The newest from OpenAI is a characteristic that reads textual content aloud in a remarkably human-like voice. This breakthrough in synthetic intelligence marks a major leap ahead, however it additionally raises issues in regards to the potential for deepfake manipulation (by way of Bloomberg).
The corporate has unveiled early outcomes from testing this characteristic, providing demos, which you’ll hearken to here. Dubbed Voice Engine, this text-to-speech mannequin is at the moment in a restricted trial section with about 10 builders. OpenAI has opted for a cautious method fairly than a widespread launch.
Following suggestions from stakeholders like policymakers and educators, OpenAI has determined to cut back its preliminary rollout. The corporate acknowledges the intense dangers of producing human-like speech, particularly throughout delicate instances like an election yr.
The corporate wrote in a weblog publish:
We acknowledge that producing speech that resembles folks’s voices has critical dangers, that are particularly prime of thoughts in an election yr. We’re partaking with US and worldwide companions from throughout authorities, media, leisure, schooling, civil society, and past to make sure we’re incorporating their suggestions as we construct.
Not like earlier audio tasks, Voice Engine stands out for its capacity to imitate particular person voices with exceptional accuracy, capturing nuances in cadence and intonation. And all it wants is simply 15 seconds to copy an individual’s voice.
Amongst OpenAI’s companions is the Norman Prince Neurosciences Institute at Lifespan, the place the know-how is used to assist sufferers in voice rehabilitation. As an illustration, it was used to revive the speech of a younger affected person who had issue talking clearly as a result of a mind tumor. The AI discovered from earlier recordings for a faculty undertaking.
Along with its purposes in healthcare, the customized speech mannequin has caught the eye of firms like Spotify, which sees potential in translating audio content material, akin to podcasts, into a number of languages. Nonetheless, OpenAI emphasizes moral tips for utilizing the know-how, together with acquiring consent from authentic audio system and disclosing AI-generated content material to listeners.
Additionally, earlier than contemplating a wider launch, OpenAI is soliciting suggestions and urging public consciousness of the challenges posed by superior AI tech. This contains advocating for the phasing out of voice authentication in delicate areas like banking.
OpenAI added in its weblog publish:
It’s vital that individuals all over the world perceive the place this know-how is headed, whether or not we finally deploy it broadly ourselves or not.
Moreover, the corporate provides that it hopes this preview sparks a dialog about addressing the dangers related to AI developments and selling societal resilience.
