Meta Says It Trained Latest AI Using Public Instagram, Facebook Posts
Meta has been using public Facebook and Instagram posts to train its latest artificial intelligence (AI) assistant, a company spokesperson said on Thursday.?
According to Nick Clegg, Meta's president of global affairs, public posts on Facebook and Instagram were fed to the new AI - both text and photos, Reuters reported.
Meta used public data to train latest AI
At the same time, Clegg assured that the AI steered clear of private posts like those shared only with friends and family. The Meta official referred to certain measures that exclude private details from being fed to this AI, but the specifics of these measures were not revealed.
In addition, Clegg told Reuters that Meta isn't using private messages on its platforms like Facebook and Instagram to train the new model. "We've tried to exclude datasets that have a heavy preponderance of personal information," he said, while adding that Meta has not used websites like LinkedIn to train its AI owing to privacy concerns.
Meta's new virtual assistant called "Meta AI" was opened to the public on Wednesday in a beta release. The new AI assistant can help users create digital stickers based on text prompts, edit images with text instructions, and chat with AI personalities based on specific prompts.
Also read:?Meta Unveils New Ray-Ban Smart Glasses With Fascinating Live-Streaming Features
Meta AI was built using the company's own AI language model called Llama 2 and a text-to-image model called Emu. Both these models were trained using public Instagram and Facebook posts.?
Clegg added that some people might legally contest whether training AI with copyrighted content is fair use. "We think it is, but I strongly suspect that's going to play out in litigation," Clegg said.
Artists have been fighting companies like Meta in court over concerns surrounding the unconsented usage of their work to train AI tools that could undermine their careers in the long-run.
Also read:?Meta Turned Down WhatsApp Founder Brian Acton, Only To Give Him Billions Later
Users of Facebook own the content posted by them as long as it doesn't infringe on someone's intellectual property rights, Meta policies claim.
What do you think about AI tools trained on public data? Let us know in the comments below.?For more in the world of?technology?and?science, keep reading?Indiatimes.com.??
