Artificial intelligence (AI) has been making significant advancements in various fields, and one area where it is becoming particularly impressive is in evaluating the similarity between voices. OpenAI’s “Sky” is an AI model that has gained attention for its vocal capabilities and has been compared to the voice of Hollywood actress Scarlett Johansson. To further explore this claim, forensic voice experts at Arizona State University conducted a study to determine how similar Sky’s voice and speech patterns are to Johansson’s.
In order to measure the similarity, the researchers compared Sky’s voice, based on audio from OpenAI’s demos, with the voices of around 600 professional actresses. The results of the study revealed that Johansson’s voice is more similar to Sky than 98% of the other actresses. This finding was supported by statistical evidence, providing credibility to what many listeners had already believed.
However, the study also revealed that Sky’s voice was reminiscent of other Hollywood stars, such as Anne Hathaway and Keri Russell. In fact, the analysis of Sky often rated Hathaway and Russell as even more similar to the AI than Johansson. This suggests that while Johansson’s voice may be highly similar to Sky, it is not the only voice that bears resemblance to the AI model.
Visar Berisha, a computer scientist at Arizona State University, who led the voice analysis in the College of Health Solutions and the College of Engineering, explained that the study demonstrates the undeniable commonalities between Sky and Johansson’s voices. While they are similar, statistical evidence indicates that they are not identical. Berisha suggests that this analysis provides valuable insights into the debate surrounding the similarity of voices and highlights the nuances that exist in AI-generated voices.
OpenAI, the organization behind the development of Sky, has maintained that the AI model was not created with Johansson in mind and was never intended to mimic her famous voice. In a conference, Altman, a representative from OpenAI, expressed regret for any confusion caused by the comparison to Johansson. He emphasized that whether one voice is truly similar to another will always be a subject of debate, highlighting the subjective nature of evaluating vocal similarities.
As AI models continue to advance, it is important to consider the ethical implications and potential challenges that arise. One concern is the creation of virtual personas that closely mimic real individuals without their consent. The development of AI models like Sky raises questions about the boundaries and rights of famous personalities. Should AI be allowed to generate voices that are nearly identical to those of celebrities?
Additionally, the study conducted by Arizona State University highlights the need for further research in the field of forensic voice analysis. By comparing AI-generated voices to the voices of professional actresses, experts can gain a better understanding of the nuances and similarities that exist. This knowledge can be applied in various domains, including criminal investigations and voice authentication systems.
Furthermore, the study also suggests that AI models have the potential to stimulate creativity and generate unique voices that resemble real individuals. While some may argue that this blurs the line between genuine human talent and AI-generated imitations, others see it as a testament to the capabilities of AI technology. By analyzing the similarities and differences between AI-generated voices and real voices, researchers can gain insights into the underlying mechanics of voice production and perception.
It is worth noting that the technology behind AI-generated voices is continuously evolving. As more data and advancements are made in the field of AI, it is possible that the voice similarity between AI models and real individuals will become even more indistinguishable. This poses challenges for voice recognition systems and raises concerns about potential misuse or unauthorized impersonation.
In conclusion, the study conducted by Arizona State University provides valuable insights into the similarity between OpenAI’s “Sky” and Scarlett Johansson’s voice. While statistical evidence supports the claim that Johansson’s voice is highly similar to Sky, other Hollywood stars like Anne Hathaway and Keri Russell also exhibit similarities. This study highlights the subjective nature of evaluating vocal similarities and raises important ethical considerations regarding the use of AI-generated voices. As AI technology continues to advance, it is crucial to carefully navigate the boundaries and implications of using AI to mimic the voices of real individuals. Additionally, further research in forensic voice analysis can contribute to advancements in various fields, such as criminal investigations and voice authentication systems.
Source link
