TL: DR — Recently AR Rahman used AI-generated voices of deceased artists for one of his albums. A R Rahman may have done his best to keep his consciousness clean and walked a thin line on the ethical aspects. But this falls under the gray area legally and also opens up more concerns. Read further for an analysis of such concerns and to learn more about what to expect in this space.
[Credit: ChatGPT, prompted to generate an image to visualize the content of this post]
AI-Generated Music/ Voice
The advent of Artificial Intelligence (AI) in music has marked a significant shift in how we create, experience, and interact with music. AI’s foray into the music industry has been transformative, offering tools to generate compositions, manipulate sound, and even recreate the voices of artists. This shift is not just about innovation but also about how it challenges our perceptions of creativity, authorship, and legacy in the digital age.
Technological advancements in music have evolved from the rudimentary use of electronic instruments and synthesizers to sophisticated AI algorithms capable of analyzing vast music datasets to create new compositions. This progression has enabled the development of AI systems that can not only compose music in several styles but also perform and produce it with minimal human intervention. These systems analyze patterns in music to understand rhythm, melody, harmony, and structure, applying this knowledge to generate new works that are both unique and stylistically consistent with existing genres.
Recent applications of AI in music have been particularly striking in the realm of voice synthesis. Technologies now enable the recreation of the voices of artists, allowing them to ‘perform’ new songs or be featured in virtual concerts. This has been made possible through developments in machine learning and digital sampling, where an artist’s voice recordings are analyzed to capture their unique vocal characteristics. These characteristics are then used to synthesize new vocal tracks that can sing new lyrics with astonishing realism. Projects like the holographic performances of iconic artists, VR headsets, and the use of AI to produce posthumous albums are testaments to the capabilities and creative potential of AI in music.
The technical foundation of voice synthesis involves collecting extensive samples of an artist’s voice, which are then analyzed by deep learning algorithms to understand the nuances of their vocal timbre, pitch, and dynamics. This analysis allows the AI to replicate the voice, applying it to new performances or compositions. The process, while complex, opens up new realms of possibility for preserving and celebrating the legacies of artists who are no longer with us. It also raises questions about authenticity, copyright, and the ethics of posthumous digital creation, reflecting the broader impacts of AI on creative industries.
A R Rahman and Technology
A R Rahman’s journey with technology in music is a testament to his innovative spirit and willingness to push the boundaries of what’s possible in music production — he has always been at the forefront of leveraging technology in music, be it synthesizers, computers, virtual reality, or now, AI.
From his early days, Rahman embraced synthesizers and digital audio workstations, incorporating them into his compositions to create sounds that were fresh and new to Indian cinema. He has been a pioneer in using surround sound in films, enhancing the cinematic experience by immersing audiences in rich, multi-dimensional audio landscapes. I vividly recollect him using a wearable device by Intel (which would later become Intel Curie) during his concert at Marymoor Park, Redmond many years ago to generate music without any instruments. (Watch his troupe use this technology at CES 2016 here to generate music of thin air.)
Rahman’s exploration didn’t stop with traditional music production; he ventured into virtual reality to provide the audiences with a cinematic sensory experience. His use of AI to generate human voices in his albums was only a matter of time, which he did for the song Thimiri Yezhuda in the recently released movie Lal Salaam. Rahman leveraged AI to generate the voices of deceased artists Sahul Hameed and Bamba Bakya, breathing new life into their legacies. He also sought permission from the family members of these artists and compensated them, paving the way for new streams of income from the legacy of the artists.
Original voice tracks of the artists were used to train AI voice models and generate equivalent using the Timeless Voices platform. Though the AI-generated voices were only about 70% match (this is the link to his BBC interview where he shares more details on the number of iterations, inspiration, and approach, check it out) to the original voices of the artists, Rahman ensured that the project paid homage to the artists in a manner that was both respectful and groundbreaking.
The response has been mixed though with some fans not even able to detect the original artist’s voice. But the intended artistic impact of this project was profound, offering listeners a unique connection to the music and its original artists, and setting a new precedent for the use of technology in creative expression.
A R Rahman recommends using AI as a slave and not letting it make decisions. However, leveraging AI for making decisions is not new, and it is just inevitable that AI will outgrow being a ‘slave’.
Questions
Thimiri Ezhuda is the first such attempt in the Indian Cinema industry where a deceased artist’s voice has been brought back to life for a complete single. This is in itself to be appreciated and welcomed, and we can expect to see more such attempts including complete videos/ movies in the future. Paul McCartney earlier had used AI to ‘purify’ John Lenon’s voice recording to create ‘what will be the last Beatles record’.
However, this also raises the following questions, along with my thoughts on these questions:
Is it ethical?
The ethical landscape of using AI to recreate the digital likenesses of deceased artists is complex, touching more than consent and ownership issues that demand careful navigation. Permissions from the artists’ estates are crucial, as they hold the rights to the artists’ works and likenesses. Without clear consent, the digital recreation of an artist could lead to legal battles and ethical dilemmas, particularly regarding the implications for the artists’ legacies and the integrity of their posthumous work.
A R Rahman obtained permission from the family members of the deceased artists, this represents a thoughtful approach to ethical considerations, showing respect for the artists’ legacies and their families’ wishes. A R Rahman claimed his consciousness was clean though.
