D&D General Can ChatGPT create a Campaign setting?

[Burt Baccara]

But what counts as work created by an AI? Lots of work that has been copyrighted already uses AI extensively. If I give it the prompt, is it still work created by an AI? What if I give it ten prompts, and ask it to incorporate specific details?

What works? This video from Corridor Crew's "Lawyer Explains Stable Diffusion Lawsuit (Major Implications!)" is my source. Followed by Google to see if it was true.

Restating, I am not a lawyer.

 

[Burt Baccara]

This likely has to do with the resources being allocated to it.

For sure, another cause may be the finetuning to get Chat-GPT to be more reliable and not make stuff up as much. At first, it was great at saying false claims with convincing conviction—it still does, but less often.

Remembering what was said upstream is an advertised feature on new chat screen. I am sure this is just a hiccup, but it's been like that for about a week now and really impacts how one interacts with Chat-GTP and in some ways its usefulness.

 

[Clint_L]

A lot of folks are not responding my argument, so I must not be explaining it well. Maybe I should enlist ChatGPT.

Will ChatGPT write the next Great American Novel on its own? No, not as it currently exists. Could it be enlisted to write the next Great American Novel? Quite possibly.

Here's an analogy: most people would agree that Andy Warhol is a highly important artist of the 20th century, despite the fact that much of what came out of Andy Warhol's Factory and was attributed to him was physically done by other artists working under his direction (many other visual artists have worked similarly, from Michelangelo to Banksy). Similarly, we consider Martin Scorsese a great film direction, despite the fact that most of the production of any of "his" films was done by someone else. That is because we recognize the creative vision guiding the process, and we see the other contributions as secondary.

Similarly, ChatGPT gives every human being a personal assistant that is a highly competent and incredibly fast writer. Stable Diffusion and similar technologies are giving us highly competent artists. We still have to provide the imagination and guidance. But how much more can we create, on our own, using these technologies?

Let's go back to the original question. Imagine that I want to create a detailed setting for my campaign, based around, I dunno, penguins. I'm a good writer, and I am capable, but do I really want to invest the time producing a 200 page guide to Pengunia? Maybe, but probably not. But a lot of that writing would effectively be drudge work. With ChatGPT, I am suddenly Martin Scorsese. I can get it produce a ton of iterations on different ideas. I can throw in my own writing whenever I want. I can tweak and massage and experiment. Artistically, I can let ideas build up far faster than I could writing on my own. And I can probably write a pretty cool guide, original to the world and specifically tailored to MY needs, in a fraction of the time.

That's what this offers: incredible creative freedom. Will other things be lost? Yes! Maybe ideas I would have had through writing it the old way will never happen. But that is what happens with every new technology. That does not make ChatGPT good or bad. It is a powerful new tool.

 

[Clint_L]

What works? This video from Corridor Crew's "Lawyer Explains Stable Diffusion Lawsuit (Major Implications!)" is my source. Followed by Google to see if it was true.

Restating, I am not a lawyer.

That is a specific case about a technology that deals with one type of AI. It is not going to come close to resolving the broad issues I raised. As I pointed out, many pieces of art that have an incredible amount of AI contribution are already copyrighted. The question isn't whether work done with AI can be copyrighted, but rather where lines should be drawn in terms of originality and methods for building the AI.

 

[Jer]

I can get it produce a ton of iterations on different ideas. I can throw in my own writing whenever I want. I can tweak and massage and experiment. Artistically, I can let ideas build up far faster than I could writing on my own. And I can probably write a pretty cool guide, original to the world and specifically tailored to MY needs, in a fraction of the time.

Go for it! I suspect you're going to find that it's actually a lot more work than you think and you'll end up abandoning the ChatGPT once your own creative energies get flowing, but maybe it'll work better for you.

I say this because I started doing exactly that, was excited at first but quickly found out that what ChatGPT was giving me kinda stank and I was doing A LOT of editing and/or random restarting to make it sound even reasonably like something I'd be happy to put my name on, and stopped using it. It started out as a fun little exercise but rapidly became more work than just writing it for myself.

 

[Asisreo]

Humans, on the other hand, have a complex system of inputs that include not just what we've read but also what we've experienced. And our emotional responses to these things. Yes if you want to get essentialist these are chemical reactions to the experiences that lead to different brain states, but we have them and they are an input that Large Language Models - which only have language - do not.

What's important to me, though, is if the inputs and memory barriers are the only thing stopping a "human" AI.

In other words, if we were reduced to only being able to interact with the world through reading or writing text, with no visuals, sounds, etc. And if our memory was nerfed to only remembering within a thread's worth of conversation, what would be the difference in the responses of ChatGPT and ourselves.

Another thing. ChatGPT isn't very original, but I don't think it's trying to be. I imagine that if ChatGPT sees a pattern across its input data, it will want to copy that pattern because it's being reinforced. Unlike humans, it's goal isn't to design something interesting or unique, it's just answering the prompt in the most human-sounding way possible.

