Microsoft Executive Calls for Emergency Brake for Advanced Artificial Intelligence
Microsoft CEO Satya Nadella proposed an emergency brake mechanism that grants the authority to stop advanced artificial intelligence models mid-task.
Microsoft Chairman and CEO Satya Nadella stated that organizations deploying advanced artificial intelligence models should treat these systems as if facing insider threats, advocating for the establishment of emergency brake mechanisms that grant authority to halt them mid-task.
Insider Threat Approach to Artificial Intelligence
Microsoft CEO Satya Nadella stated that institutions utilizing advanced artificial intelligence models must take into account the possibility that any model could be compromised.
Emergency Brake Mechanism
Nadella emphasized that authorized personnel should possess emergency brake-like mechanisms to ensure a model can be stopped or shut down at any moment during a task.
Governance and Observability Principles
Stating that organizations cannot rely solely on the assurances of artificial intelligence providers, Nadella proposed a new governance approach that includes utilizing different models together, human-readable logs, and third-party testing.