With Messenger, we’ve made it simple to connect with the people who matter in your life, whether that means sending Reels to friends or video calling loved ones. Around the world, people  collectively spend over 7 billion minutes on calls across Facebook and Messenger every day, and we’re constantly working on new ways to enhance the experience. 

That’s why today, we’re announcing new Messenger Calling features, many of which are some of our most requested. Now, calling is easier, more reliable and more fun. 

AI Backgrounds in Video Calling

In September, we announced new ways to personalize Messenger chat themes with unique, AI-generated images from Meta AI. Soon, you’ll be able to use AI backgrounds in Messenger video calls to reflect how you’re feeling or just add a little excitement to your call. 

To start creating your personalized AI background, tap on the effects icon in the sidebar of your next video call and select “Backgrounds.” 

HD Video Calls and Noise Suppression

We’re introducing HD video calls, background noise suppression and voice isolation to help you make clearer, higher-quality calls directly from Messenger. Talking digitally will feel more than ever like being in a room together. 

Click to view slideshow.

HD will be enabled by default for video calls made on WiFi.  To enable HD for video calls on cellular data, go to your call settings and turn on “Mobile data for HD video.” 

Background noise suppression and voice isolation can be enabled via call settings in Messenger. 

Audio and Video Messages

Messenger Calling is becoming more and more like a fully outfitted phone, plus some extra features.  Now, you can leave audio or video voice messages when your friends can’t pick up. 

Click to view slideshow.

Tap the “Record message” button to leave an audio message for unanswered audio calls or a video message for unanswered video calls. 

Hands-Free Calling and Messaging

When your hands are full or your phone is out of reach, you can now ask Siri to help you make your calls and messages wherever you are. Try something like: “Hey Siri, send a message to Cassandra on Messenger” and then dictate what that message will be.  

The post Introducing AI Backgrounds, HD Video Calls, Noise Suppression and More for Messenger Calling appeared first on Meta.

We want to make sure everyone on Instagram – especially teens – has safe, positive, age-appropriate experiences and feels the time they’re spending on Instagram is valuable. In addition to providing built-in protections from sensitive content with Teen Accounts, we want to give teens new ways to shape their Instagram experience, so it can continue to reflect their passions and interests as they evolve.

That’s why we’ve started testing the ability for everyone on Instagram – including teens – to reset their recommendations. In just a few taps, you’ll be able to clear your recommended content across Explore, Reels and Feed and start fresh. 

Your recommendations will start to personalize again over time, showing new content based on the content and accounts you interact with. When resetting, you’ll also have the option to review the accounts you’re following and unfollow any that share content you no longer want to see.

We’re excited to partner with organization experts Joanna and Clea from The Home Edit to share tips for parents on how both they and their teens can organize their Instagram accounts, just like they’d organize their homes, taking stock of what they’re looking at every now and then to see whether it still suits them and whether there’s other content they may prefer. 

Ways to Shape Your Instagram Experience

This new feature, which will soon roll out globally, builds on the tools we already offer to let people curate what they see in their recommendations. 

For example, people can tell us they like the content recommended in their Explore page by tapping the three dots in the corner of the post and selecting “Interested.” We recently announced a new way for teens in the US, UK, Canada and Australia to choose topics they want to see more of – everything from books and travel to cooking and sports. We also offer people ways to tell us what they don’t want to see in their recommendations by selecting “Not interested” from the three-dot menu, or using our Hidden Words feature to hide content with certain words or phrases in the caption.

We also offer a range of tools to help teens shape their Instagram experience, beyond recommendations. For example, teens can switch to a Following Feed to see, in chronological order, content from accounts they follow, with the most recent post first. Alternatively, they can add accounts to a Favorites list so they see content from those accounts more often and higher in their Feed, and they can see a dedicated feed of just their favorites when they want to catch up on their posts quickly. 

Features like Close Friends give teens more control over who sees their Stories, while the “Your activity” section in settings lets teens see and manage all their previous interactions in one place, giving them more control over their digital footprint. We also encourage teens to regularly review their following lists, to make sure the accounts they’re following are still accounts whose content they enjoy seeing. 

Our Approach to Age-Appropriate Content

We recognize parents are concerned that their teens might see mature or inappropriate content online, which is why we have stricter rules around the kinds of content teens can see on our apps. In addition to removing content completely when it breaks our rules, we take steps to help prevent teens from seeing content that may be sensitive or mature. We avoid recommending it, and in some cases we hide it from teens altogether, even if it’s shared by someone they follow.

Today we’ve published a new page in our Transparency Center, to help parents and teens better understand how this approach to content works. The new page sets out the different layers of protection and provides examples of the types of content we remove, hide from teens only and avoid recommending. 

The post Reshape Your Instagram With a Recommendations Reset appeared first on Meta.

Meta yesterday concluded its groundbreaking Llama Impact Hackathon in London, marking a significant milestone in AI innovation aimed at transforming public services. The event, held in collaboration with Cerebral Valley brought together over 200 developers across 56 teams, all leveraging Meta’s open source Llama 3.2 model to address critical challenges in healthcare, clean energy, and social mobility.

