GPT-4

GPT-4 was released by OpenAI on March 14, 2023, as the fourth generation of their Generative Pre-trained Transformer series. It arrived roughly sixteen months after GPT-3.5 powered the launch of ChatGPT, a period in which public expectations for large language models had been radically reset. Unlike its predecessors, GPT-4 was announced as a multimodal model capable of accepting both text and image inputs, though image input remained accessible only to a limited set of API partners at launch, with broader rollout following later in the year.

GPT-4 mattered because it moved AI performance from 'impressive but unreliable' to 'reliably competent across a broad range of expert-level tasks.' The bar exam result was symbolic but also precise: the 90th-percentile score meant GPT-4 would have passed the licensing threshold required to practice law in most U.S. jurisdictions. Similar scores appeared across medical licensing exams, coding competitions, and graduate admissions tests. For the first time, the question of whether AI could substitute for credentialed human professionals in narrow, well-defined tasks became not theoretical but empirical.