ChatGPT writes a character background

[DougGraves]

5 hours ago, GroovyJivester said:

How can you say a person is lazy by using A.I. as a writing assistant? 

 

People are lazy using AI as a writing assistant.  But not nearly as lazy for that as for using a computer to play a superhero game.

 

Proper humans play Champions TTRPG and don't use computers for any of it.

[eiynp]

All chatgpt and similar generative models will ever be able to do is mimic; they can mimic some stuff pretty well right now, and in the future it seems likely they'll be able to mimic more things. But that's all they do, and imo people will find their limits pretty quickly (stable diffusion for example is already starting to have problems filtering input that was generated by the engine itself.)

 

it'll have applications for writing very formulaic material (like say, level 1 support, or clickbait news) but it's never gonna replace human authorship because it can't

[Crysis]

On 2/17/2024 at 7:42 PM, FFFF said:

Here's the prompt I used:

 

 

Here's the description it wrote up:
 

 

 

Obviously, you don't have to be a literary genius to write character backgrounds, but I was pretty impressed with the output by ChatGPT. It even came up with a name and a civilian identity.

 

You can also have ChatGPT limit the number of characters used to get a perfecly fitting bio for copy/paste.  And for extra fun, have it write the bio in a particular style, like “…using hillbilly Appalachian slang” or “….Roman-styled prose.”

[sbloyd]

I dunno, I get tells complimenting my costumes and bios all the time.  And I do the same to others.  What, you think I stand under Atlas all day doing nothing?

[Clave Dark 5]

7 hours ago, Doomguide2005 said:

Chat GPT and similar AI are as awful at it as they currently are.  It's only going to improve.  Just think about how long they've existed and what the original outputs looked like.  There's material out there now that the vast majority of folks won't even notice is AI generated when they look at it and many of them couldn't pick out the 'fake' even if told one was AI generated.

 

You kind of missed the point though it wouldn't surprise me if that was deliberate.  I think you had the post highlighting all the bits of unimaginative writing up thread.  Well those various phrases are unimaginative crud commonly known as tropes and virtually all human writing is full of them which is why there's a word for it.  Sure Chat GPT writes like a poorly plagiarizing high schooler at the moment.  But that's just in the current state, it was only a short bit of time ago when it would have barely passed muster as a poorly plagiarizing kindergartner.  And I've no idea how skilled at writing prompts FFFF is either.

I'm sure it will get better at not being completely bland, but it still, by definition, can't come up with an original angle  or make a thematic connection between things, unless one of it's scraped sources did so, or its human wielder tells it to.  I figure a lot of it's going to stay in "fairly dumb" territory, the occasional peak aside.  People are treating it like it's going to drive our cars and do our homework for us, like a train smoothly following previously laid tracks... without realizing all the work that goes into laying down tracks (planning, design, materials considerations, traffic laws and zoning, track maintenance, etc.)  Humanity is basically a lazy bunch, so there's going to be some really telegentic screw-ups as time goes by too.

[Sovera]

You're looking at the current iteration as if it was the end all. Not too long ago people would laugh if we said we can have an 'AI' do a complete program by itself. Not too long ago the most context GPT had was 4k. It then went up to 8k. Currently we are at 128k. There are others with 200k now. And a new one is/has been released with 10 million context. And that's with tech not even geared (seriously, using GPUs and graphic cards) where the tech is now catching up with GPUs tailor made for AIs being sold to the public.

 

A lot of these improvements are even being made by the open source models so anyone can download them and run them in their PC instead of having to rely on GPT and big companies.

 

So yeah, it's not driving cars yet, but poo pooing at the progress made as 'just' repeating brings back what I said earlier on how every book and song has been 'just' recombining the same alphabet over and over.

[Outrider_01]

I just asked Chatgpt to generate a bio with a 1024 limit, it was still 400 short

 

Aya, an Egyptian priestess of Hathor, was revered for her wisdom and beauty. But her fate took a dark turn when she became possessed by the demon Amunet, seeking vengeance for ancient wrongs. Consumed by darkness, Aya's once gentle spirit was twisted by the demon's malevolent influence. As chaos spread, her fellow priests, fearing the demon's power, sealed her within a tomb deep beneath the desert sands. There, entombed alive, Aya's consciousness fought against the demon's grip, trapped in a nightmarish existence. Over centuries, her story faded into myth, until archaeologists unearthed her tomb, releasing her from her ancient prison. Now, Aya wanders the modern world, grappling with her past and the lingering presence of Amunet, seeking redemption for sins not entirely her own.