Revolutionising NHS Frontline Services

The winning team, Guardian, developed a concept of an AI-powered triage assistant which could reduce waiting times and better allocate resources in A&E departments. This innovative solution, built on Llama 3.2, enhances patient care through intelligent patient intake and real-time risk assessments. Guardian’s tool, Atlas, acts as a clinical AI agent, providing crucial support to frontline medical staff and facilitating communication in multiple languages to improve patient evaluations.

Celebrating Innovation and Collaboration

The hackathon showcased the potential of open source AI to drive positive social impact. The top three teams shared a prize fund of $50,000 and will receive six weeks of technical mentorship to further develop their projects. Participation in this hackathon also enables these teams to apply for both regional grants of up to $100K and global grants of up to $500K through the Llama 3.1 Impact Grants which close on 1st December 2024.

Empowering Developers and Researchers

“It was inspiring to be at Meta and discuss how open source AI can be harnessed for public benefit. Whether accelerating cancer diagnoses, boosting productivity, or developing new tools to combat climate change, this technology holds immense potential. That’s why we’re placing AI at the forefront to not only improve public services and stimulate economic growth but also deliver a brighter future for communities across the country,” – UK Minister for AI Feryal Clark.

“The UK has the developer talent, research base, and creativity to lead in deploying AI. With access to open source AI models like Llama, developers and researchers can craft tools and systems tailored to Britain’s challenges, from enhancing public service delivery to boosting workplace productivity and aiding scientific breakthroughs. We eagerly anticipate these projects coming to life and look forward to working with our winners as well as the UK Government to develop them further.” – Nick Clegg, President of Global Affairs at Meta.

“The talent at this London Llama Hackathon was some of the best we’ve ever seen. More than 200 developers come out to build projects for the public good and demonstrate how open source AI can create positive social impact.” – Ivan Porollo Founder, Cerebral Valley.

Meet the Finalists

We would like to extend our congratulations to all the finalists who participated in the UK Llama Impact Hackathon. Their innovative projects demonstrate the vast potential of AI to drive positive change in the UK’s public services. Here were the finalists:

• First place – Guardian: An AI-powered triage assistant to support frontline NHS staff. Built on Llama 3.2, Guardian transforms A&E departments through intelligent patient intake, real-time risk assessment, and Atlas – a breakthrough clinical AI agent which provides a second pair of eyes to doctors and nurses. It will also support patients across multiple languages helping them communicate their symptoms better. 
• Second Place – Gripmind: Is a solution which makes robotics in assisted living more useful, scalable and affordable. An open source project combining the Llama 3.2 Vision model, which processes signals from the brain, your voice or via images to control a robotic arm. It’s envisaged this solution could help support people with mobility issues.
• Third place – Pharmallama: An on-device app which allows patients to engage with their pharmacist to discuss potential side effects they may be experiencing; understand what medication they need to be on and centralises the patient records to spot potential conflicting medications. This will be particularly beneficial for those with mobility issues or who are unable to access a more local pharmacy.  
• ClimaticAI: Heat the person, not the home, and save up to a third on energy bills. Their AI-powered concept works with smart devices, infrared-controlled appliances, and IoT sensors to optimise energy use based on presence, cost, and savings goals. Affordable, open source, and easy to use—comfort and efficiency, simplified.
• Team WinAmp: An intelligent companion app for sustainable eating which makes the healthy choice the easy choice by comparing supermarket offers on healthy food. It uses Llama to create highly personalised meal plans and habit-forming coaching, which is accessible at scale.
• The GoodPath: A platform which creates a roadmap into a socially impactful career – including culturally-relevant explanations, visual diagrams and translations into underserved regional dialects. It has a tutor function which can guide you on everything from how to give strong interview answers to flash cards for exam preparation. 

A Commitment to Open Source

We believe an open approach best ensures AI delivers broad benefits to society. This is why we make Llama free and openly available for anyone to access and download. Since 2023, there have been more than 400 million downloads of all Llama versions, demonstrating how Llama has been a bedrock for AI innovation globally. Since the release of Llama 2 in July 2023, Meta has issued over $2 million in Llama Impact Grants and Awards, fostering innovation and collaboration across various sectors.

Our partners

Our partners for this event included Cerebral Valley, Groq, Nebius, and Neon, who provided inference, compute, and data retrieval support for the participants over the weekend. 

Cerebral Valley is the biggest AI community in the world with 40,000+ of the best developers in the industry.

Groq provided 1,000+ tok/s Llama model inference for hackers. Nebius provisioned each team an H100 VM for compute and model fine-tuning. Neon provided a simple serverless database environment. 

Looking Ahead

The Llama Impact Hackathon is just one of many initiatives under Meta’s Llama Impact program, which includes accelerator programs, training, and workshops. These efforts aim to harness the power of open source AI to solve important local and global challenges, ensuring a brighter future for communities worldwide.

The post Meta’s Llama Impact Hackathon: Pioneering AI Solutions for Public Good appeared first on Meta.