VWA with Chat-GPT?

ChatGPT also makes directly detectable errors, because it basically does not understand the topics, only combines texts. This was largely very successful. You don’t recognize any errors, but the examiner probably does.

Even now, many such texts already recognize or at least have a suspicion. There are typical features, rafting, repetitions, phrases like “don’t go home” and certainly others. A comparison with other texts of the same author was a further certainty.

What brings the future is difficult to predict. Certainly, progress has also been made in KI-KI recognition, but the extent can be badly assessed.

One can certainly also develop defence mechanisms, a Stealth-KI. In the end, each test subject or author is enclosed, cut off from any communication.