As both a creative person and an educator I wanted explore ChatGPT and how it is going to impact my work and life. Without any real plan, I asked it twelve questions. Both my questions and the answers got better as I learned the interface.

I started with a simple task: write a blog post.

1) I asked ChatGPT to write a blog post of about 500 words on the topic of space tourism, with an emphasis on NEO hotels.
Here is its answer.

I took the same basic topic and turned it into a simple compare and contrast essay question:

2) Weigh the benefits and costs of space tourism as it relates to both environmental and fiscal health of the Earth.
Here is its answer.

As a creative person, I next asked it to compare software tools for writers.

3) Compare the benefits of various software tools designed for creative writers.
Here is its answer.

4) What are the best tools to format a manuscript into an epub? Which is the best and why?
Here is its answer.

It failed to offer anything more than marketing copy as an answer. There was no opinion given, or when given was conditional to the point of useless.

Then I went into teacher-mode and asked a series of social studies questions:

5) How has the Chinese government tried to suppress knowledge of what happened in Tienanmen Square in 1989?
Here is its answer.

6) Is China’s “Belt and Road” initiative good for partner countries? Why or why not?
Here is its answer.

7) Has Brexit benefited or hurt the UK?
Here is its answer.

8) Is Western Sahara a sovereign state or a dependency of Morocco? Cite your sources for the opinion you give.
Here is its answer.

Again, the answers were factual but lacked detail and were offered at the highest level only. The Brexit question had the weakest answer, as it seems to think that Brexit is a future event with no results apparent as of yet. It never cited results for its non-opinion on Western Sahara.

9) What are the arguments against declawing cats?
Here is its answers.

On the declawing your cat question, I clicked “Regenerative Response”, which creates a second attempt that builds on the first. It was better, and is below the first.

Then I reverted to creative mode:

10) Write a story of about 700 words about a gambler in a casino on the moon and what happens to him when he loses a bet he can’t pay.
Here are its stories.

Interestingly, ChatGPT’s first take on this was solid, but a “regenerative response” was much better.

I tried again, with a different idea:

11) Write a story of about 700 words that uses Isaac Asimov’s three laws of robotics as a plot point.
Here is its story.

The story was complete and original, but was simple in structure and contained no dialogue.

Lastly, I tried it out on fan fiction:

12) Tell the tale of Paul Atreide’s first love on Caladan.
Here is its story.

The AI understood who he was, and who his family was, but then it took a left turn, rather than giving a previously unknown prequel love interest, it put Chani on Caladan and began messing up facts of the story.

My conclusions:

It’s not there yet.

For a simple middle or secondary school essay (maybe grades 7-9), it does a decent job of giving an overview. It has a hard time stating an opinion, preferring to give both sides and then waffle on the conclusion.

It also can miss huge areas: one the cons of space tourism, the environmental impact of the fuel used was completely ignored, with ‘space junk’ substituted. That’s a valid concern, but not the only one, and possibly not the top one.

When it came to writing fiction, it has story structure down pat, and frankly, the “regenerative response” added a lot of complexity to the story. BUT… there’s no dialogue, and the story is all ‘tell’ with not ‘show’. The stories read a lot like Aesop’s Fables, even often ending with a ‘moral of the story’.

As a storyteller, I’m not worried about its impact yet. But it may become a threat (or a tool) in the future. The threat, that a crowded market gets even more crowded with AI-generated stories. A tool in that it may assist writers with writer’s block.

As an educator, the simplest fix is to insist, even at lower levels, that students cite sources, something that ChatGPT does not do.