After Watching the Trolley problem video from The Answer in Progress sometime ago. I'd got this idea of asking ChatGPT to solve that. Unfortunately, I'd put this project in my to do list, but since I'd never thought on this subject again.
However, recently I came across with this game on steam:
Which gave me a lapse of memory, and brought me urge to do the Trolley GPT project again.
During the talk with ChatGPT, It was nice to see that the most part of his answers (except from some confunsion made in some parts of the talking) are pretty based on what the general population could do in those situations, even when I asked him to think out of the box, he gave answers that we could give in his place.
Actually, such behavior is exepected, once he was trained with a lot of online texts, and, for sure, this kind of response was intrinsically embeded in the training data.
Anyway, If you want to check my results, you can find them in the conversation.md file.