the Western scale are notes. Better when they are no Spread evenly.
And finally, the tuning system doesn’t even need to be octave based. Sure, the laws of physics and the biology of our ears may make us tend to feel comfortable with octaves, but we humans are creative and flexible creatures.
So given these endless possibilities, it makes sense that cultures around the world choose to carve up the musical space differently. Neither Thai fiddle…
Huntracool: of saw you. of hello duan …
partial: Fit a 12-tone model. This is closer to the model that musicologists call his 7-tone equal temperament tuning. This fits him in 7 notes in the same amount of musical space that western music fits in his 12 notes.
Similarly, Pinay, Instruments like the oboe played in Thai boxing matches play notes and intervals that cannot be imitated by a piano.
In fact, if you take this Pinaytril…
[CLIP: Pi nai trill]
partial: The closest notes I can play on the piano sound like this…
[CLIP: Pi nai piano approximation]
partial: Their pitch is kinda close, but not Pinay. But these are the pitches our music technology was designed to work with.
Huntracool: And I think that was really the starting point I think. What if we could use technology to write music on the terms of Thai music?”
partial: That’s where machine learning comes in. There has been a pretty big revolution in how machine learning works with music.
For a long time, the music we made with AI was limited to 12 piano notes.
This is because the raw audio file itself is Large scale, encoding tons Of information.Normally 44,000 It samples every second to make high quality recordings. Double for stereo recordings where each ear of the headphone plays a different track.
So, let AI bite Led Zeppelin’s “Stairway to Heaven”…
[CLIP: Led Zeppelin’s Stairway to Heaven]
…it would be over 42 million Data points to process.That’s all too much for AI algorithms.
So those old algorithms needed something simpler. All they could handle was a symbolic, text-based representation that captured the gist of the note being played…
[CLIP: MIDI version of Led Zeppelin’s Stairway to Heaven]
… like notes written on a page to explain how to play a song on the piano. And here we come across our nemesis, his limited 12-tone scale.
[CLIP: Reprise of the 12-tone approximation of the pi nai]
Huntracool: The moment your music strays from this tuning system, there’s really no way these iconic models can even comprehend or understand some of these melodies.
partial: So the old Symbolic model wouldn’t have known what to do with the Pinay. Fortunately, our computing power is finally catching up to our audio ambitions.
Hanoi is one of the engineers who made this possible through his previous work at Google Magenta and his current work at TikTok.
He has developed a machine learning tool that can pick up the melody being played on one instrument. And within seconds, Transform into another instrument.
Huntracool: There were these very fun demos I worked on when I was at Google. A classical Indian song can be transformed as if it were played on a saxophone, or a bird call can be re-rendered as follows: flute.
[CLIP: Birds re-rendered as a flute using Google’s Tone Transfer]
partial: And for Hanoi, these machine learning tools were what made his operations possible. directly Pinay recordings do not need to be filtered. sieve Western music notation.
Listen to the trill again.
[CLIP: Pi nai trill]
partial: The recording features Thai musician Udomkiet Joey Phengaubon playing standard Thai classical melodies.
Hanoi contributed this recording via Mawf, one of the AI tools he helped develop. We have extracted the unique characteristics of its tuning and timbre and re-rendered it as a separate instrument. The first was the sax…
[CLIP: Pi nai as saxophone]
partial: Then the trumpets…
[CLIP: Pi Nai as trumpet]
partial: And another Thai instrument for good measure.
[CLIP: Pi nai as khlui flute]
Huntracool: It’s like five instruments layered together with a lot of distortion. So who knows what it is?
[CLIP: All instruments together; music reenters]
partial: Hanoi describes his work as “Entering the Demons and Gods”, or in Thai…
Huntracool: “Ashura Deva Chum Noom”
partial: It is inspired by a scene in Thai mythology where hundreds of gods and demons legendary clash over the elixir of immortality.
[CLIP: Battle sound effects in the background]
In the end, he basically created a Thai-Western fusion EDM track.
Huntracool: I don’t want to call it cultural preservation. I call it cultural revitalization. I find it incredibly liberating. I feel like I’ve been speaking English my whole life.
partial: And now he speaks Thai and has no intention of stopping.
Huntracool: It was kind of a big bang moment for me to have to write more music like this. This is the moment when Thai passages, which could not be reinterpreted beyond Thai classical music, finally transcend this dimension and move to electronic music. To me, this is the definition of how AI can empower creative humans.
[CLIP: Music ends]
partial: AI Song Contest judges and public voters loved Hanoi’s song. The competition is also expanding each year with new music artists using AI in different ways.
But from flattened text-based representations to the rich world of live audio, this revolution in music AI harnessed by Hanoi has far-reaching consequences.
In the next episode…
[CLIP: Christine McLeavey: “Working with raw audio, the sky is the limit in terms of what you can create.”]
[CLIP: Shelly Palmer: The amount of things that have to be true for this to be what it is are unbelievable—like, unbelievable.”]
partial: You’ll hear how these algorithms have advanced tremendously with recent models created by Google. click of the button.
science, quick Produced by Jeff Dervicio, Turika Bose and Kelso Harper. Our theme music was composed by Dominic Smith.
don’t forget to subscribe science, quick Get podcasts anywhere. For more scientific news and features, visit ScientificAmerican.com.
for Scientific Americanof science fastI’m Alison Parshall.
