AI-generated celebrity voice mimicry and protection measures discussed by Hume's CEO in the realm of advanced artificial voice creation technology
In a recent demonstration, Dr. Alan Cowen, CEO and chief scientist of Hume, unveiled the latest update to their AI voice creation service, EVI 3. The technology, which can create a near-perfect replica of someone's voice from just 30 seconds of audio, has the potential to revolutionise various industries. However, it also raises significant ethical concerns.
EVI 3 uses large amounts of voice data and reinforcement learning to learn inflections, accent, and personality from a small clip. This technology offers numerous legitimate uses, such as enhancing creative workflows in media production, improving accessibility for individuals with speech impairments, and reducing costs by eliminating the need for repeated recording sessions.
One of the most exciting applications of EVI 3 is its potential to expand multilingual reach with the same voice profile in content creation. For instance, podcast production can benefit from quick voiceover generation, maintaining voice consistency across episodes, and providing a multilingual option for listeners.
However, the combination of AI voice cloning, video, and image generation could potentially escalate the problem of deepfakes. Dr. Cowen expressed concern about the potential misuse of voice cloning technology, such as creating deceptive calls, speeches, or fake audio leading to fraud and misinformation campaigns.
Consent is another major concern. Using someone's voice without explicit permission risks infringing on personal rights and can be legally actionable. The technology also raises concerns about privacy, as voice data can reveal sensitive personal information beyond speech. For example, emotions and health can potentially be inferred from voice samples.
Reputational harm is another issue. Voice cloning without provenance can lead to misuse in offensive or controversial contexts, damaging the original voice owner's reputation. The technology also threatens voice actors' livelihoods by replicating their unique vocal identities without proper credit or compensation.
To address these concerns, Hume has established ethical guidelines for empathetic technologies, known as the Hume Initiative. The initiative lists six principles: benefits should outweigh costs, technology should serve emotional well-being, claims should be supported by rigorous science, diverse groups should have equal access, people should have access to necessary information, and informed consent is required.
Dr. Cowen emphasised the importance of these principles and Hume's commitment to the ethical side of the debate. He believes that voice cloning technology is a technology that Hume had to build, given its potential benefits. However, he also highlighted legitimate use cases such as live translation, dubbing, accessibility, and celebrity interaction.
Despite these efforts, Cowen acknowledged that the AI space moves quickly, and bad actors may have access to voice cloning technology in the future. He urged continued efforts to update legal protections, establish clear consent protocols, ensure transparency, and balance innovation with ethical responsibility to mitigate risks of abuse and harm.
This debate is not unique to Hume. Sam Altman, CEO of OpenAI, has also warned about the risks of AI voice cloning and its potential use in scams and bank voice activations. As AI voice cloning technology becomes more prevalent, it is crucial that all parties involved work together to ensure its safe and ethical use.
Technology like EVI 3, capable of creating near-perfect replicas of someone's voice from minimal audio, has enormous potential to transform various industries, such as media production and accessibility solutions. However, the ethical concerns surrounding its misuse, particularly in regards to deepfakes, consent, privacy, and reputational harm, necessitate careful consideration and the establishment of guidelines to promote safe and ethical usage.
