Meta rolls out Generative AI experiences across its apps

At the recent Connect 2023 event, Meta unveiled exciting AI innovations across their apps and devices.

They introduced 28 distinct AI personalities, each with its own interests, for users to engage with. These AIs offer a deeper dive into your areas of interest, complemented by new generative AI features.

Social Profiles for Meta’s AIs

Starting today, you can chat with these 28 AIs on Instagram, Messenger, and WhatsApp. They also have social profiles on Facebook and Instagram, allowing you to follow and learn about them.

These profiles feature recognizable figures representing the AIs, though they symbolize the characters, not the individuals. Meta manages the creation of social content generated by AI, marked with #ImaginedWithAI and a watermark.

If you’re part of the U.S. beta rollout, you can message the AIs from their social profiles on Instagram or Messenger. Their responses are powered by Meta’s AI technology.

AI Stickers for Chat Customization: Meta has introduced AI stickers for creating personalized chat and story stickers, providing more expressive options to users.

AI Image Editing: Soon, users can transform and co-create images using AI through features like Restyle and Backdrop on Instagram. Markers will indicate AI-generated content for transparency.

Meta AI Assistant: Meta AI, available on platforms including WhatsApp, Messenger, Instagram, and smart glasses, offers real-time information and image generation in text chats.

Meta is launching AI Studio, allowing developers and non-coders to create AIs for messaging services. Businesses can create brand-specific AIs for better customer engagement, and creators can expand their virtual presence with controlled AIs. A sandbox for AI experimentation is also in the works.

Responsible Development of Generative AI Features

Responsible development includes safeguards like notices, integrity classifiers, and alignment with industry best practices outlined in the Llama 2 Responsible Use Guide.

Meta conducts rigorous testing, collaborates with experts, and seeks feedback to ensure safety and reliability. They plan a gradual feature rollout, keeping an eye on how people incorporate them into their daily lives.

Prioritizing Safety and Privacy

Meta emphasizes safety and privacy in AI development. Custom AI models, like Meta AI, build upon the foundation of Llama 2, incorporating safety and responsibility training. Meta actively evaluates and fine-tunes models, trains them on guidelines, and strives to reduce bias.

Feedback mechanisms, algorithms to filter out harmful content, and collaboration with security researchers contribute to ongoing improvements. Privacy protection is paramount, and Meta does not use private messages for AI training.

Promoting Awareness and Transparency

Meta ensures users understand when they’re interacting with AI and its limitations. They publish “System Cards” to explain AI decision-making and provide interactive demos. They also implement visible markers on AI-generated content, making it clear that AI created it.

While working toward industry standards for labeling AI-generated content, Meta is actively combatting misinformation with AI technologies. Tools like Few-Shot Learner can swiftly adapt to address emerging harmful content in multiple languages, reducing the need for extensive data sets.

  • AI stickers are rolling out across Meta’s apps, and AI editing tools, Restyle and Backdrop, are coming to Instagram.
  • Meta AI is available in beta in the U.S., offering advanced conversation and image generation capabilities.
  • Meta is also introducing 28 more AIs in beta, each with unique interests and personalities, some represented by cultural icons and influencers.

Additionally, Meta has plans to make AIs accessible to businesses and creators and release AI studio for broader use in the future.