Meta, the parent company of Facebook and Instagram, recently revealed that it used public posts from these platforms to train its new AI assistants. However, Meta made it clear that it excluded private posts and conversations from the training data. According to Meta’s President of Global Affairs, Nick Clegg, they aimed to avoid datasets containing personal information. Clegg stated that the majority of the data used to train the AI assistants was publicly available.
The AI chatbots developed by Meta can assist users with various tasks such as answering questions, planning trips, and generating images from text prompts. These chatbots are being rolled out across messaging platforms including WhatsApp, Instagram, and Messenger. They are also capable of providing real-time search results from Microsoft’s Bing search engine.
Meta’s AI assistants were partially based on its LLaMA 2 language model, which was introduced in July. In addition to the general assistant, Meta is launching 28 AI-powered characters on its messaging platforms, some of which are modeled after celebrities like Snoop Dogg, Kendall Jenner, and Tom Brady. While the chatbots were launched in limited beta, they have yet to be made available to the general public.
It’s worth noting that Meta’s use of public data for training AI models aligns with recent controversies surrounding companies using web content without the original author’s permission. This includes Elon Musk’s xAI project, which plans to train its AI on tweets and has faced backlash from artists who share their content on Twitter.
