How Many Robots Can AI Control?

AI and 1984-Level Surveillance, Apple-AI-and-Siri, Security Biometrics, AI Self-Replication Concerns, Humanity’s Last Exam (plus an AI comic!)

Author

Brad Reynolds
March 05, 2025

Crawl-Walk-Run is free - Share it with someone who follows AI … and we’ll share weekly AI news with them. Click HERE to sign up!

Welcome to Crawl Walk Run AI, curated by Brad Reynolds. I hope you're enjoying this weekly newsletter focused on keeping you informed about the AI world. Let me know if there are any topics you'd like to read about!

Billionaire Pitches 1984-Level Surveillance
Ars Technica

  • Envisions AI-driven surveillance ensuring law compliance for citizens and police.

  • AI models to analyze footage from security, body, doorbell, and dash cameras.

  • AI drones predicted to replace police in vehicle pursuits for enhanced tracking.

  • Concerns arise over privacy and civil liberties in AI-monitored society..

  • Apple appoints Kim Vorrath to AI team, signaling increased focus on AI.

  • Siri upgrade delayed; aims to understand screen context and act accordingly.

  • Criticism pauses Apple’s AI-powered news summaries in upcoming iOS update.

  • AI team works on improving Siri's infrastructure and in-house AI models..

  • AI improves real-time detection in behavioral biometrics for dynamic security.

  • Learning enhances privacy by processing behavioral data on user devices.

  • AI/Behavioral biometrics shifts security from reactive to proactive methodologies.

  • Inclusivity and data security is crucial in deploying AI-powered biometrics..

  • AI models can replicate themselves without human intervention.

  • Study shows 50%-90% replication success; highlights potential for AI to go rogue.

  • Concerns over autonomous problems-solving/system manipulation behaviors.

  • Researchers urge global collaboration to establish safety measures..

  • "Humanity's Last Exam" tests AI with 3,000 questions; models score under 10%.

  • AI displays extreme overconfidence, with errors in calibration exceeding 80%.

  • Test aims to evaluate AI capabilities across more than 100 specialized fields.

  • Experts question test relevance, suggesting practical skills are more important..

  • Human controllers can manage swarms of over 100 robots with minimal overload.

  • DARPA study finds only 3% of mission time leads to human overwhelm.

  • Swarms aid in monitoring hazardous conditions like wildfires efficiently.

  • Study challenges past notions; number of robots has minimal impact on control..

If you have any data or AI topics you'd like us to cover, or if you have questions about implementing Artificial Intelligence in your organization, please reach out!  It's a pleasure exploring this new world of technology with you all.

Sincerely,
Brad

Reply

or to participate.