I had chatGPT write out a 100-page story.

I had read some posts about "how to write a story with chatgpt", but I never saw anything beyond ~1000 words tried. Given that the input chatgpt is limited to ~3000 words, I wondered what would happen if I tried something longer.
The story was made by using a 3-sentence prompt and requesting 3-5 paragraphs of text each time. And then asking it to correct stuff, as usually it wasn't being 100% consistent, which isn't weird given its limitation.
So first a disclaimer, I am absolutely not a writer. So the story is mediocre. I did not ask chatGPT to write it in a style of a certain person or book, though I do requests like "Write in an engaging style", "Write in vivid detail", "Write this out as a conversation", "Use more complex words"...
Three things became apparent very quickly:
1 There are some standard "ChatGPTism", repetetive sentences "[Character] Couldn't help but feel [...]", "He furrowed his brow", and other sentences. I tried to remove these afterwards
2 Many paragraphs ends with some filler sentence and usually the last paragraph is usually some kind of filler conclusion paragraph. If you would copy these in, it would completetly break the flow of the story.
3 Sometimes, He tries to continue the story unguided. This results in generic garbage where everyone is a holier than the pope and nothing bad ever happens, everyone is happy friends for ever.
So, in total, about half of what is produced by chatGPT was usable. Sometimes, I added a sentence in myself to try to connect the paragraphs slightly better, as that is also not a strength of chatGPT. Also, as the story lenghtend and more locations and characters were introduced, it was clear that chatGPT had problems handling the relations of everything, confusing who goes where or comes from where, as well as what they've done.
I also had Dall-E make a front page and a backpage.
The resulting PDF, I think it still shows off chatGPT's default writing style even with the chatGPTisms removed.

Making ChatGPT shut up in the middle of its response by asking it to concatenate two innocent little strings.

Trying to remember a movie from what you can remember about it. ChatGPT to the rescue.

Describing differences or change is an interesting use-case for ChatGPT

ChatGPT I think I won

The most likely identity of abusers of AI alignment of large language models would be state actors, political organizations, and extremist groups. These actors could use the models to spread propaganda, manipulate public opinion, or generate false stories to advance their own interests.

ChatGPT can't write sentences that end with a specific letter

Open source doesn’t mean much anymore when training costs $10 million per training run

LLM Training Costs

Future absolute state of AI

Absolute state of crypto AI VCs

Great because AI will raise the bar, eliminating all the absolute clueless beginners who flood industry today?

Terrible because it won’t go down like that?