GPT-3 wrote a college essay that fooled professors

OpenAI's GPT-3, released in 2020, was the first language model large enough to produce genuinely human-quality writing across diverse topics. In experiments, academics failed to reliably identify GPT-3-written essays, and the model passed law school entrance tests in the top percentile. Its successor, GPT-4, passed the bar exam in the top 10% of test-takers. The implication — that white-collar cognitive work is automatable — has prompted economists to revise their predictions about which jobs AI will affect first.

Source: OpenAI Technical Report, GPT-4 (2023)