Predicting AI job exposure

Short summary

Benedict Evans argues that predicting job exposure to AI is fundamentally impossible because job definitions are fluid, external factors are unpredictable, and work itself can't be objectively measured. The piece challenges frameworks attempting to score automation risk by highlighting these epistemological limits.

• •Job definitions evolve; you can't predict how roles will change under AI
• •Second-order effects and external factors make forecasting unreliable
• •Work as a concept resists the measurement automation risk scores require

Generated with AI, which can make mistakes.