Twenty-five years of research into complex systems shows why artificial intelligence will always produce errors in healthcare decisions, regardless of technological improvements or funding.