Convergent Robotics

Building the foundations of safe physical AI:
🔴 Red-teaming intelligent robots
⚒️ Building a safety layer for physical AI
🔗

[Convergent Robotics @convergentrbtx]

To address this, we will red-team frontier physical AI on real robots, in scenarios drawn from real threat models, and build a safety layer for physical AI. We want the future with intelligent robots to be safe. Let's collaborate!
💡Blog post: https://t.co/BcSw3UEIN9

[Convergent Robotics @convergentrbtx]

We are launching Convergent Robotics, an independent research lab for physical AI safety. We think robot capabilities are improving faster than the technologies we need to make them safe.
It is time to build the foundations of safe physical AI.
https://t.co/HOPhk1e6VJ
🧵

[Convergent Robotics @convergentrbtx]

Advanced physical AI also poses severe sociotechnical risks: By blindly trusting opaque, powerful AI systems with most physical tasks in our lives, we risk gradually disempowering ourselves. Robots that empower people require new alignment and control technologies. 6/7