Meet Zeynep Demirbas, the New York eighth-grader who tested whether AI can recognise stress; a basic machine-learning model beat ChatGPT-4o
Zeynep’s work earned her a finalist place in the 2025 Thermo Fisher Scientific Junior Innovators Challenge Fourteen-year-old Zeynep Demirbas has found that ChatGPT-4o is less accurate than a mental health-focused AI model and a simpler machine-learning system at detecting stress in human written text.Zeynep, an eighth-grade student at Transit Middle School in East Amherst, New…