EXAM CHEATING
Crack bots
DESPITE evidence that use of AI cheating by students is rife and human markers struggle to spot it, many universities are refusing to use AI detection tools.
An experiment at the University of Reading last year saw 63 essays and assessed answers that were fully written by GPT-4submitted to markers alongside student work in five different undergraduate modules. Only four were flagged as concerning, only two of which were identified as possible AI. The answers from GPT-4, a creation of OpenAI, averaged slightly higher marks than real students.