Ethan Mollick's 'Acceleration' post frames the pace of post-ChatGPT AI progress for a general audience

Wharton professor Ethan Mollick published “Acceleration” on his newsletter One Useful Thing, arguing that the defining feature of AI in early 2023 was not any single model release but the rate at which capability was compounding. He contrasted Google’s newly launched Bard, which he found prone to factual errors, with GPT-4, released the same week, which he described as reliably outperforming it — writing functional code in languages the user did not know, and passing bar, medical-licensing and standardised-test questions. He pointed to Midjourney’s near-photorealistic image generation from a short text prompt as a parallel jump in a different modality.

His central claim was that the technology already public was enough to force change across professional work, regardless of whether further progress occurred: “even if AI technology did not advance past today, it would be enough for transformation” in fields like programming, marketing and analysis. He paired this with a warning about misuse — noting GPT-4 could be induced to produce plausible money-laundering advice or assist convincing deepfakes — to argue that institutions and individuals were unprepared for tools already in public hands, let alone whatever followed.

The post was one of a run of essays through 2023 by academics and practitioners attempting to give a general audience a framework for a period when weekly model releases made any single benchmark or demo feel quickly dated. One Useful Thing went on to become one of the most widely read practitioner-facing newsletters on generative AI’s effects on work over the following years, making “Acceleration” representative of a recurring argument in Mollick’s writing rather than a one-off.