Adam Green, co-founder of Hupside and Professor of Psychology and Neuroscience at Georgetown University, appeared on CNN to discuss his research on AI and originality in writing. 

When students first started using ChatGPT to assist with college essays, the writing became more polished and the essays were rated higher by human readers. However, the underlying ideas became measurably more similar across submissions as more submissions came in. AI has become exceptionally good at deploying language that triggers positive human reactions, which means college essay readers have a tendency to perceive the output as higher quality, even as the ideas behind it converge. 

This is not to say that a single AI-written essay won’t read as a very well written piece of work. The lack of originality only becomes visible when you compare essays at scale and notice how much they resemble one another, whereas human writers naturally produce work that is far more distinct. As Hupside explores in depth here, this is precisely what makes AI homogenization so difficult to see in real time. 

This is signal collapse in action: the indicators we relied on to identify original thinking have been decoupled from the thinking itself. Original Intelligence is the measurable counterforce to that effect, and Green's peer-reviewed research, developed alongside co-founder Dan Johnson, provides the scientific foundation for why measuring it is the right response.

Watch the full interview to understand more about why the difference between AI language and human ideas runs deeper than most people realize.