IEEE Spectrum breed · Technology

The Shift to Governing AI

As AI capabilities advance, the focus for professionals is shifting from building AI systems to governing them. This involves defining their operational boundaries, decision-making autonomy, and ethical constraints.

Many organizations struggle to achieve profit from generative AI, a 'GenAI Divide.' This gap stems from a shortage of personnel to direct AI systems and validate outputs.

Professionals should recognize 'human middleware' roles. Focus on judging data significance and risks, which AI agents cannot perform.

Replace rigid rules with guiding principles. Prioritized principles allow AI to resolve conflicts autonomously, aligning with company values.

Embed company culture into AI governance as machine-readable code, including a 'constitution,' 'doctrine,' and 'playbook.'

Use a 'trust thermostat' approach. AI decisions have confidence scores; they proceed autonomously above a threshold or escalate for human review.

Effective AI governance requires fixing 'context' issues, enabling AI to integrate disparate information for a comprehensive understanding.

Lead by exception. Focus on AI-identified ambiguous or high-stakes cases, allowing scalable growth without proportional headcount increases.

AI-samenvatting op basis van de bron.