Why Learning Still Matters in the Age of AI

Summary

The essay begins with the author’s unsettling interaction with ChatGPT, which suggested that learning addition may be unnecessary when calculators are available. It uses that exchange to examine whether machines can possess knowledge or merely produce accurate-looking outputs. The author defines knowledge as access to truth supported by justification, rather than a lucky or correct prediction. Because justification depends on belief, self-conviction, social validation, and the ability to assess how a conclusion was reached, the essay argues that knowledge is fundamentally human; it also acknowledges that machines might someday develop a self, but asks readers to assume current machines do not. The practical argument focuses on reliability: opaque AI systems make it difficult to know why a claim was produced, while stochastic behavior can undermine confidence even when a system has a strong track record. Full delegation of decision-making would therefore require either better observability into an AI’s process or testing every claim, neither of which the author considers available. The proposed middle ground is to use AI as a source of ideas while keeping humans responsible for verification and deciding what becomes accepted knowledge. The essay concludes by urging readers to keep learning mathematics, read code, and remain curious rather than surrendering intellectual judgment.