OpenAI recently said that they have a tool for cloning people’s voices, but they won’t be giving it to the public because they are worried that it will be hacked.
Voice Engine is a voice-replication tool that can use a 15-second audio clip of a person to accurately copy their voice. ChatGPT’s author has said that they are being careful with how they use this technology.
“We know that making speech that sounds like people’s voices comes with big risks,” OpenAI said. “These risks are especially important this election year.”
They went on to say, “We are working with U.S. and international partners in government, media, entertainment, education, civil society, and other areas to make sure we take their feedback into account as we build.”
People say that the company will keep trying the technology and then make a “more informed decision” about whether to make it available to everyone. They are working on two things: making sure that the person being taped knows that they are being recorded, and making sure that voices that sound too much like celebrities or politicians are not made.
People have a lot of problems with AI being used in bad ways, and this has made people worried about the upcoming election. Since we’ve seen this happen before, voice changes have already been a problem. A lot of different ideas are being worked on to make sure that the Voice engine isn’t abused.
OpenAI says they’ve done things like adding a watermark to any Voice Engine-generated music so that it can be tracked back to its source and keeping an eye on how it’s being used.