But that doesn't mean a different AI can't be programmed specifically for novelty or engagement.

 

[Reynard]

How much do I need to massage ChatGPT output for me to have a legitimate copyright claim?

 

[Andvari]

How much do I need to massage ChatGPT output for me to have a legitimate copyright claim?

Or how would anyone know and then prove your massaged text came from AI?

 

[Jer]

What's important to me, though, is if the inputs and memory barriers are the only thing stopping a "human" AI.

This is an open question in AI research -whether "strong" AI is possible or not. The jury is still out. But LLMs are not that and may not even be a fruitful path towards getting that. The jury is still out on that too (though I'm in the skeptical side - I think it's likely to end up being a neat toy that can do some things very well but ultimately a dead end on the path towards strong AI).

In other words, if we were reduced to only being able to interact with the world through reading or writing text, with no visuals, sounds, etc. And if our memory was nerfed to only remembering within a thread's worth of conversation, what would be the difference in the responses of ChatGPT and ourselves.

I mean, yes, if you take away everything that makes us human and reduce us down to a state machine that gives the same performance as a GPT algorithm you'll get the same performance. It's kind of a tautology there.

I know what you're getting at - "is there something magical about humans such that an AI can't replicate us". And the answer to that is "the jury is still out" (see above). But I can tell you we ARE more than machines that collect statistics on the syntax of a language and repeat back convincing sounding text based on randomly perturbing our way through those statistics. If that was all we were entire branches of philosophy would never have been invented for starters. A large language model cannot analyze things, it can only generate text based on the distribution of words its discovered in the data it's trained on (which is why they can't really be stopped from lying either. "The sun is the source of its own light." and "The moon is the source of its own light." are both quite reasonable sentences for it to construct and the first one is only slightly more probable than the second absent any ability to do more than construct strings of words into a plausible syntactically correct sentence. And depending on the training set they may be equally likely. Or the second may even be more likely.)

Another thing. ChatGPT isn't very original, but I don't think it's trying to be. I imagine that if ChatGPT sees a pattern across its input data, it will want to copy that pattern because it's being reinforced. Unlike humans, it's goal isn't to design something interesting or unique, it's just answering the prompt in the most human-sounding way possible.

But this is actually something that is part of the basic idea of a large language model - it's a model that is built from basically collecting statistics across large datasets of text. So if you want true novelty you need to be thinking about a different algorithm. So this technology is not going to get better at it, it will only get better at generating realistic sounding text and keeping a consistent "train of thought" going as resources increase and it can keep a wider and wider context of the previous discussion it's had. What you'd want is a new breakthrough - a new algorithm that approaches the problem in a different way.

Now the LLM could be used as a front end to an algorithm that generates novel ideas - a way to express the ideas that it comes up with. That was the original purpose of LLMs at one point - to be used in tandem with another algorithm to help make the text generated read more like actual natural language. Using them by themselves is an artifact of researchers being somewhat surprised at how much of the meaning of what we say can actually be gleaned from the syntax that it has learned. (I'm actually not sure how well this research is generalizing to other languages right now - I haven't seen a lot of publications about non-English LLM algorithms. I should go looking I suppose.)

 

[Asisreo]

I know what you're getting at - "is there something magical about humans such that an AI can't replicate us". And the answer to that is "the jury is still out" (see above). But I can tell you we ARE more than machines that collect statistics on the syntax of a language and repeat back convincing sounding text based on randomly perturbing our way through those statistics. If that was all we were entire branches of philosophy would never have been invented for starters. A large language model cannot analyze things, it can only generate text based on the distribution of words its discovered in the data it's trained on (which is why they can't really be stopped from lying either. "The sun is the source of its own light." and "The moon is the source of its own light." are both quite reasonable sentences for it to construct and the first one is only slightly more probable than the second absent any ability to do more than construct strings of words into a plausible syntactically correct sentence. And depending on the training set they may be equally

I don't think anyone thinks there isn't more to human language than what ChatGPT outputs. But those limitations theoretically can be overcome. It wouldn't strictly be an LLM, but if we're going for "more human," then we'd have to expand past that anyways.

Anyways, it's a bit off-topic on whether ChatGPT can make adventures.

But this is actually something that is part of the basic idea of a large language model - it's a model that is built from basically collecting statistics across large datasets of text. So if you want true novelty you need to be thinking about a different algorithm. So this technology is not going to get better at it, it will only get better at generating realistic sounding text and keeping a consistent "train of thought" going as resources increase and it can keep a wider and wider context of the previous discussion it's had. What you'd want is a new breakthrough - a new algorithm that approaches the problem in a different way.

We would need a new algorithm to imitate human's creativity, but maybe not insurmountably complex.

I mean, human creativity is also based on inputs and association. Even if the inputs are a few degrees apart, having an algorithm query similarities to two concepts and merge them together contextually wouldn't be impossible. Nor would taking how many times it's seen a concept and trying to put a twist on it. Although they might be pretty difficult to program in.