OpenAI has announced that it has developed a tool that can replicate the human voice from just 15 seconds of recorded audio, but it has not yet released it to the public due to concerns that it could be misused, especially during the 2024 election. .
The software, called Voice Engine, was first developed in 2022 and integrated into ChatGPT’s text-to-speech functionality.
However, starting in late 2023, OpenAI “began testing privately with a small group of trusted partners,” the artificial intelligence giant said in a previously reported blog post. Guardian.
The company said it was “impressed” with Voice Engine’s applications, including providing reading assistance to non-readers, serving as an educational tool for children, and translating content.
In one of the most striking use cases, researchers at the Norman Prince Neuroscience Institute in Rhode Island created a low-quality 15-second clip of a young woman giving a school presentation who had lost her voice to a vascular brain tumor. to “restore” her speech.
However, OpenAI has not yet released Voice Engine to the public. That’s because, according to the blog post, there are “significant risks that are particularly concerning during an election year.”
OpenAI says, “We are choosing to preview this technology at this time, but not broadly release it. This will strengthen society’s resilience to the challenges posed by more compelling generative models.” This is for the purpose of doing so.”
It wasn’t immediately clear when OpenAI would debut Voice Engine at scale, but it said it would be in the near future. “We encourage measures such as phasing out voice-based authentication as a security measure for accessing bank accounts and other sensitive information.”
“We want to start a dialogue about the responsible adoption of synthetic speech and how society can adapt to these new capabilities,” OpenAI added. “Based on the results of these conversations and small-scale testing, we will make more informed decisions about whether and how to deploy this technology at scale.”
Already, the use of AI has fueled misinformation, including a viral deepfake image of President Donald Trump and his wife, Melania Trump, yelling at police and resisting arrest.
As a result, Google updated its policies last year to require “all verified election advertisers” to “clearly disclose” when AI is used in their ads. However, OpenAI did not follow suit.
However, the San Francisco-based startup said in a blog post that its usage policy prohibits partners with exclusive access to Voice Engine from impersonating other individuals or organizations without their consent or legal rights. certify that he agrees with the.
“We are working with U.S. and international partners in government, media, entertainment, education, and civil society to ensure we incorporate their feedback during development,” OpenAI asserted.
Representatives for OpenAI did not immediately respond to The Post’s request for comment.
