Will 2023 be the year that AI music and humans go head-to-head?

Author | Wan Yi Edit | Fan Zhihui

On his performance, renowned French DJ David Duetta believes that "the future of music lies in AI". 

Recently, he shared on Twitter a clip of his song during the show, and with several sound effects, Eminem's voice sounded, singing "This is a future rave song..." and the audience was "crazy about it." According to David Duetta, this is a song he finished with the help of AI tools. He first typed "write a song about Future Rave in Eminem style" on the lyrics generation website, and then put the lyrics in Eminem's voice on the virtual vocal website. 

The effect of AI music is certainly surprising, but some people soon questioned whether such a practice was allegedly infringing. It can be said that the alleged infringement is indeed the most worrying issue in the application of AI music. 

Since the explosion of ChatGPT, artificial intelligence no longer seems to be just a gimmick or a secretive high-end technology, but a future that ordinary people can reach. Especially in the past two years, music companies and streaming media platforms have accelerated the layout of AI technology, high-end game-breakers such as Music LM and Jukebox are still rubbing their hands, and ordinary AI composition platforms have received commercial cooperation from big names such as Audi. 

So, will 2023 be the year that AI music and human music go head-to-head? 

Does AI concert trigger the "Napster effect"?

The gradual development of AI has brought real fear to the music industry in the first place. 

It is true that AI music represents a weapon to expand the boundaries of creation and improve creative efficiency, but it also threatens the careers of a group of musicians. The most worrying thing is that the record companies, AI is based on the music that already exists, but does not have to pay a penny to the record company, the music created can even steal some of the jobs of human creators, record companies naturally will not easily allow this kind of thing to happen. 

In January this year, the RIAA (Recording Industry Association of America), which represents the interests of the three major record companies, submitted a report to the US government, lobbying the US government to pay attention to the issue of AI music infringement. Since last year, the RIAA has listed software such as RemoveVocals and Songmastr, accusing them of generating new music from existing music as an infringement on musicians and record labels.

The British Recording Industry Association (BPI) has made the same sound. Because last year, in order to promote the development of artificial intelligence, the British government wanted to allow AI training to use creative works, including music, without the permission of creators and rights holders, which was strongly opposed by several music industry organizations in the British music industry, including the Recording Industry Association (BPI), so the government finally decided to re-examine this proposal. 

From January to September last year, the RIAA appeared in 15 government lobbying reports, declaring nearly $5.4 million in lobbying expenses, the highest lobbying expenditure since 2018, which shows how seriously record companies take it.

Recently, this concern has continued unabated. In February, Universal Music said that AI song generation could cause lasting and widespread harm to musicians, calling for not relaxing copyright laws to promote AI development, and could create a Napster-level crisis for copyright protection. 

Back in 1999, the global recession was unfolding and the old world was on the verge of collapse with the advent of new technologies. In the chaos, the record industry is enjoying unprecedented light because of the huge profits of the CD business, and in the dormitory of Northeastern University in Boston, a precocious programmer has choked the music industry by the throat and almost strangled it. 

The programmer, called Shawn Fanning, launched Napster, a music sharing platform focused on P2P, that allows young people to share each other's mp3 files over the Internet. Just 4 months after launching, Napster attracted more than 150,000 users, 8 months later it exceeded 26.4 million, and some even suspected that it exceeded 80 million. Sixteen months later, its founder was on the cover of Time magazine. 

But Napster also became a nightmare for record labels. Because of Napster, pirated music is rampant, people can easily obtain music resources through the Internet, sales of tapes, CDs and vinyl records have plummeted, the scale of music industry revenue has not exceeded the peak of 1999 in the next 20 years, and physical music revenue has disappeared, almost every year can only be lower than last year. In 2019, the RIAA angrily sued Napster, and artists represented by Dr Dre and record companies represented by EMI expressed strong protests against it. 

Finally, Napster was shut down in 2001 after multiple lawsuits with record labels and artists, and Blistmann, who provided loans behind Napster, paid tens to hundreds of millions to hundreds of millions of dollars to Universal Music, Warner Music, EMI, and the National Music Publishers Association (NMPA) from 2006 to 2007, and the rising star Spotify also became the strongest worker of copyright owners through agreements with copyright owners. 

