After Suno became popular, China's AI music circle once again came into the spotlight.
Xuanxuan, who was the head of AI music products at a former leading company four years ago, observed that some capital and large companies had re-entered this field. Many AI music startups, including Aiyin, had once again received a new round of financing. He also heard the echo of what he had insisted on four years ago: AI music had finally entered the public eye.
But a hidden concern is emerging: Will AI lead to more electronic waste?
In late June, I met Li Li, CEO of Aiyin, in an office building near Shanghai IFC. Aiyin is one of the earliest AI music startups in China and received investment from a large company around 2020.
The logic behind the big companies’ investment in Aiyin was not that they believed AI music would have a disruptive effect on the industry, but rather out of a FOMO (fear of missing out) mentality, because Aiyin had received a TS (investment intention) from another competing big company at the time.
Looking back at the original intention of making AI music, Li Li, who was a musician, told Huxiu: “Ten years ago, the Chinese music scene began to decline visibly. “The assembly line mass production of divine songs and “electronic waste” flooded the public's vision, so they began to explore the possibility of using AI music to replace these “electronic waste” to achieve commercialization.
I asked Li Li: “What proportion of the music industry's revenue is made up of these electronic wastes?”
Li Li revealed an exclusive data: squeezed by these electronic wastes, Sony Music, one of the three traditional record companies in the golden age, now has only a few percent of the market share in Tencent Music. Five years ago, the three major record companies including Sony accounted for 70% to 80% of the market share in Tencent Music.
So, from a technical point of view, can AI music reach the level of electronic waste?
Li Li told me that their technology had reached this level in 2020. After providing AI music BGM technical services to a certain music platform, “our music not only passed the Turing test, but 10 million users used it and none of them criticized our AI. I really felt pretty good at the time.”
The implementation scenarios and technical means are already in place, so according to the normal story development, as one of the earliest startups to enter the AI music industry, it should have already disrupted part of the traditional music industry. However, the cramped office and the occasional confusion in his eyes told me: the story is not that simple.
Excessive electronic waste is not a good business
Li Li, who was originally an idealist, found after several years of exploration that this path was not feasible. There were many obstacles to the commercialization of AI music, one of the main reasons being determined by the economic structure of the music industry.
It was not until an industry report released by Tencent Music in 2020 that the chain logic of this complex industry truly surfaced.
The report shows that there are 1 million new songs, and this year the number may reach 10 million. Among them, 200,000 songs are of perfect quality, but only 10,000 have the opportunity to be heard by users, and 100 are listened to repeatedly by users. And those 100 songs earn 80% of the money of the entire industry.
This means that users’ music consumption demand is fixed, and AI music will not bring about a greater increase in demand; at the same time, the already over-produced content side does not need the more “electronic waste” brought by AI music.
Perhaps among tens of thousands of AI music, a few will be heard by the public, but most of them will be like passing clouds and sink into the electronic wasteland.
But who cares? After all, before Suno and Udio were launched, countless artificial music was uploaded to streaming media every day, but it never had a chance to be listened to by humans. AI evolves by devouring human creativity, but these inferior songs will not even become training materials for Suno and Udio.
In this industry where the head effect is extremely obvious, the emergence of AI may just be a reminder to creators: if you don’t reach a certain level, all your beautiful talents can be replicated and erased by AI.
“It was a bolt from the blue to me when I learned that millions of new songs were produced each year,” Li Li recalled. “A key question was: Can AI really bring more revenue by improving the efficiency of music creation?” This means “the supply side doesn’t need you, so what should your business model be like?”
Later, they began to change their route and launched new products, but I could still feel his confusion and uncertainty about the current products. But even so, with a prestigious school background and the support of top investment funds, he still received tens of millions of dollars in financing. After Suno came out, the snowball rolled bigger and bigger. Not long after our conversation ended, news came out that they were going to raise the next round of financing.
Who is holding back AI music?
Li Li, who was invested by a music giant's industrial fund, still retains a key secret that is holding back AI music.
The secret lies in this: in addition to the music industry structure itself, platform-based giants such as Tencent Music and NetEase Cloud Music, which control the content distribution rights, are not happy with AI music shaking up the current industry landscape and ushering in this new wave of Renaissance.
“Tencent and NetEase will definitely fight us to death.”
