Meta Expands Generative AI Features to Social Media and Teases Invisible Watermarking
Meta has released a host of updates and enhancements to the generative AI features available on its platforms. The nearly two dozen improvements include upgrades to its Meta AI assistant and streamlining the creation and sharing of synthetic images, which will soon come with a way to detect their artificial origin without being visible to humans.
Meta Generative AI
The upgrades to Meta AI stand out as particularly significant. The chatbot now offers more detailed responses on mobile devices and improved accuracy in summarizing search results. The assistant’s reach extends beyond chat functionalities too, as it now assists in powering various behind-the-scenes features on Facebook and Instagram. These include AI-generated post comment suggestions, community chat topic suggestions, and enhanced product copy in Shops.
Meta AI’s role on Facebook includes streamlining user experiences, assisting in creating and editing posts, drafting introductions for Facebook Dating profiles, setting up new Groups, and testing AI-generated images for sharing in Stories. For content creators, Meta is testing AI tools to suggest replies in Direct Messages, aiding them in engaging more efficiently with their audiences. These tools draft potential replies based on the tone and content of incoming messages. The AI is also being utilized to enhance the relevancy of information in Groups and Marketplace, as well as to improve search capabilities. You can see how it works on the right.
On the synthetic media front, Meta has created a text-to-image tool named “imagine with Meta AI,” designed for creative hobbyists, as seen at the top of the page. The imagine tool is incorporated into the realm of messaging through the creatively named “reimagine” feature. Messenger and Instagram users can create, modify, and share images in group chats. Moreover, the integration of Reels into Meta AI chats offers users a more visually immersive way to discover content and plan activities. The imagine tool is available outside of chats as a standalone web experience.
“Meta AI is our virtual assistant you can access to answer questions, generate photorealistic images and more. We’re making it more helpful, with more detailed responses on mobile and more accurate summaries of search results. We’ve even made it so you’re more likely to get a helpful response to a wider range of requests,” Meta wrote in a blog post. “And Meta AI is now helping you outside of chats, too. It’s doing some of the heavy lifting behind the scenes to make our product experiences on Facebook and Instagram more fun and useful than ever before.”
Other AI services within Meta’s ecosystem are also seeing improvements. The company is introducing search functionality to more of its AI services and experimenting with long-term memory features for select AIs, enhancing the continuity of conversations. In a move to ensure transparency and safety, Meta plans to introduce invisible watermarking on AI-generated images. This measure, aimed at increasing the traceability of AI-generated content, will be complemented by the company’s investment in red teaming techniques, including the Multi-round Automatic Red-Teaming (MART) framework, to improve the safety of its large language models.
Meta has been on something of a tear in terms of expanding generative AI’s footprint in its businesses. The company unveiled a stable of LLM-powered conversational AI chatbots for its messaging services not long ago, which is where Meta AI was born. That chatbot employs a version of Meta’s Llama 2 LLM enhanced by some of its recent research. But, the company promised there is more to come.
“We can’t wait for what’s to come next year with AI advancements in content generation, voice and multimodality that will enable us to deliver new creative and immersive applications,” Meta wrote. “We’re still just scratching the surface of what AI can do. Stay tuned for more updates in the new year.”