An essay argues that AI cannot eliminate uncertainty in an inherently random world, challenging the narrative that AI will automate away most jobs.

What happened: A data scientist reflects on a decade of experience with machine learning and argues that despite AI's sophistication, real-world systems remain fundamentally probabilistic and unpredictable. The author notes that in a supply-chain optimization exercise, a simple rolling 3-week average outperformed advanced algorithms like XGBoost, because real data is often noisier than pure trend.

Why it matters: The author contends that the "AI will replace humans" argument rests on a flawed assumption—that the universe is deterministic and fully explainable with better algorithms. Quantum mechanics and empirical evidence suggest otherwise: AI surfaces noise faster but cannot eliminate it in an inherently probabilistic world. This means many feared disruptions may not materialize as predicted.

What to watch: Jobs involving rule-based, digital systems (data entry, scheduling, accounting, roles with more than 80% repeatable patterns) face the highest automation risk. By contrast, roles operating in probabilistic spaces—where decisions must be made amid uncertainty—will likely see AI as a tool rather than a replacement, making the ability to navigate uncertainty the most valuable skill to develop.