Harmful content can evolve rapidly and it’s crucial for AI systems to evolve alongside it. @AIatMeta
Harmful content can evolve rapidly and it’s crucial for AI systems to evolve alongside it. @AIatMeta
Uploaded December 2021 | Updated September 2026, 1 hour ago
To address this bottleneck, we’ve built and deployed a new AI technology called Few-Shot Learner that can adapt to take action on new or evolving types of harmful content within weeks instead of months. It not only works in more than 200 different languages, it also learns from different kinds of data and it can strengthen existing AI models that are already deployed to detect other types of harmful content.

Learn more:
Harmful content can evolve rapidly and it’s crucial for AI systems to evolve alongside it.Using AI to deliver more inclusive biographical content on WikipediaDiplomacy Gameplay with CICERO | AI at MetaMeet Llama 3.1Knowledge Distillation with Llama 3.1 405B | Llama for DevelopersBuilt with Llama: Solo Tech | AI at MetaCICERO: The first AI to play Diplomacy at a human level | AI at MetaMilestones to CICERO | Meta AIMaking conversation models more empathicFacebook, Carnegie Mellon build first AI that beats pros in 6-player pokerMake-A-Scene: Mike Schroepfer talks with Oran Gafni about text and sketch promptsCreating real-time depth maps for occlusions in AR
AI at Meta |

Harmful content can evolve rapidly and it’s crucial for AI systems to evolve alongside it.

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER