Meta AI trained with your Instagram posts!

One of the managers at Meta has confirmed that the company’s new Meta AI service has been trained on public posts of Instagram users.

In an interview with Reuters, the Senior Policy Manager at Meta stated that the company has used public posts from Facebook and Instagram to train parts of the Meta AI assistant. However, they have never utilized private posts shared exclusively with family and friends to respect privacy. Meta has also used the text of these public posts. I suggest you see the article on Instagram Tutorial: From 0 To 100 For Beginners.

Meta also stated that the Meta AI service does not utilize private chat conversations from messaging services like WhatsApp and Messenger. The company has tried to filter out personal details from public datasets to prevent their use in training artificial intelligence.

Nick Clegg, Head of Global Affairs at Meta, said, “We have tried not to use datasets that are predominantly made up of personal data.” He further added that most of the data used for training artificial intelligence were publicly accessible data.

Nick Clegg stated that LinkedIn is an example of a websites that Meta never uses for training artificial intelligence because the company is concerned about user privacy.

The Meta AI service was unveiled on Thursday night during the Meta Connect 2023 event. This service introduces new AI-based features to the Instagram, Facebook, and WhatsApp platforms. Meta has also introduced AI personalities for Instagram.

Meta AI assistant

The Meta AI assistant is built on a large language model called Llama 2, made publicly available a few months ago. According to Meta, the Meta AI text-to-image conversion service utilizes a new language model called Emu.

Meta’s artificial intelligence can generate text, audio files, and images. Thanks to the Microsoft Bing search engine, this service also has real-time access to web data.