EERIE new AI that can clone your voice with only a 15 second sample has been revealed by the creators of ChatGPT.
OpenAI has shared early clips that bosses claim can create “emotive and realistic voices”.
3
3
But the firm admits its tech is too risky in the hands of the general public right now with looming elections in the US and UK.
Experts fear that Russia, China, and North Korea are already working on hijacking the presidential election using their own deceptive AI and deepfakes to manipulate results.
We recognize that generating speech that resembles people’s voices has serious risks, which are especially top of mind in an election year
OpenAI
Even billionaire Elon Musk, 52 – who has launched legal action against OpenAI – believes AI could be one of the biggest threats to humanity.
OpenAI has already showed the world how it can generate text in a matter of seconds with ChatGPT, as well as images and more recently videos.
Read more about AI
Now the company is going a step further, revealing that they started privately testing a so-called “Voice Engine” late last year.
Despite the concerns, OpenAI continued development to see how the tool could be “used for good”.
It shared several examples, such as providing assistance to children or those who can’t read with “natural-sounding, emotive voices”.
Another demonstrates how the tech could help a person who’s lost their fluent voice due to a vascular brain tumour.
Most read in Science
Doctors were able to restore the voice of a young patient using 15 seconds of audio from an old video recording, OpenAI claims.
The alarmingly precise AI can even translate a person’s voice and dialogue into different languages that still sound like them.
Can you spot the two mistakes in this AI video?
For now, OpenAI says it is “choosing to preview but not widely release this technology”.
“We hope this preview of Voice Engine both underscores its potential and also motivates the need to bolster societal resilience against the challenges brought by ever more convincing generative models,” the company wrote in a blog post.
“We recognize that generating speech that resembles people’s voices has serious risks, which are especially top of mind in an election year.”
It also notes that early Voice Engine testers have agreed to not impersonate a person without their consent and to disclose that the voices are AI-generated.
A trademark application filed on March 19 suggests OpenAI may be plotting a move into the business of speech recognition and digital voice assistants to rival the likes of Alexa.
3
Read more about Artificial Intelligence
Everything you need to know about the latest developments in Artificial Intelligence
