Did you know a virtual assistant was being trained using your social media posts?
The parent company of Facebook, Meta, has admitted that it is training its new AI assistant, Meta AI, using public posts from Instagram and Facebook members.
What is this “Meta AI,” exactly?
It’s a chatbot that can produce lifelike visuals in response to written questions. It will soon be on Ray-Ban Meta smart glasses and Quest 3 and is now in beta on WhatsApp and Messenger. Generative technologies like Llama 2 and Emu fuel Meta AI by generating text and visuals from massive datasets.
Meta’s president of global relations, Nick Clegg, claims that the internet giant is training its artificial intelligence on text and photographs from public posts on Instagram and Facebook. He claims that the posts are culled from the most popular and actively discussed and that any identifying information is removed before being given to the AI. He adds that Meta has included safety features in Meta AI to protect against abuse and misapplication, such as blocking access to potentially offending material.
Users have voiced worries about the privacy and ethics of public posts being used to develop Meta AI. Meta, they say, should not have utilized their posts without their permission, and they should have been informed of the data’s intended use.
They also wonder if Meta has taken adequate precautions to prevent Meta AI from producing either damaging or misleading content or infringing on the rights of their work.
Neither the number of posts utilized to train Meta AI nor the method by which members whose posts are used will be made public by Meta at this time. How it will deal with members who seek to have their postings removed from being used to train AI is also unknown. Meta has promised to continue using user-generated content to train its artificial intelligence (AI) and develop new AI capabilities, including AI stickers, restyle, and backdrop.
Concerns about the members’ right to privacy and the validity of their permission have been raised. Meta needs to be more open and accountable with its users’ data, protecting it while considering users’ privacy preferences and rights.