[ad_1]
It’s no secret that Meta has a vision for generative AI coming to all of its platforms — Instagram, WhatsApp, and Facebook — in every way possible. But generative AI isn’t trained on nothing. According to a report from Reuters, Meta AI, the social media tech giant’s new assistant, will be trained on your public Facebook and Instagram posts.
At Meta Connect 2023, CEO Mark Zuckerberg announced new AI features for image generation, AI-generated celebrities, and a personal assistant. The personal assistant will, ideally, provide users information they need (think ChatGPT) and also perform tasks across its products and services, including WhatsApp, Facebook Messenger and Instagram. Even the new-generation Ray-Ban Meta smart glasses and Quest 3 are getting a taste of Meta AI.
It is powered by an open-source large language model called Llama 2, which can already chat with users to answer questions. Llama 2 works in partnership with Bing, so it has access to real-time information.
But some of the training for the Meta AI virtual assistant comes from public Facebook and Instagram posts, Meta President of Global Affairs Nick Clegg told Reuters in an interview on Thursday. Clegg said the company doesn’t use private posts nor private chats on messaging services in order to respect users’ privacy.
“We’ve tried to exclude datasets that have a heavy preponderance of personal information,” Clegg told Reuters. He added that the “vast majority” of training data used was publicly available.
Think about the posts you’ve made publicly on Facebook. Now imagine an AI virtual assistant with all that knowledge. I, for one, cannot wait for a personal assistant who was meticulously trained on my shitposts.
[ad_2]
Source link