In front of Huxiu, Wang Wei, one of the earliest AI music entrepreneurs in China, told a standard story of an individual entrepreneur fighting against the monopoly of a large company.
His approach is to use AI music to allow ordinary users to generate their own musical expressions, and gradually build an AI music UGC platform. This AI music platform supports the distribution of user-generated AI music to short video platforms to earn distribution fees. “We will definitely not touch Tencent Music and NetEase Cloud Platform. They are not vegetarians. We will definitely have a tough battle with them, and we will definitely subvert them in the end.”
But I know very well in my heart that the ending of this kind of story often deviates from expectations. A practitioner close to Tencent Music told us that Tencent and NetEase will not give these AI music startups even a second's chance.
Therefore, after large companies entered the AI music market in 2020, some variables were brought to AI music. Ran Ran, CEO of another AI music startup, told Huxiu that after a large company failed to invest in them in 2020, “headhunters kept calling my employees to try to poach them."Soon after, another platform began to copy their products. But the good thing was that they also received investment from another large company.
During that one or two years, he kept a close eye on the data trends of the AI music product “similar to Suno” of the big company, “to see what the big company was doing and whether I should make adjustments.” But now he thinks that mentality was a pitfall. Because in the end, the AI music product of the big company “collapsed” and did not come out.
By the beginning of 2021, the AI music product team within the big company had disbanded.
I heard a more complete version from Xuanxuan, a former head of a major music company who witnessed the AI music melee among major companies four years ago.
“If you want to practice this skill, you must first castrate yourself. Even if you castrate yourself, you may not succeed.” In front of Huxiu, Xuanxuan described why the big company he worked for failed to do AI music four years ago.
“When large companies were working on innovative projects, they encountered many obstacles,” Xuanxuan recalled. “The complex interests within the companies and the immature technology at the time caused many obstacles to the advancement of the AI music business line.”
He then cited the example of Kodak cameras being killed by the digital camera that Kodak invented. “This is equivalent to giving up all the industrial chains, channels, and departments that you have accumulated over the years. But even if you give up, you may not succeed.”
The shadow of the giant not only affects the development of AI music startups, but also hangs over the AI music departments within these large companies. On the one hand, they are afraid of missing out on the AI music wave, and they bet on both sides through investment or internal project establishment; on the other hand, they dare not go all in completely, after all, this poses a certain threat to the existing music business of the large companies.
Xuanxuan revealed to Huxiu: “In 2020, we used AI music to provide technical support for a feature of our karaoke product – a one-click MIX accompaniment change feature. After the new feature was launched for a while, another large company felt a sense of crisis and urgently invested in an AI music startup company.” Then, he added: “This tangled and contradictory mentality is not unique to a certain large company, but is determined by their organizational form.”
Therefore, even though they have extensive industry resource accumulation and channel advantages, the AI music products of these large companies have never made much noise.
After Suno became popular, the story of big companies fighting over AI music four years ago was re-echoed. Although the business model is still to be explored, more big companies have chosen to get involved in this foggy sea.
The consensus among these practitioners is: “Big companies still have some FOMO mentality, and some are also for secondary market management.” Most of them still say: “First make the product, and then continue to iterate and explore new business routes in the subsequent versions.”
Ranran told me: “The AI music circle is very small. After all, the number of top students with a background in both music and technology can be counted on one hand.” Putting aside various conflicts and disagreements, what this group of people have in common is: their extreme love for music.
After graduating from a top 2 mathematics department, he, who loves music, rejected a high-paying job from a top VC, studied classical music at the Curtis Institute of Music for a few years, and then returned to China to start an AI music startup. But after starting his own business, he had almost no time to appreciate and create music.
He recalled his original intention of creating AI music – equal rights for music: he hoped that more people could easily create music at the level of entry-level musicians.
After being poached by large companies and copied by competitors, the original ideals seemed to be no more important than keeping the company alive.
The story of AI Music is perhaps a microcosm of many generative AI practitioners. With a passion for content, they hope to use AI to change the current situation of the industry being flooded with electronic waste, but they find that their efforts are like a drop in the bucket in the face of the original industrial structure, and commercialization is difficult to achieve. In the new round of AI bubble, such stories are still emerging, and Ran Ran is not the only one trapped in such a dilemma.