Today's AI music is indeed similar to Napster.

AI tools trained with the help of a wide range of existing content to generate content according to the user's requirements, like Napster in the past, represents a newer trend, more open and efficient knowledge use technology, but at the same time, AI tools are also accused of "stealing" the fruits of creators. 

If AI music is really widely used, it is indeed likely to cause unprecedented damage to music copyright content. Of course, the new generation of AI game-breakers certainly don't want to be the Napster that breaks the game to bankruptcy, nor does it want to become Spotify who has revolutionized but still worked for others, and the key depends on how AI music companies will agree with copyright owners. 

But in the face of yet another technological innovation, record labels have learned a valuable lesson from Napster. 

The music companies that strongly refused to stream have fallen under the wheels of the times, and the surviving winners are old drivers who know how to ride on the wheel to make themselves faster. This time, music companies, while worried, have scrambled to join the wave of AI music in a way that suits them. 

Music companies embrace AI music

Record companies represented by Universal Music have asked the RIAA to lobby the government to worry about the infringement of AI music; On the one hand, he also became a player, for fear of falling behind. 

Today, with the help of many people in the music industry, AI generation tools have covered many aspects such as lyrics, composition, voice generation, arrangement, mixing, and even distribution.

In terms of the composition that has attracted the most attention, music created by AI has been widely used in functional music such as games, commercials, and ambient music. 

This is a lifesaver for streamers who want to reduce the cost of copyright. But abroad, perhaps due to infringement concerns, streaming platforms have not been high-profile in the acquisition of AI composition tools, and Spotify's rival Anghami in the Middle East and North Africa has only claimed that it will soon become the world's first platform to host more than 200,000 AI songs. 

YouTube's "brother of a mother" Google's recently launched music creation killer Music LM is still "sealed" by Google due to copyright-related concerns. It has been likened to the musical version of ChatGPT, capable of composing songs based on text entered by the user. However, Vice gave the evaluation that it "does a good job of copying the music of the game, but so far, it seems that it can't write songs that can bankrupt your favorite artists." OpenAI's Jukebox and MuseNet are in a similar situation. 

In China, there are more entrants to this track. ByteDance acquires AI composition company Jukedeck; Based on Huawei's original AI Singer model, "Singing Duck" allows users to complete the creation by simply entering lyrics. NetEase also launched NetEase Tianyin, and launched the "NetEase Tianyin" New Year Greeting Mini Program, allowing users to generate New Year greeting songs by entering lyrics and sing them with AI singers; Kuaishou also independently researched AI music creation models and AI singers, and launched "Komori Singing" to create exclusive songs for users through AI. 

Interestingly, the record labels are also positive about this.

AI music company Endel, which focuses on music that provides focus, relaxation and sleep scenes, has received investment from music companies such as Warner Music, AVEX, Sony, Concord Music, and Royalty Exchange, and currently cooperates with Mercedes-Benz Research, Japan's All Nippon Airways, and Australian consumer electronics company Nura; Warner Music also helped release the album, cooperated with Sony to study headphones, and co-created sleep music called "AI Lullaby" with musician Grimes, which made headlines at the time. 

In 2022, Warner Music also invested in AI music company Lifescore, whose application scenarios include live broadcasting, cars, sleep, fitness, games and metaverse, and there are many big brands such as Audi. 

Perhaps it is the copyright issue that has not been solved, making AI composition software at risk of infringement, making music companies more inclined to let AI recommend the development of music creation tools.

For example, Sony CSL (Computer Science Research Laboratory) launched the AI-assisted music creation application Flow Machines in 2020, and in 2022 also launched a machine learning model called "SampleMatch", which can retrieve instrument samples that match specific tracks from archives to help create. 

Universal Music, Disney, Microsoft and others have also invested in Soundful, an AI-powered music creation platform. Perhaps due to the support of all the "copyright maniacs", Soundful has solved all copyright licensing issues in the closed beta stage, so that users can use music with video, advertising, social media and other platforms with confidence. 

Warner Music participated in the seed round of investment in the AI arranging software Landr, which can provide remote teamwork, automated mastering, and distribution to more than 150 streaming platforms to help musicians complete the "last mile" of creation. Spotify's Basic Pitch can help musicians transcribe music efficiently. 

