Meta announces a new AI model which can mimic the voices of loved one, but claims it is too risky for release.
The company stated in a recent research article that there are many interesting use cases for generative models. However, due to the risk of misuse, the code or model of Voicebox is not publicly available. While we think it’s important to be transparent with the AI community, and to share our work to advance the current state of AI, it’s equally necessary to strike a balance between openness and responsibility.
Insider’s comment request made after normal business hours was not immediately responded to by Meta.
Meta stated in a news release that this model would allow people with visual impairments to hear their friends’ voices in their messages or speak foreign languages themselves. The company said that the technology also allows creators to edit audio for videos or create more realistic-sounding virtual assistants.
Source:
https://www.businessinsider.com/meta-generative-ai-voice-model-too-risky-release-2023-6?utm_campaign=business-sf&utm_source=facebook&utm_medium=social
