How artificial intelligence is changing music


SecondEarlier this year, Bad Bunny flatly denied rumors that he was about to release a new song with Justin Bieber. “That’s false,” he told Time magazine in an interview for a cover story about his meteoric rise. “You never know what I’m going to do.”

But last month, a song that sounded like his and Bieber’s voices began circulating on TikTok, racking up millions of likes. Bad Bunny didn’t lie in the interview, though: the song was created using artificial intelligence. An artist named FlowGPT used artificial intelligence technology to recreate the sounds of Bad Bunny, Bieber and Daddy Yankee in reggae songs. Bad Bunny himself hated the song, calling it a “shit song” in Spanish and discouraging his fans from listening to it, and the clip has since been removed from TikTok. But many fans of all three stars loved it equally.

The song and the polarizing reactions to it are emblematic of the worrying ways in which artificial intelligence is sweeping the music industry. Over the past few years, advances in machine learning have made it possible for anyone sitting at home to recreate the sounds of their musical idols. An artist named Ghostwriter became popular for his impersonations of Drake and The Weeknd. Another creator jokingly set Frank Sinatra’s smoky vocals to profane Lil Jon lyrics. Other AI tools that allow users to conjure up songs simply by typing in prompts are essentially audio versions of text-to-image tools like DALL-E.

More from Time Magazine

Some proponents believe these advances will further democratize music, allowing anyone with an idea to create music in their bedroom. But some artists are outraged that something as personal as their sound or musical style could be exploited and commodified for the benefit of others. In the coming years, we will explore the push and pull between protecting artists, driving innovation, and identifying the complementary roles of humans and machines in music creation.

“If a huge explosion of music were created at infinite scale and infinite speed, would we rethink what we as humans actually bring to the table?” Lex Dromgoole, musician and artificial intelligence technologist asked. “Where’s the imagination in this? How do we inject personality into our creations?


Music producers are already using artificial intelligence for more mundane parts of their jobs. Artificial intelligence can help correct pitch and allow engineers to mix and master recordings faster and cheaper. The Beatles recently used artificial intelligence to isolate John Lennon’s voice from a 1978 demo, stripping out other instruments and ambient noise to create a new, pristinely produced song. Artificial intelligence is also ingrained in many people’s listening experience: Streaming platforms such as Spotify and Apple Music rely on artificial intelligence algorithms to recommend songs based on people’s listening habits.

read more: The Complete Guide to Spotify Wrapped, 2023

Then there’s the use of artificial intelligence to actually create music, which sparks both excitement and alarm. Musicians have begun to embrace music tools like BandLab, which suggests unique musical loops based on prompts, as an escape valve for writer’s block. The artificial intelligence app Endel generates customized, ever-changing audio tracks for concentration, relaxation or sleep based on people’s preferences and biometric data. Other AI tools build complete recordings based on text prompts. A new YouTube tool powered by Lyria, GoogleDeepMind’s large-scale language model, allows users to enter something like “a ballad about opposites attracting, cheerful sounds” and it will instantly generate a clip by Charlie Puth. ) sung excerpts of songs that sound very similar to.

These technologies raise a variety of concerns. If artificial intelligence could instantly compose a “Charlie Puth song,” what would that mean for Charlie Puth himself, or for all the other aspiring musicians who worry about being replaced? Should AI companies be allowed to train large language models on songs without the creators’ permission? AI is already being used to summon the voices of the dead: a new Edith Piaf biopic, for example, will include a recombined version of her voice created by AI. If any voice from history could be reactivated, how would our understanding of memory and heritage change?

Even those most excited about the technology are becoming worried. Last month, Edward Newton-Rex, vice president of audio at artificial intelligence company Stability AI, resigned from the company, saying he feared he might put musicians out of work. “Billion-dollar companies are training generative AI models on creators’ works without permission and then using them to create new content, in many cases,” he wrote in an open letter. Can compete with original works.

These issues will likely be decided by the courts in the coming years. In October, Universal Music Group and other major labels sued startup Anthropic after its artificial intelligence model Claude 2 began spitting out copyrighted lyrics verbatim. A Sony Music executive told Congress that the company has issued nearly 10,000 takedown requests for unauthorized deepfakes of sounds. Many artists want to quit entirely: Dolly Parton recently called AI voice cloning “the mark of the beast.” In contrast, AI companies argue that their use of copyrighted songs is “fair use,” more akin to tributes, parodies, or covers.

Singer-songwriter Holly Herndon is one of the artists trying to stay ahead of these seismic changes. In 2021, she created a deepfake voice called Holly+ that allows anyone to transform their own voice into hers. She said the goal of the program is not to force other artists to give up their voices, but to encourage them to take an active role in these larger conversations and gain autonomy in the top-down music industry. In this industry, technology giants are playing an increasingly important role. “I think this is a huge opportunity to rethink the role of the artist,” she told Time magazine. “There’s a way to maintain some agency over a digital version of yourself, but be more fun and less punishing.”

Dromgoole, a musician and co-founder of the AI ​​company Bronze, hopes that AI music can get rid of the current stage of imitating the singer’s voice and immediately generating music. Over the past few years, Bronze has collaborated with musicians like Disclosure and Jai Paul to create ever-evolving AI versions of their music that never sound the same two times played. Our goal is not to use artificial intelligence to create perfect, monetizable static songs, but to use it to challenge the way we think about music. “The tech industry seems to think that everyone wants a shortcut or a creative solution,” he said. “That’s not how imagination works. Anyone who has studied flow states or spent time with music makers knows that we enjoy this process.





Source link

Leave a Reply

Your email address will not be published. Required fields are marked *