- Email: [email protected]
AI melody maker
TECHNOLOGY
ONE PER CENT
AI melody maker Machine learning can turn lyrics into songs, finds Gian Volpicelli it will fit with the previous five I’m all alone”. They also fed it the notes. To write a song with ALYSIA, lyrics to the vaudeville classic I’m the songwriter feeds it lyrics Always Chasing Rainbows to see broken down into individual lines. how it could reimagine the song The system will give a variety of in the pop genre (arxiv.org/ melody options for each segment abs/1612.01058). that the user can choose between. The idea of trying to automate Alternatively, ALYSIA can write musical composition is not new, a complete tune by selecting its but David Cope at the University of first melody option for each line. California, Santa Cruz, says ALYSIA Ackerman and Loker developed is unusual in taking lyrics as its the system to produce pop songs, starting point. He is impressed but say it could be adapted for that the system manages to match different genres. “Over time, I started to The pair trained two models, think of computers separately focused on rhythm as creative partners and pitch, on the melody and instead of tools” lyrics of 24 pop songs. They then used the system to make melodies to accompany two the metre of the melody with sets of words written by Ackerman, that of the lyrics, but says the with lyrics such as “Now that compositions show an “almost you’re gone / I just realised that annoying” lack of harmony. Rebecca Fiebrink, a researcher in machine learning and music at Goldsmiths, University of London, questions how useful the lyrics-to-melody approach is. “Is this really solving the compositional process for people who want to make music?” she says. “Creating a melody without additional accompaniment, like this system does, is the easiest thing to achieve.” The songs admittedly aren’t about to win any Grammys, but Ackerman says this is just the start. She initially imagined targeting ALYSIA at the electronic music community, but is now working with classical composers to repurpose it for professional songwriters. Ultimately, Ackerman hopes to create a system capable of composing all aspects of a song on its own. “We want to design a program able to generate the music, the lyrics, and ideally even the production and the singing by –La la la… now what next?– itself,” she says. n
22 | NewScientist | 17/24/31 December 2016
Blow-up batteries Ever taken the back off your phone and seen that your battery has swollen? Scientists at the Canadian Light Source synchrotron facility used X-rays to see what is happening inside lithium-ion batteries when they swell. Part of the problem, they say, is their “Swiss roll” structure, which can lead to pillowing if gas forms between the layers of electrodes (Journal of the Electrochemical Society, doi.org/bvms).
163
million: the view count of the most-watched Facebook Live video, Candace Payne’s viral “Chewbacca Mom”
Blast heat to space A cooling method that blasts heat into space has achieved record levels of efficiency. Zhen Chen and his team at Stanford University cooled a thermal emitter to 42.2 °C below the ambient temperature using radiative cooling, which releases heat at wavelengths that pass through the atmosphere (Nature Communications, DOI: 10.1038/ncomms13729). It could eventually be used to refrigerate food and medicine in hot places, Chen says.
TOBY BOND, CANADIAN LIGHT SOURCE
KIKO JIMENEZ/PLAINPICTURE
GOT words but no melody? A machine learning system turns lyrics into song by composing a pop music score to suit the words it’s given. “I was studying singing while I was doing my PhD in computer science,” says Margareta Ackerman at San Jose State University in California. “Over time, I started to think of computers as creative partners instead of tools, which could maybe help me write songs.” This inspired her to develop the AI system, along with David Loker at technology consultancy Orbitwerks. ALYSIA, as the system is called, processes short lines of text and associates each syllable with a musical note. It chooses the pairing based on features including the syllable’s position in the word and, where relevant, how
ONE PER CENT
AI melody maker Machine learning can turn lyrics into songs, finds Gian Volpicelli it will fit with the previous five I’m all alone”. They also fed it the notes. To write a song with ALYSIA, lyrics to the vaudeville classic I’m the songwriter feeds it lyrics Always Chasing Rainbows to see broken down into individual lines. how it could reimagine the song The system will give a variety of in the pop genre (arxiv.org/ melody options for each segment abs/1612.01058). that the user can choose between. The idea of trying to automate Alternatively, ALYSIA can write musical composition is not new, a complete tune by selecting its but David Cope at the University of first melody option for each line. California, Santa Cruz, says ALYSIA Ackerman and Loker developed is unusual in taking lyrics as its the system to produce pop songs, starting point. He is impressed but say it could be adapted for that the system manages to match different genres. “Over time, I started to The pair trained two models, think of computers separately focused on rhythm as creative partners and pitch, on the melody and instead of tools” lyrics of 24 pop songs. They then used the system to make melodies to accompany two the metre of the melody with sets of words written by Ackerman, that of the lyrics, but says the with lyrics such as “Now that compositions show an “almost you’re gone / I just realised that annoying” lack of harmony. Rebecca Fiebrink, a researcher in machine learning and music at Goldsmiths, University of London, questions how useful the lyrics-to-melody approach is. “Is this really solving the compositional process for people who want to make music?” she says. “Creating a melody without additional accompaniment, like this system does, is the easiest thing to achieve.” The songs admittedly aren’t about to win any Grammys, but Ackerman says this is just the start. She initially imagined targeting ALYSIA at the electronic music community, but is now working with classical composers to repurpose it for professional songwriters. Ultimately, Ackerman hopes to create a system capable of composing all aspects of a song on its own. “We want to design a program able to generate the music, the lyrics, and ideally even the production and the singing by –La la la… now what next?– itself,” she says. n
22 | NewScientist | 17/24/31 December 2016
Blow-up batteries Ever taken the back off your phone and seen that your battery has swollen? Scientists at the Canadian Light Source synchrotron facility used X-rays to see what is happening inside lithium-ion batteries when they swell. Part of the problem, they say, is their “Swiss roll” structure, which can lead to pillowing if gas forms between the layers of electrodes (Journal of the Electrochemical Society, doi.org/bvms).
163
million: the view count of the most-watched Facebook Live video, Candace Payne’s viral “Chewbacca Mom”
Blast heat to space A cooling method that blasts heat into space has achieved record levels of efficiency. Zhen Chen and his team at Stanford University cooled a thermal emitter to 42.2 °C below the ambient temperature using radiative cooling, which releases heat at wavelengths that pass through the atmosphere (Nature Communications, DOI: 10.1038/ncomms13729). It could eventually be used to refrigerate food and medicine in hot places, Chen says.
TOBY BOND, CANADIAN LIGHT SOURCE
KIKO JIMENEZ/PLAINPICTURE
GOT words but no melody? A machine learning system turns lyrics into song by composing a pop music score to suit the words it’s given. “I was studying singing while I was doing my PhD in computer science,” says Margareta Ackerman at San Jose State University in California. “Over time, I started to think of computers as creative partners instead of tools, which could maybe help me write songs.” This inspired her to develop the AI system, along with David Loker at technology consultancy Orbitwerks. ALYSIA, as the system is called, processes short lines of text and associates each syllable with a musical note. It chooses the pairing based on features including the syllable’s position in the word and, where relevant, how