Last week OpenAI unveiled its speech-cloning technology Voice Engine, which you may have read about under various news headlines in the vein of “The AI technology too dangerous to be released”.

This is because, in OpenAI’s blog post, it explained why it’s creating the model and what benefits it could bring, while also saying the world was not ready for it and that it could be a menace in the wrong hands. We’ve seen this before from OpenAI, which talks a lot about safe and responsible AI, but which, it’s worth remembering, has both a non-profit and a commercial arm.

OpenAI says its new tool can teach a machine to speak with any person’s voice after just 15 seconds of training.Credit: Marija Ercegovac

The company has frequently introduced new developments by talking about how potentially ruinous they could be, which it has to know also makes them seem more cool, powerful and valuable. So will Voice Engine be a force for good, bad, or both?

OpenAI is developing models across all kinds of media – from text to video – which can train on examples and then generate “original” content according a prompt. These so-called generative AI models power its consumer products, including GPT for text, DALL-E for images, and Whisper for audio-to-text transcription.

Voice Engine is an in-development model that can train on an individual’s voice, and then read any text using that voice. Imagine Siri or Google Assistant reading the content of a web page, except it sounds just like you. Or anyone, for that matter.

We don’t have a lot to go on in terms of judging Voice Engine’s capabilities, apart from the five examples OpenAI has provided. And while they appear impressive, they are likely to be the best-case scenario and not a typical result. Similar technology tends to sound very accurate in certain output, and hollow or robotic in others.

QOSHE - OpenAI’s new Voice Engine could bring real benefits, and real dangers - Tim Biggs
menu_open
Columnists Actual . Favourites . Archive
We use cookies to provide some features and experiences in QOSHE

More information  .  Close
Aa Aa Aa
- A +

OpenAI’s new Voice Engine could bring real benefits, and real dangers

9 17
08.04.2024

Last week OpenAI unveiled its speech-cloning technology Voice Engine, which you may have read about under various news headlines in the vein of “The AI technology too dangerous to be released”.

This is because, in OpenAI’s blog post, it explained why it’s creating the model and what benefits it could bring, while also saying the world was not ready for it and that it could be a menace in the wrong hands. We’ve seen this before from OpenAI, which........

© The Sydney Morning Herald


Get it on Google Play