Meta
Platforms will soon begin training its own AI models on data from its app users in the EU. This includes public posts and comments on apps like Facebook and Instagram, as well as the history of user interactions with the Meta AI bot.
Image source: Steve Johnson/unsplash.com
It is noted that the data of personal correspondence of users with friends and family members will not be used to train neural networks. In addition, the algorithms will be trained only on the data of adult users. Meta will begin notifying EU residents about this this week. The notification will contain an opt-out form for those who do not want their data to be used to train neural networks.
It should be noted that Meta planned to use EU user data last year, but this process was suspended at the request of the industry regulator in the region. The company claims that training on EU user data will allow AI models to better take into account regional characteristics. It is noted that this is especially important for content generation by multimodal neural networks, which are capable of creating not only text responses to queries, but also images and videos.
Last year, Meta announced its intention to train its neural networks on data from UK users. However, the company is unlikely to receive as much data as it has already managed to use. The fact is that last year, Meta admitted to using all text posts and photos that Facebook users have publicly posted since 2007 to train its AI.