Meta Platforms, the parent company of Facebook and Instagram, has revealed that it used public posts on these social media platforms to train parts of its new Meta AI virtual assistant. This information was shared by the company’s top policy executive, who also emphasized that Meta made a conscious effort to exclude private posts shared only with family and friends in order to prioritize consumer privacy.
To further protect user privacy, Meta did not utilize private chats on its messaging services as training data. Additionally, the company took steps to filter out private details from public datasets that were used for training the AI. Nick Clegg, the President of Global Affairs at Meta, stated that the company made an effort to avoid datasets that contained a high volume of personal information. In fact, platforms like LinkedIn were intentionally not utilized due to concerns surrounding privacy.
The use of scraped information from the internet without permission to train AI models has recently faced criticism. However, Meta is committed to respecting privacy and ensuring that proper protocols are followed. The company’s annual Connect conference, which focused on artificial intelligence this year, showcased Meta AI as its most significant product. This virtual assistant was developed using Meta’s custom models called Llama 2 and Emu, enabling it to generate text, audio, and imagery.
It is important to note that Meta AI used public posts from Facebook and Instagram for training purposes, including both text and photos. However, Meta has implemented safety restrictions to regulate the content generated by Meta AI. For instance, the assistant is barred from creating photo-realistic images of public figures. This measure aims to prevent any potential misuse or manipulation of the AI technology.
As Meta moves forward with this groundbreaking development, the company anticipates potential litigation surrounding the use of copyrighted materials and the application of fair use doctrine. Nevertheless, Meta has updated its terms of service, explicitly prohibiting users from generating content that violates privacy and intellectual property rights.
In conclusion, Meta Platforms has utilized public posts from Facebook and Instagram to train its new Meta AI virtual assistant, while respecting consumer privacy by excluding private posts shared with family and friends. The company has also taken measures to filter private information and has implemented safety restrictions on the content generated by Meta AI. As Meta faces potential legal challenges, it remains committed to maintaining user privacy and adhering to intellectual property rights.