Since A R Rahman obtained permission from the family members of the deceased artists, this represents a thoughtful approach to ethical considerations, showing respect for the artists’ legacies and their families’ wishes. However, the legal authorization of family members to provide such consent depends on the copyright laws, their right to publicity, the artists’ wills, and any existing contracts (more on the legal aspects in the next sub-section).
Is it Legal?
Copyright law and rights of publicity provide some protection for artists’ works and likenesses, but the rapid advancement of technology, particularly the creation of AI generation of voice, images, and videos, presents new challenges. These gaps highlight the need for legal frameworks to evolve in step with technological innovations, ensuring that artists’ rights are protected after their death while fostering creativity.
Legally, it is crucial to ensure that the individuals providing consent have the authority to do so on behalf of the artist’s estate. Currently, there is no dedicated US, European, or Indian law that forbids AI generation of voices of a deceased person [standard legal disclaimer applies here, please contact your attorney for more details]. However, there are laws around the rights of publicity, who owns this right, and for how long, which help determine if such AI generation is legal or not. Within the US, these vary between states, which determine the legality of such AI generation.
Unless the deceased has willed otherwise, the heirs have a right of publicity post-mortem as allowed by the state of residence of the artist when they were alive. Some actors like Robin Williams were proactive in this regard; some are following suit as this space evolves [for example, popular Bollywood actor Anil Kapoor has ensured legal protection of his personality rights (and that of his family after his demise) including his name, image, likeness, voice, and other attributes of his personality against any misuse].
In this specific case of AR Rahman generating voices of Shahul Hameed and Bamba Bakya, the legality boils down to whether the family had the right to publicity of their voices or not.
Is it moral? Can a dead person be brought back ‘alive’?
AI generation of the voices, images, and videos brings up the possibility of bringing back the dead alive, at least in virtual environments. Along with recent developments in VR technology, it is only a matter of time before we can interact with Avatars of our loved ones who are no longer with us physically. While human cloning was outlawed by nations when Dolly was cloned using Somatic Cell Nuclear Transfer (SCNT) technology in 1996, such AI generation might be received more positively. I do not expect to see laws that outlaw such AI creations; however, we can expect to see a lot of pushbacks from various quarters.
Such AI generation also opens up the possibility of works of popular artists being used instead of providing opportunities to upcoming/ other artists. While this is very much possible, this is market-driven and is similar to popular artists who are alive and in demand.
Is it only about royalties/ copyright/ consent?
There are more aspects than ethical, legal, and commercial aspects to this — legacy preservation, emotional connection, and societal impact to name a few.
Fans connect emotionally with artists. Any recreation of an artist’s work is bound to bring up such emotions. Creators must respect such connections while leveraging AI for such recreation. They also need to ensure the authenticity of such works when compared to the works of the original artist. Authenticity is also important in enforcing copyright laws.
Emotional and authenticity concerns also play a significant role. AI recreations can deeply impact audiences, offering a semblance of the artist’s presence but potentially distorting their legacy if not handled with sensitivity and respect. The authenticity of such recreations is a contentious issue, as it raises questions about the artistic integrity of new works produced under the artist’s name without their direct input.
It is also important to note that AI-generated work does not affect the legacy of the deceased artist. What if such work is used against the artist’s principles and values, or negatively portrays the artist? Countries such as India have laws around defamation and libel that may offer protection against such instances, but such concerns must be addressed, and the legal system may need to catch up.
Such re-creations have the potential to have a huge societal impact — both positive and negative! Imagine Mahatma Gandhi trying to broker peace in Gaza, MLK speaking for the BLM movement, or President Truman addressing if he would defend using atomic bombs! [ I am purposefully leaving the possibility that may create a negative impact here]
We also witness AI-generated videos of yesteryear political figures being used for promotional purposes already. With elections around the corner, in India, we can expect to see such AI-generated figures actively campaigning for their parties as well.
What to expect?
Future regulations must address these challenges, establishing clear guidelines on consent, attribution, and preventing misuse. This would involve crafting legislation that respects the rights of artists and their estates, safeguards the emotional well-being of audiences, and maintains the authenticity of the artists’ legacies.
We are seeing disjoint efforts happening, such as the EU’s proposal on new copyright rules for generative AI, President Biden’s Executive Order on Safe, Secure, and Trustworthy Development/ Use of AI, etc. However, more efforts are needed to prevent misuse. In countries such as India, one can expect to see stricter laws/ protection against recreating voices/ images/ videos of popular leaders/ celebrities/religious figures.
Conclusion
A R Rahman’s use of AI to generate voices of deceased artists for music albums walks a fine line ethically but is not clean legally. He has done the best to his consciousness though.
The importance of balancing technological advancement with ethical considerations in the arts cannot be overstated. By setting a framework that balances innovation with ethical considerations, the industry can navigate the complexities of AI in music, ensuring that technological advancements enhance artistic expression without compromising integrity or rights.
Artists like A R Rahman, who navigate these complexities, demonstrate how innovation can coexist with respect for legacy and ethical standards. There is a critical need for ongoing dialogue among technologists, artists, legal experts, and policymakers to ensure that creativity flourishes within ethical boundaries. Honoring the legacy of artists through innovation, while advocating for a thoughtful approach to technology, ensures that advancements enrich the arts without compromising the values we hold dear.