Got a bit of time with #ChatGPT #GPT4. Asked it for the story described in the image text / alt text. Story image follows in next toot, then a sketch of the cause-effect diagram of the story created. Full chat thread here: https://chat.openai.com/share/d67b2962-e03b-49dd-9583-e728c90da6d4 @tomstafford 1/3
ChatGPT

A conversational AI system that listens, learns, and challenges

Image of the resulting story. Text too long for alt text, here's the full chat thread. Sketch of cause effect in next toot. https://chat.openai.com/share/d67b2962-e03b-49dd-9583-e728c90da6d4 2/3
ChatGPT

A conversational AI system that listens, learns, and challenges

And sketch of the cause-effect paths in the story. It answered the prompt pretty well huh? 3/3 #ChatGPT #gpt4
p.s. I tried to get it to write an even more elaborate construction but it just made something very similar again. Only one attempt, will press on...
Also asked it to explain the causal structure it created: "Could please explain what the causal structure of the story is? In doing this, can you please separate out the linear cause-effect chains from any cause-effect elements that jump across from one chain to another? I would appreciate any other insights you have into how best to describe the causal structure of the story." See image for (accurate) response or here for full chat text: https://chat.openai.com/share/d67b2962-e03b-49dd-9583-e728c90da6d4 #ChatGPT #gpt4
ChatGPT

A conversational AI system that listens, learns, and challenges

More on #gpt4 #ChatGPT causal chain stories. In its last response describing the causal structure of its story, it said (correctly) there are "two distinct linear cause-effect chains that are interconnected at two points, forming a kind of 'H' shape if visualized." I was quite taken by this idea & asked it to do a few further things based on this idea. Starting with: "Work out which letters in the alphabet you think could be used as diagrams to create causal stories & which ones can't?" 1/x
I then asked it (not the most well-formed question, but it managed): "Could you categorise the letters that do work as causal story structures into whatever you think are the best small number of categories, for those that have a similar causal flow?" Its response in the image / or read the full thread here. Particularly like its response about how M could fit into more than one category. https://chat.openai.com/share/d67b2962-e03b-49dd-9583-e728c90da6d4 2/x
ChatGPT

A conversational AI system that listens, learns, and challenges

Then asked it to write a detective story based on a criminal spelling out a message using crimes with causal structures based on these letters. Not convinced it quite managed it, but it's a pretty good attempt. See the full chat thread for this and subsequent attempts at similar stories: https://chat.openai.com/share/d67b2962-e03b-49dd-9583-e728c90da6d4 #gpt4 #ChatGPT 3/3
ChatGPT

A conversational AI system that listens, learns, and challenges

p.s. I haven't yet tried the same with gpt3.5, would be worth checking to see if / what the difference is.