Meta admits its new AI chatbot is trained on posts by users on Facebook and Instagram

Meta executives have revealed that their AI virtual assistant, dubbed Meta AI, was trained on public posts made by its users on Facebook and Instagram.

Meta President of Global Affairs Nick Clegg, however, clarified that they didn’t scrape through the private Facebook messenger and Instagram chats that are shared only with family and friends so as to respect consumers’ privacy.

"We've tried to exclude datasets that have a heavy preponderance of personal information," Clegg told Reuters news agency on Thursday (Sept 28), adding that the "vast majority" of the data used by Meta for training was publicly available.

Add WION as a Preferred Source

The clarification seeks to assuage the privacy concerns over tech companies like Meta, OpenAI, and Alphabet, the parent company of Google, using information scraped from the internet without permission to train their AI models.

Built on powerful Lama2 language model

Meta AI, which was unveiled by CEO Mark Zuckerberg on Wednesday at Meta's annual products conference, Connect, is built on a custom model based on the powerful Llama 2 large language model, and a new model called Emu that generates images in response to text prompts.

The product will be able to generate text, audio and imagery and will have access to real-time information via a partnership with Microsoft's Bing search engine, Clegg said.

“The public Facebook and Instagram posts that were used to train Meta AI included both text and photos,” he added.

Clegg assured that Meta imposed safety restrictions on what content its virtual assistant AI tool could generate, like a ban on the creation of photo-realistic images of public figures.

Speaking on copyrighted materials, Clegg said he was expecting a "fair amount of litigation" over the matter of "whether creative content is covered or not by existing fair use doctrine," which permits the limited use of protected works for purposes such as commentary, research and parody.

"We think it is, but I strongly suspect that's going to play out in litigation," Clegg said.

AI still in beta version

Meta, which is currently in beta version, is an “advanced conversational assistant” that can generate text, audio, and imagery, and have real-time access to information through a partnership with Microsoft's Bing search engine.

The company states that the AI chatbot will be limited to information that largely existed prior to 2023,

Meta will eventually launch an AI creator studio for businesses and developers to build their own personalized AI chatbot.

Currently, Meta’s new AI tool is only available in the US.

(With inputs from agencies)

WATCH WION LIVE HERE