GPTs and Large Language Models (LLM)

• What does GPT mean?
• What is a language model (light technical answer)?

Reliability

• How does the model account for essays written by a group?
• Will the author verification scores reflect that a student’s writing improves over time?
• Writing style differs between subjects, e.g. a poem analysis and a technical lab report — do we need authentic text in all subjects?
• How reliable are the results from the author verification model?
• What is the precision of an AI detection tool?
• What is accuracy, and why is this not a good measure for an AI detection tool?

AI writing

• Can students alter a GPT or train their own language model to mimic the student's original writing to circumvent the authorship verification method?
• Will future improvements in language models be able to circumvent the authorship authentication method?
• Will a language model on the student's local PC circumvent the authorship verification method?
• When does the authorship authentication method not have to be in use?
• What is the difference between AI for paraphrasing and AI writing?
• Why is AI writing the same as the age-old problem of ghostwriting?
• What are the mechanisms of AI detectors and why can they be fooled?

Data & Privacy

• Where can we get access to the necessary training data for these models?