I've been amusing myself with ChatGPT.

[Ocelot II]

I asked it to write a poem about a zombie who doesn't want to eat brains but prefers broccoli, and I got this:

In the realm of the undead, a zombie roams,
A creature of darkness, but with tastes unknown,
While others crave brains, with a ravenous desire,
This zombie, unique, has a different fire.

For broccoli, not brains, is what it seeks,
A veggie delight, it hungrily sneaks,
In the moonlit night, it stalks the greens,
With hollow eyes and rotting seams.

No appetite for flesh, but for the verdant fare,
It craves the crunch, the leafy affair,
Through haunted gardens, it silently creeps,
Feasting on broccoli, while the world sleeps.

Its zombie brethren mock, with disdain and dread,
"Why broccoli, dear fiend?" they question with dread,
But this zombie, defiant, holds true to its taste,
A peaceful choice, a deviation embraced.

In the midnight hour, amidst eerie moans,
The zombie munches on broccoli, bones to be sown,
For even the undead can find a new way,
To nourish their bodies, in their own undead sway.

So let us celebrate this zombie's delight,
Embracing uniqueness, breaking from the night,
For even in the darkest places, new preferences may bloom,
And even the undead can find joy in a vegetable's room.

[delisen]

1. PrettY good.

[LetMyPeopleVote]

2. LOL

[highplainsdem]

3. Yes, ChatGPT will spew out what it's told to. RWers also find it useful

for things like Dr. Seuss-style poems about how wonderful guns are. See this (reply 22):

https://www.democraticunderground.com/100217753275#post22

We could easily fill DU with text from ChatGPT - text produced mindlessly by algorithm, using a data set of text taken without permission or compensation. We could also post endless messages with AI's fake art, produced thanks to real art and photos that were ripped off.

You could generate a ChatGPT response to any post on DU in seconds.

But none of it would mean anything. It's spewed out by a machine. And if you gave the exact same prompt that produced that text to ChatGPT in another session, it wouldn't be able to provide the same response.

Ocelot II, you write fine poetry of your own - https://www.democraticunderground.com/10181815377 .

And it's YOUR writing ability, your own thoughts. Which are much more interesting than anything ChatGPT could produce.

I'll bet YOU could have written a poem of your own about a zombie who doesn't want to eat brains but prefers broccoli.

It's an amusing idea. It deserved YOUR words, your own poem.

I would've liked to read that one.

As for that ChatGPT-written poem I found on Twitter, a Dr. Seuss-style poem about using guns against liberals and government officials - and their mothers - it's a vicious misuse of Dr. Seuss's style, by someone who probably couldn't write a poem on his own.

But ChatGPT will produce anything those RWers want, making them seem more literate than they are.

[Ocelot II]

4. I just wanted to see what it could do. I asked it to write all kinds of other, absurd things,

which it did, and they were funny but clumsy. They would be what you might expect from a clever high school student but they were not artful or literate at all. It will have to get a lot smoother before it can produce anything like actual literature, but if one wanted to write pop songs it could turn out lyrics that might pass muster with a mass and not particularly discriminating audience. I've been having fun, though, trying to tie it in knots by making it write completely surreal poetry. And yes, I could do a much better job.

[highplainsdem]

5. It's already being used to write pop songs, including in the style of various

artists, and then their AI-generated voices are used to sing those lyrics.

https://www.democraticunderground.com/10355945

https://www.democraticunderground.com/103493633

I can guarantee you that poetry markets, like short-story markets, are being flooded by AI-written submissions. And AI poems are also being turned in as supposedly student-written assigned writing. By kids who are NOT learning anything about writing.

I don't even want to think what's happened to sales of children's books.

Artists - real artists - have to fight back.

AI is one reason WGA is on strike.

I've seen articles on commercial artists whose income has been cut to a fraction of what it used to be by image-generating AI - which copies those artists' styles.

I know these programs can be entertaining. I was playing with an image generator the other day, a brand new one requiring no registration, and it was interesting to see it come up with 4 different images in seconds from one prompt.

But the more I looked at the better quality images, the more I wondered what real photo or human-created art it was mostly borrowed from.

In any event, none of it was mine.

And none of it was really under my control. Any more than the images you see in a kaleidoscope toy are. Those are entertaining, too, for a while.

ChatGPT's non-creative writing is causing a lot of harm, too. I've seen so many articles and social media posts from writers who've suddenly lost most or all of their writing income. And ChatGPT is being used for academic papers about subjects the papers' authors have no real expertise in.

AI detectors don't work well, so you not only have frauds and cheaters getting away with it, but people submitting their own genuine writing being accused of using AI. Those detectors can give very different results analyzing the same text in different sessions.

IMO it does much more harm than good.

And we're really going to see how much harm AI does as we get closer to next year's election.

Which is why I wish it had never been developed and released. It will make a few people and businesses much wealthier, and a lot more much poorer. I don't think AI providing entertainment while substituting for real artistic creativity, or real writing ability, is worth the harm it does. Including to all the people whose work was ripped off to train the AI.

[Ocelot II]

7. I hope someone comes up with a way of identifying it.

When I was teaching I was very good at identifying written work that could not have been created by the student. I warned them that they'd better not plagiarize because I'd know, and they'd get no credit for the work (or worse). Now I wonder whether I could identify AI work; based on what I've seen, maybe I could; it's pretty clumsy. I just worry that it will get better.

[highplainsdem]

8. If you look through the ChatGPT subreddit or forum on Reddit, which I've

posted a thread from in the other reply I just posted.- here's the main link https://www.reddit.com/r/ChatGPT/ - you'll see lots of posts from people about how bad the AI detectors are. They're still being used, and new ones are constantly being marketed, because teachers and school systems are desperate. But none of the detectors work worth a damn. I've seen people on Reddit advise anyone who didn't use AI, but is accused of doing so, to tell whoever made the accusation to run their own writing through the detector. And to run the same text through the detector in different sessions to see how the results will differ.

I tried one of the detectors, GPTZero. Used a couple of my DU posts. Was told one was human, but the other was AI. Then I took AI-generated fiction from a Reddit post and was told some parts were clearly AI and some parts were clearly human.

I've read that these detectors label the Constitution as AI-written.

[highplainsdem]

6. Here's an example of the sort of thing I've been reading

about ChatGPT destroying livelihoods. This thread in the ChatGPT subreddit has nearly 4,000 replies since it was started 3 days ago:

https://www.reddit.com/r/ChatGPT/comments/139o1q6/lost_all_my_content_writing_contracts_feelin/

Anyway, that sort of thing is why I don't find AI entertaining.

That, and wanting to hear people's own voices through their writing, and see their own creativity.