The other day, I decided to have a chat with the AI chatbot that has recently caused such a stir, ChatGPT.
The topic of our chat was the history of psychology. Many of the answers were pretty reasonable. A good number were problematic in terms of accuracy. Some of the answers were bizarre and outrageous!
Don’t be daunted by the size of the document. The paper proper is just the first 11 pp. The 40-some pages after that are a complete transcription of the “conversation,” for those who are interested.
I should point out that ChatGPT’s answer to the same question can change from moment to moment. Where I got bizarre answers, some others have gotten reasonable ones. One of the oddest things I found is that, when ChatGPT does not have access to correct
information, rather than simply responding “I don’t know,” it sometimes “confabulates” faintly plausible but totally fictional answers.
I should also point out that my understanding is that ChatGPT was never intended as a universal-true-content-provider (which, unfortunately is how many people have been using it). It was developed to produce fluid English-language answers to questions
that are put to it. It does that pretty well. But the content is often far from reliable.
Best,
Chris
Editor of the journal History of Psychology
…….
“Civilization is in a race between education and catastrophe.”
- H. G. Wells
---------------------