For the 10th anniversary of Meta Connect, Mark Zuckerberg has shared how rising AI and metaverse technologies reshape the way consumers experience physical and digital worlds, as well as Meta’s progress on AI, and AR glasses.
Although AI technology keeps advancing, many still haven’t experienced it first-hand, and Meta is figuring out how to change that by building state-of-the-art AI into apps already in use by billions.
Meta have unveiled their image generation model of its Llama ecosystem, named EMU (Expressive Media Universe), it uses text prompts to generate high-quality images in seconds. Pairing with Llama 2, users can also create their own custom AI stickers.
Also introduced was restyle and backdrop, two new features on their way to Instagram, using EMU technology to allow users to transform photos, or co-create AI generated images. Restyle allows for the reimagining of images by applying visual styles described, while backdrop leverages learning from the Segment Anything Model to change the image’s scenery/background.
Images created using restyle and backdrop indicate the use of AI to reduce odds of mistaking the images for human generated content. Meta are also experiencing with forms of visible and invisible markers to better distinguish AI generated content.
Meta are also building AI studio, a platform to be used to create AIs that help users get things done while having fun. Individuals can interact with the AIs across the Meta universe, and they will have Instagram and Facebook profiles where people can chat with them. Eventually, they’ll also be embodied in the metaverse.
AI Studio has created some AIs that have started rolling out in the US.
MetaAI is a new assistant that can be interacted with like a person, using a custom model based on Llama 2 technology, and has access to real time information through a Bing search partnership. EMU has also been built in for the generating of high-quality images for free in seconds.
The company have created some AIs with more personality, opinions, and interests. There’s Victor, a motivational coach that encourages goals, the Dungeon Master that takes users on an old school, text based adventure, and sous chef Max who can create recipes through a random assortment of ingredients. Several more are expected over the next few weeks.
AI Studio will let developers build third party AIs for messaging service, and Meta are creating a sandbox which will allow those who don’t code, to create their own AIs.
Generative AI does come with its challenges, but Meta are ensuring they’re putting in time and effort to get it right, including training and fine-tuning models that fit into the safety and responsibility guidelines, internal teams that help ensure the models safer and inclusive, and red teaming with external experts.
Now onto a new product, built in partnership with EssilorLuxottica, the next generation smart glasses were designed to help users stay connected and capture the moment without the need to take out a phone.
They’ve been upgraded from the first gen models, and for the first time, users can livestream directly from the smart glasses to friends and followers on Facebook and Instagram.
These are also the first to ship with Meta AI built in. Shipping begins in the US in beta, with a free update rollout scheduled for next year. The Ray-Ban Meta smart glasses start retailing at $449.00 AUD, and will launch on October 17th, but are available for pre-order now on meta.com and ray-ban.com.
Meta are currently hard at work attempting to deliver new technologies to the world, as well as making them available to everyone.
