Developments in the field of artificial intelligence (AI) have led to the advent of a new, real-time voice clone called Sesame. Despite growing scrutiny and concerns regarding the ethical implications of voice cloning, AI companies continue advancing the technology with immense potential.
Sesame AI, a pioneer in this arena, is said to be capable of replicating a person’s voice by using a mere three-second sample. Although, the precision and realism are said to increase significantly when the sample provided is at least five minutes. The technology elongates a short voice clip up to four hours of continuous speech.
This technology has been developed based on an innovative machine learning technique known as GANs (Generative Adversarial Networks). GANs enable an AI to learn from the complexities of human speech, thereby capturing and mimicking the nuanced inflections, cadence, and tone.
While the potential applications of the technology are abundant, it also brings a myriad of ethical questions to the forefront. Concerns arise as this technology could be exploited to create ‘deepfakes’, counterfeit voices used in fraud or misleading news narratives.
To counteract potential misuse, Sesame AI has implemented a policy where approval is needed from the individual whose voice is to be cloned. Applications are then reviewed to determine eligibility. Sesame further notes that it keeps a recorded digital fingerprint of each voice to track potential misuse.
Undoubtedly, the promises of individualized AI voice technologies are vast. They can create personalized characters in video games or provide assistance to those who have lost their ability to speak. However, the potential misuse of the technology warrants rigorous controls and policies to safeguard against any unethical applications.
Despite these challenges, Sesame AI remains optimistic about the future of real voice AI, emphasizing that the safeguards in place are currently sufficient to deter potential misuse. The company plans to continually enhance and refine their regulations as the technology advances.
Last modified: March 8, 2025
