The parent company of Facebook has launched a new voice and sound cloning tool that will not only create a cloned sounds or words, but will also record your voice and have it cloned, too.
Called Audiobox, the platform was released by Meta and built by researchers at the Facebook AI Research (FAIR) lab.
Why This Matters: This is yet the latest in a growing body of technology and tools that give individuals the power to create artificial versions of themselves or others which can be used in any number of situations, including interactions with collection agents, customer service representatives, and others over the phone.
- Audiobox is a Self-Supervised Learning (SSL) platform that uses artificial intelligence algorithms to assign labels to unstructured data. This allows the platform to train itself without supervision using only the audio that is supplied by the users.
- In building the platform, researchers relied upon 160,000 hours of speech, 20,000 hours of music, and 6,000 hours of sound samples.
- Meta did announce that the tool is not to be used for commercial purposes and is only allowed to be used outside the states of Illinois and Texas.
Just Imagine the Possibilities: Consumers can use this tool to replicate their voice or create whole new voices. Imagine creating background noise to make it sound like you are typing or working in an office when you’re really working from home.
- The blurring lines between reality and artificial reality are getting increasingly blurrier by the day. Companies in the ARM industry need to be prepared for situations where the individuals they are interacting with on a phone call or a webchat are not actually humans.
Not to be Outdone: Similarly, I ran across this post on Reddit last week, which shows a user using ChatGPT to place 10,000 calls simultaneously using the cloned voice of anyone. In this case, the user opted for the voice of former president Barack Obama.