AI Struggles With Simple Focus Tests

AI models like GPT-4o and Claude dazzle with short, simple tasks—but crumble under pressure when faced with longer, more complex color-word tests, dropping accuracy from 91% to just 15%. While humans rely on executive control to filter distractions, AI defaults to reading words, unable to consistently follow instructions. Similar failures across models like GPT-5 and Gemini reveal a fundamental gap: AI mimics human skills but lacks the sustained focus and cognitive control that make us resilient under pressure. Support the show:Get a discount at https://solipillow.com/discount/dnn. Advertise on DNN:[email protected] This is an automated, high-level news summary based on public reporting.Report issues to [email protected]. View sources & latest updates:https://sources.thednn.ai/e7c0bfba5f004936