Sign up to save your podcasts
Or
Share E16: Technology Thursdays | Invitation Required: Why Robots Need Your Explicit Permission to Speak
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
E16: Technology Thursdays | Invitation Required: Why Robots Need Your Explicit Permission to Speak
Welcome to Technology Thursdays on the Deep dAIve podcast!
In this episode, we break down a compelling robotics and AI safety paper from the University of New South Wales. While developers typically think of robot safety in terms of collision avoidance (motion) or content filtering (dialogue), this paper highlights a major missing dimension: Initiation Safety. Just because a generalist humanoid or smart assistant sees a human doesn't mean it has corporate or social permission to interact with them. We look at why relying entirely on a high "engagement score" or a confident neural rollout can lead to invasive, hard-to-undo social actions—like an uninvited grasp or stepping directly into your personal space. We also look at the newly proposed PAS (Probe–Authorize–Speak) protocol built to change how AI bots cross social boundaries.
Read the Original Research Paper:
Conference/Preprint: 2026 Human-Robot Interaction Reference Track
Article Title: Initiation Safety: A Missing Dimension in Generalist-Robot Safety
Authors: Zhijin Meng and Francisco Cruz (University of New South Wales)
View all episodes
By Deep Daive PodcastWelcome to Technology Thursdays on the Deep dAIve podcast!
In this episode, we break down a compelling robotics and AI safety paper from the University of New South Wales. While developers typically think of robot safety in terms of collision avoidance (motion) or content filtering (dialogue), this paper highlights a major missing dimension: Initiation Safety. Just because a generalist humanoid or smart assistant sees a human doesn't mean it has corporate or social permission to interact with them. We look at why relying entirely on a high "engagement score" or a confident neural rollout can lead to invasive, hard-to-undo social actions—like an uninvited grasp or stepping directly into your personal space. We also look at the newly proposed PAS (Probe–Authorize–Speak) protocol built to change how AI bots cross social boundaries.
Read the Original Research Paper:
Conference/Preprint: 2026 Human-Robot Interaction Reference Track
Article Title: Initiation Safety: A Missing Dimension in Generalist-Robot Safety
Authors: Zhijin Meng and Francisco Cruz (University of New South Wales)