Meta cancels AI voice generator launch amid misuse and harm concerns

Meta, the parent company of Facebook, has introduced its latest AI tool called ‘Voicebox’, touting it as a groundbreaking advancement in AI-driven speech generation. Voicebox is capable of generating audio clips of speech in six European languages, producing reasonably accurate text-to-speech replications with just a two-second audio sample. However, Meta has decided not to release the tool to the public due to concerns about potential misuse and harmful consequences.

One major worry is the tool’s potential for creating lifelike voice imitations of public figures, such as politicians, by compiling existing audio recordings available on the internet. This raises serious implications for misinformation and conflicts that could arise from such manipulations.

Acknowledging these risks, Meta expressed its recognition of the potential for misuse and unintended harm associated with Voicebox and similar technologies. To mitigate these concerns, Meta plans to release a research paper and a classifier tool that can distinguish between Voicebox-generated speech and genuine human speech. While the classifier is described as “highly effective,” it is not claimed to be perfect.

Despite the associated risks, Meta highlighted the potential benefits of AI speech generation. Voicebox has the potential to provide more naturalistic speech for individuals who are mute or have communication difficulties, thereby reducing barriers to interaction. Additionally, it could facilitate real-time translation, bringing us closer to the concept of a “universal translator” depicted in science fiction.

However, in the interest of preventing potential misuse and harm, Meta has taken a cautious approach and opted not to release Voicebox publicly at this time.