 

Like, thats about 80% of what I had in mind.  Im going to edit and modify it...but dang the AI is reading minds now.

 

Down side if you generate the same request, its not much different each time.  Names change, theme is the same.

[Clave Dark 5]

6 hours ago, Sovera said:

... brings back what I said earlier on how every book and song has been 'just' recombining the same alphabet over and over.

... which brings back what I said earlier on how AI requires the intention and direction of intelligent users; that's more than just infinite monkeys banging away on typewriters, and adding more monkeys and typewriters doesn't change that.  

 

Without the guidance before and refining afterwards, you risk unforced errors; with those, you get bland stuff - which is fine enough for a bread bag label, but one does not live on bread bags alone.  Maybe I'm wrong, sure, but get back to me when it can conceive of and compose a good poem.

[Marshal_General]

I need to look at it for characters I made based off of look and power sets to give me inspiration towards a back story for them.

 

The problem I see with ChatGTP and Grammerly, is that soon companies will be hiring semi illiterate people who can't string together a coherent sentence without ai help.

[Jack Power]

16 hours ago, DougGraves said:

 

People are lazy using AI as a writing assistant.  But not nearly as lazy for that as for using a computer to play a superhero game.

 

Proper humans play Champions TTRPG and don't use computers for any of it.

Myself I go out to tackle busses and leap from rooftops. TTRPG. Pfft.

 

//Jack

Edited February 21 by Jack Power

[Doomguide2005]

14 hours ago, Clave Dark 5 said:

I'm sure it will get better at not being completely bland, but it still, by definition, can't come up with an original angle  or make a thematic connection between things, unless one of it's scraped sources did so, or its human wielder tells it to.  I figure a lot of it's going to stay in "fairly dumb" territory, the occasional peak aside.  People are treating it like it's going to drive our cars and do our homework for us, like a train smoothly following previously laid tracks... without realizing all the work that goes into laying down tracks (planning, design, materials considerations, traffic laws and zoning, track maintenance, etc.)  Humanity is basically a lazy bunch, so there's going to be some really telegentic screw-ups as time goes by too.

Two words I'm told i should say or ask about "Emergent properties"

[Clave Dark 5]

9 hours ago, Doomguide2005 said:

Two words I'm told i should say or ask about "Emergent properties"

Interesting philosophical theory!  Not one I'm qualified to get into so: maybe? Who knows, we'll see.

[Judasace]

Motorcars are a fad, nothing will replace the horse and buggy. Those calculators are never going to replace the slide rule. Computers are as big as a house, never gonna fit one in every house. Technology will never improve or surpass the limitations of my personal experiences, biases and limitations.

 

 

 

 

[Krimson]

You wanna hear about what people were saying about digital art and Wacom tablets 20 years ago? I still have the Graphire 3. 😄 People were NOT nice about digital art, and it certainly wasn't considered REAL art. So, I have AI language models and Stable Diffusion running on my desktop, and I could care less what people think because I remember what they said about digital art, and I remember what people said about my pen and ink art before that. So I'll make what I want. I don't care what the naysayers complain about because I know that if AI didn't exist, they would find something else to complain about.

[Clave Dark 5]

4 hours ago, Krimson said:

You wanna hear about what people were saying about digital art and Wacom tablets 20 years ago? I still have the Graphire 3. 😄 People were NOT nice about digital art, and it certainly wasn't considered REAL art. So, I have AI language models and Stable Diffusion running on my desktop, and I could care less what people think because I remember what they said about digital art, and I remember what people said about my pen and ink art before that. So I'll make what I want. I don't care what the naysayers complain about because I know that if AI didn't exist, they would find something else to complain about.

Not the same thing, you drew with your tablet (as I did with mine), you didn't click "create" and call it a day.  You can make what you want, just as I am free to not have to say it's perfect; that's not complaining.

 

Some other words to words to consider:  https://locusmag.com/2023/12/commentary-cory-doctorow-what-kind-of-bubble-is-ai/

[Krimson]

11 minutes ago, Clave Dark 5 said:

...you didn't click "create" and call it a day.

Heh... That's what you think the process is, is it?