Let’s Reason About (Your) Job Security!

Analyst Gergely Máté has published a detailed framework on LessWrong for assessing individual job security in the face of advancing AI. The core argument begins with a first-principles economic model: if an AI system can achieve a 50% success rate on an 8-hour workday task at a cost lower than a human wage, an employer becomes financially incentivized to automate. At a 99% success rate, entire departments could be reduced to a single human overseer. The model accounts for real-world adoption factors like upfront transformation costs, workplace social dynamics, and unionization, which can slow the process.

The framework's innovation is a five-dimensional decomposition of any job to compare human and AI capabilities. The dimensions are Cognitive Processing (thinking, planning), Physical Execution (movement, dexterity), Social Interaction (communication, emotional intelligence), Sensory Perception, and Environment Adaptability. By scoring AI vs. human performance in each dimension relevant to a specific role and weighting the scores, one can estimate overall automation risk. Máté acknowledges limitations, including the exclusion of pure cost/benefit ratios and intrinsic human value in certain fields (like professional chess), but provides a structured method for moving beyond vague anxiety to data-informed forecasting.