Posted by
For the second time in a matter of months, OpenAI has found itself explaining its text-to-audio tool, reminding everyone that it is not, and may never be, widely available.
“It’s important that people around the world understand where this technology is headed, whether we ultimately deploy it widely ourselves or not,” the company said in a statement posted to its website on Friday. “Which is why we want to explain how the model works, how we use it for research and education, and how we are implementing our safety measures around it.
Late last year, OpenAI shared its Voice Engine, which relies on text inputs and 15-second audio clips of human voices to “generate natural-sounding speech that closely resembles the original speaker,” with a small group of users outside the company. The tool can create voices that sound convincingly human in several languages.
At the time, the company said it was choosing to preview the technology but not widely release it to “bolster societal resilience” against the threat of “ever more convincing generative models.”
As part of those efforts, OpenAI said it was actively working on phasing out voice-based authentication for accessing bank accounts, exploring policies to protect the use of individual’s voices in AI, educating the public on the risks of AI, and accelerating development on tracking audiovisual content so users know whether they’re interacting with real or synthetic content.
But despite such efforts, fear of the technology persists.
President Joe Biden’s AI chief, Bruce Reed, once said that voice cloning is the one thing that keeps him up at night. And The Federal Trade Commission said in March that scammers were using AI to elevate their work, using voice cloning tools that make it harder to distinguish between AI-generated voices and human ones.
In its updated statement on Friday, OpenAI sought to assuage those worries.
“We continue to engage with US and international partners from across government, media, entertainment, education, civil society, and beyond to ensure we are incorporating their feedback as we build,” the company said.
It also noted that once Voice Engine is equipped with its latest model, GPT4o, it’ll also pose new threats. Internally, the company said it’s “actively red-teaming GPT-4o to identify and address both known and unforeseen risks across various fields such as social psychology, bias and fairness, and misinformation.”
The bigger question, of course, is what will happen when the technology is widely released. And it looks like OpenAI might be bracing itself, too.
OpenAI did not immediately respond to a request for comment from Business Insider.