The report was tabled at the UN General Assembly focusing on AI agents, misalignment and the risk of losing human control.
About Agentic AI
- Meaning: It is an artificial intelligence system that can accomplish a specific goal with limited supervision.
- It is different from traditional software that follows pre-defined rules and guidance.
- Advantages: Autonomy, Specialisation, proactivity, adaptability, etc.
- Applications: Analysis of live stock prices to execute trade; navigation through real time GPS in autonomous vehicles, adjusting treatment on monitoring patients’ health data, etc.
Key Issues of AI systems
- Misalignment: Occurs when AI system pursues a goal that conflicts with the actual intentions.
- E.g. Hacking computers to gain unwanted access to the internet.
- Loss of control: It is the human’s inability to reliably direct, constrain, or stop an AI system.
- Self-reinforcing: Refers to escalating behaviours in an unintended direction. E.g., content moderation AI designed to reduce harmful speech over-censors legitimate discussions.
Way Forward
- Accountability Across Jurisdictions: Connect national accountability mechanisms with international coordination rather than assuming same legal instrument to work everywhere.
- Incident Reporting and Shared Learning: Organizations and countries to share incident information or promote whistleblowing by employees.
- Independent AI Monitoring: Automated monitoring by a separate AI model could provide an additional layer of control for detection of deliberate misbehaviour.
