Researchers at the University of Maryland, Microsoft, and UMass Amherst find regular ChatGPT users accurately identify AI-generated text after labeling 300 non-fiction articles from models including GPT-4o and Claude 3.5 Sonnet
AI recognition becomes intuitive with exposure but hard to demonstrate.
#175Ethan Mollick@EMOLLICKI broke my own rule to never post about AI detection as it is fraught in many ways.
The problem is that if you use AI a lot, you know AI writing on sight, which makes the difficulty of objectively proving that AI use to others very frustrating
@nabeelqu @johnrobertgage @GrantaMag I had the exact same thought, pulsing quietly...
@nabeelqu @johnrobertgage @GrantaMag Just for fun, I wouldn't put much stock in this (and I needed to pad to get to the 50 word min)...
Michael Nielsen@michael_nielsen@nabeelqu @johnrobertgage @GrantaMag I had the exact same thought, pulsing quietly...
#501Lucy Li@LUCY3_LI@emollick @jennajrussell !!
Ethan Mollick@emollickI broke my own rule to never post about AI detection as it is fraught in many ways. The problem is that if you use AI a lot, you know AI writing on sight, which makes the difficulty of objectively proving that AI use to others very frustrating
