May 31, 2025 News
Meta Automates 90% of Product Risk Assessments Using AI Systems
Meta plans to use AI-powered systems to automatically evaluate potential harms and privacy risks for up to 90% of updates to its apps like Instagram and WhatsApp, replacing human evaluators. The new system would provide instant decisions on AI-identified risks through questionnaires, allowing faster product updates but potentially creating higher risks according to former executives.
Skynet Chance (+0.04%): Automating risk assessment reduces human oversight of AI systems' safety evaluations, potentially allowing harmful features to pass through automated filters that lack nuanced understanding of complex risks.
Skynet Date (+0 days): The acceleration of product deployment through automated reviews could lead to faster iteration and deployment of AI features, slightly accelerating the timeline for advanced AI systems.
AGI Progress (+0.01%): This represents practical application of AI for complex decision-making tasks like risk assessment, demonstrating incremental progress in AI's ability to handle sophisticated evaluations previously requiring human judgment.
AGI Date (+0 days): Meta's investment in automated decision-making systems reflects continued industry push toward AI automation, contributing marginally to the pace of AI development across practical applications.
Google Launches AI Edge Gallery App for Local Model Execution on Mobile Devices
Google has quietly released an experimental app called AI Edge Gallery that allows users to download and run AI models from Hugging Face directly on their Android phones without internet connectivity. The app enables local execution of various AI tasks including image generation, question answering, and code editing using models like Google's Gemma 3n. The app is currently in alpha and will soon be available for iOS, with performance varying based on device hardware and model size.
Skynet Chance (-0.03%): Local AI execution reduces dependency on centralized cloud systems and gives users more control over their data and AI interactions. This decentralization slightly reduces risks associated with centralized AI control mechanisms.
Skynet Date (+0 days): This is a deployment optimization rather than a capability advancement, so it doesn't meaningfully accelerate or decelerate the timeline toward potential AI control scenarios.
AGI Progress (+0.01%): Democratizing access to AI models and enabling broader experimentation through local deployment represents incremental progress in AI adoption and accessibility. However, the models themselves aren't fundamentally more capable than existing ones.
AGI Date (+0 days): By making AI models more accessible to developers and users for experimentation and development, this could slightly accelerate overall AI research and development pace through increased adoption and use cases.