Yhn TechnologyRobotics first seen 9 h ago, last 17 min ago, peak #4
Roboharm benchmarks test whether robots refuse unsafe commands
Original: Roboharm: Do frontier robot policies refuse unsafe instructions?
A project called Roboharm is asking whether frontier robot policies, the AI models now being wired into embodied machines, actually refuse unsafe instructions. It examines how large models behave when given harmful physical commands, a question that matters as robots move into homes and workplaces. The work is drawing attention among AI safety and robotics researchers discussing whether current safeguards transfer to the physical world.
Why now: As AI models are deployed in robots, researchers are worried they may follow harmful physical instructions instead of refusing them.
Roboharmfrontier robot policies
Evidence
- Roboharm: Do frontier robot policies refuse unsafe instructions? · msadowski · 60
API: https://socialmediatrends-api.osmike.com/v1/trends/3957