OpenAI introduces voice cloning device, however it's deemed too dangerous for public launch


The ChatGPIT creator says the voice engine can mimic an individual's voice primarily based on a 15-second audio pattern.

OpenAI has unveiled a device for cloning individuals's voices, however has been holding off on releasing it publicly over issues of potential misuse in a key election yr.

Based on an OpenAI weblog put up showcasing the device, the voice engine can mimic an individual's voice primarily based on a 15-second audio pattern.

However the ChatGPIT creator is “taking a cautious and knowledgeable method” to the know-how and hopes to start out a dialog on the “accountable deployment of artificial voices,” the corporate mentioned in a weblog put up printed Friday.

“We acknowledge that there are severe dangers in crafting a speech that sounds just like the voice of the individuals, who’re particularly high of thoughts in an election yr,” the San Francisco-based start-up mentioned.

“We’re participating with U.S. and worldwide companions throughout authorities, media, leisure, academia, civil society and past to make sure we’re incorporating their suggestions as we construct.”

OpenAI mentioned it might make “extra knowledgeable choices” about whether or not to deploy the know-how at scale primarily based on testing and public debate.

The corporate mentioned it believes the know-how ought to solely be launched with measures in place that guarantee “native audio system are deliberately including their voices to the service” and “forestall the creation of voices that resemble distinguished personalities.” be just like.”

Abuse of AI has emerged as a significant concern forward of elections this yr in international locations representing nearly half the world's inhabitants.

Voters in additional than 80 international locations, together with Mexico, South Africa and the US, are headed to the polls in 2024, in what has been billed as the most important election yr in historical past.

The affect of AI on voters has already come below scrutiny in a number of elections.

Pakistan's jailed former Prime Minister Imran Khan used AI-generated speeches to enchantment to supporters within the run-up to the nation's parliamentary elections in February.

In January, a political operative of longtime US presidential candidate Dean Phillips issued a robocall impersonating US President Joe Biden, urging voters to not forged their ballots in New Hampshire's Democratic Occasion main.

OpenAI mentioned it has applied quite a few safety measures for its companions testing the voice engine, together with watermarking to hint the origin of any audio generated by the voice engine, in addition to how it’s getting used. Energetic surveillance is concerned.

Leave a Comment