The new software SingSong recently launched by Google allows users to generate the corresponding accompaniment only by humming, which simply allows music idiots to complete their dreams of singers. 

In China, NetEase Cloud Music has made a strategic investment in AIVA, an AI music creation assistance platform, whose music is inspired by music masters such as "Mozart, Beethoven, Bach" and other music masters, and is also a public version of music, including NVIDIA, TED, Luxembourg Ministry of Culture and so on. Sponge Band, a music creation tool owned by ByteDance, has also introduced AI technology, similar to the "clipping" of music production, which can complete various simple and efficient creations and optimizations for creators. 

In addition, AI voice generation has also become an important direction for music companies, not only to improve the efficiency of music creation, but also to create virtual idols to welcome the meta-universe. 

South Korean idol entertainment giant HYBE acquired AI voice company Supertone in 2022, which is speculated to be used for idols' digital avatars to provide fans with more customized content. That same year, Spotify acquired Sonantic, an artificial intelligence voice platform. This year, its speech generation technology, combined with OpenAI and the platform's human editors, launched AI DJ, which provides users with personalized song services and can instantly adjust recommended content based on user feedback, which is currently only available to paying users in the United States and Canada. 

In China, iFLYTEK launched the "iFLYTEK Voice" label based on its own AI technology, involving singing synthesis, singing voice change, singing evaluation and other technologies. Tencent AI Lab launched the digital human "Ailing", which can automatically generate lyrics and sing based on keywords provided by users. 

Kugou Voice Apollo Sound Lab has developed the sound synthesis black technology "Lingyin Engine", which can highly restore and reproduce the voice characteristics of singers, not only To B, so that singers no longer break sounds, but also provide fans with personalized idol companionship, but also To C, becoming the "Meitu App" for ordinary people in sound. 

In addition to generative AI, another major use of AI is data analysis. 

In the streaming era, some of A&R's work has been replaced by data, and music companies are trying to find the next Billie Eilish or BTS from the fountain of data. 

As early as 2018, Warner Music acquired AI-powered A&R platform Sodatone, which uses machine learning to mine songs uploaded daily to platforms like Spotify and analyzes social media and tour data to find potential newcomers. In 2020, there were twice as many artists and songwriters locked and signed through Sodatone than in 2019, which shows that Warner Music has a green eye for Sodatone. 

Another AI-powered A&R platform that Warner Music has invested in, Instrumental, has a stake in Tencent, and the three major records and Live Nation are partners. SoundCloud also acquired Musiio to detect the condition and A&R of songs on the platform, hoping to build relationships with rising stars early and retain them. 

In addition to finding artists, music companies are also using AI to find listeners. 

This does not refer to streaming recommendation algorithms, or streaming platforms using AI to label or censor music, which has become quite popular. Even Apple's acquisition of AI Music allows music recommendations to be more accurate based on data such as heartbeats to predict scene changes, or Spotify's new AI DJ function, which allows users' feedback to instantly change the recommended content, can only be regarded as the icing on the cake. 

What is more noteworthy is that record companies have begun to change the way artists attract fans and solid fans through AI in the past two years. Ingrooves Music Group, a global label, has obtained three patents in AI marketing in the past two years, which is committed to using AI to automate marketing campaigns, identify high-value audiences, and predict high-value social media audience behavior. 

All in all, from songwriting, A&R, copywriting, marketing, library management and other aspects of the music industry, AI may eliminate some jobs, or AI can become a powerful assistant to help people get a little relief from complicated work. 

epilogue

By reviewing the layout of music companies in AI, it can be seen that the role of AI has penetrated into all aspects of the music industry. 

As for the most revolutionary AI music creation, because the copyright issue has not been properly resolved, AI music creation is more like a Pandora's box that has not yet been allowed to open, although people can't wait to open it, but the result after opening can be fatal. 

The record companies will never easily allow it to become the second Napster and change the life of the music industry again. However, music companies that are opposed to it and invested heavily have long understood that rejection is definitely not the right gesture in the face of changes in the general trend, and it is the law of winners to go with the flow and find a way to use it for me. In the face of the enormous benefits represented by AI creation, it seems that accepting it is only a matter of time and how. 

So, stop asking if AI music is coming, because it's already coming.