Meta, the owner of Facebook, Instagram and WhatsApp, has announced that the new “speech generating” artificial intelligence system known as “Voicebox” will not be released yet due to the risk of abuse.
In the statement made by the company about the system that Meta announced and can produce speech in 6 languages, it was stated that Voicebox will not be released yet due to risks such as abuse.
Trained with 50 thousand hours of recordings in English, French, Spanish, German, Polish and Portuguese, Voicebox can fill the gap in speech by inferring from the context. The system is also capable of producing different styles of audio from raw audio and transcription.
Voicebox, which can create a realistic sound even from only two seconds of speech, preserves the style between two languages and helps communication.
Meta stated that speech-generating AI models are exciting, but that Voicebox or the code for the system is not shared due to potential risks.
Numerous reports state that artificial intelligence can copy people’s voices without their consent, and the recordings in question can be used for fraudulent activities or to create fake videos.