Epoch report says frontier AI still falls short on human judgment and creativity

Epoch Automation Reports, an AI research organization, has published a report saying frontier AI models perform strongly on structured tasks but still struggle with open-ended work. The finding points to a clear limit in current systems: they are not yet able to replace human judgment and creativity. The item was carried by Techub News, with Crypto Briefing cited as the original source. The report’s core conclusion is narrowly framed around task performance. It does not say frontier models are weak across the board. Instead, it draws a distinction between environments with defined rules and outputs, where AI performs well, and less structured work, where the models continue to face difficulty. That gap, according to the report summary, highlights why human decision-making and creative capacity remain difficult to substitute.