sierralee
patreon
A Somber Announcement
🕑 Added 2023-04-01 16:53:03 +0000 UTCWell, it's finally happened. ChatGPT has gotten too good and there's no point being a writer anymore. Artificial intelligence will be writing all of TLS going forward, starting with this real dialogue actually generated by ChatGPT-4:
-
Aka: "I can't believe we have to work with a succubus like Yarra. It's not appropriate."
Yarra: "Oh, come on Aka, lighten up! We're all on the same team here."
Qum: "Hehe, yeah, Aka, don't be such a prude. You need to loosen up a little."
Aka: "I'm not a prude! I just think there are more appropriate ways to conduct ourselves."
Yarra: "Look, I understand that you come from a different culture, Aka, but here in the demon world, things are a little different. We don't have the same hang-ups about sexuality."
Qum: "That's right! Us succubi love sex and pleasure. We're not afraid to embrace our desires and share them with others."
Aka: "I understand that, but I still think we should be more professional."
Yarra: "Well, if it makes you feel any better, we'll try to tone it down a bit. But you have to understand that we're not going to change who we are just to make you feel more comfortable."
Qum: "Hehe, yeah, we're succubi after all! We love to have fun and explore our desires. If you can't handle it, maybe you should find another team."
Aka: "I'll be fine. I just think we should focus on the task at hand and not get distracted by... other things."
Yarra: "Fair enough, Aka. Let's focus on the mission and get the job done. But if you ever change your mind, just let me know. I'd love to show you the pleasures of succubi."
Qum: "Hehe, me too! I'm always ready for some fun."
-
Okay, I'm not going to maintain a facade, obviously this is an April Fools' Day joke. XD Everything demarcated by hyphens is unedited output from ChatGPT. After prompting the chatbot with information about TLS, I described the three characters and asked it for a dialogue. The above was the first output I received; I tried further to get Qum to refer to herself properly but never could, perhaps because the chatbot has "cumdump" forbidden from output.
It was all roughly like this. ChatGPT can easily produce a coherent dialogue, and is happy to confabulate information, but I had a devil of a time trying to pin it down on details. No matter how many specifics you feed it, the chatbot fills in a ton of information with generic assumptions.
Here's another one:
-
Simon sat at the table, nursing a tankard of ale. Orcent walked over, carrying a plate of roasted meat.
Orcent: "Greetings, Simon. I see you're deep in thought."
Simon: "Yeah, just thinking about how messed up the world is."
Orcent: "Ah, the cynicism of a veteran soldier. What specifically troubles you?"
Simon: "Capitalism. It's all about greed and profit, no matter who gets hurt. And with lust and purity magic in the world, it just gets more complicated."
Orcent: "I see. Do you believe capitalism is inherently unethical?"
Simon: "Of course. It's built on the exploitation of the working class and the natural resources of the world. And with magic in the mix, it becomes even easier to manipulate and control people."
Orcent: "I can see why you would think that. But what if we approached capitalism differently? What if we valued the well-being of everyone involved in the economic process, rather than just profit?"
Simon: "That's a nice thought, but it's not realistic. The system is rigged against us, and those in power will always prioritize their own interests over anyone else's."
Orcent: "I understand your skepticism, but I believe that change is possible. And with lust and purity magic, we have the ability to shape people's desires and values in a more ethical direction."
Simon: "I don't know about that. Magic is just as susceptible to corruption as anything else."
Orcent: "True, but it's not the magic itself that is corrupt. It's the intentions and actions of those who wield it. We can use magic to empower people to make more ethical choices, rather than just satisfying their own desires."
Simon: "I see your point. But it's still a tall order to expect people to act selflessly, especially in a capitalist system that rewards greed and selfishness."
Orcent: "I agree that it's not easy. But we have to start somewhere. By questioning the status quo and advocating for more ethical practices, we can slowly shift the values of society. It won't happen overnight, but it's better than doing nothing."
Simon: "I suppose you're right. It's easy to get cynical and give up hope, but maybe we can make a difference if we try."
Orcent: "Exactly. We can't control everything in the world, but we can control our own actions and attitudes. And that's a powerful starting point."
-
It was extremely easy to get a typical dialogue about subjects like capitalism, so I pushed hard with my prompts to try to get Simon and Orcent discussing magic. It... kind of worked.
Honestly, though? As much as that dialogue made me laugh, I think lots of writing-indifferent developers/producers would be happy to replace their writers with ChatGPT. You don't generally need to discuss topics in any depth, just gesture vaguely at them. Chatbots can replicate archetypes easily, and that's what many people want.
But this was honestly meant to be a silly post! I won't share every single prompt, but I did find it interesting to see what ChatGPT said when I asked it to describe TLS:
-
The Last Sovereign is a role-playing video game developed by Sierra Lee, released in 2014. The game features a rich storyline with complex characters and strategic gameplay.
The game is set in a world called Arclent, where humans and demons coexist. The player takes on the role of the last sovereign of humanity, a powerful mage who must navigate political intrigue, fight battles, and gather allies to save their people from a mysterious threat.
The Last Sovereign has a non-linear storyline with multiple paths and endings, allowing players to make choices that affect the outcome of the game. The game also features a unique combat system, where players must carefully choose the skills and equipment of their party members to defeat challenging enemies.
One of the standout features of The Last Sovereign is its mature themes and well-developed characters. The game deals with topics such as power, politics, religion, and sexuality, and the characters are complex and multi-dimensional, with their own motivations and backstories.
Overall, The Last Sovereign is a highly regarded game within the RPG genre, praised for its engaging story, memorable characters, and strategic gameplay.
-
That's vaguely accurate! It seems likely that ChatGPT scraped the Steam page for TLS and derived the information mostly from there, which means that all the positive reviews led it to construct effusive praise. Some of the text is clearly just looking at certain tags and spitting out common text for game writing.
What's interesting to me is the inclusion of "demon" - you also saw it in a previous dialogue. The word doesn't appear anywhere on the Steam page or any other major pages. I suspect that in the ChatGPT black box there is a "fantasy = demons" connection that got deployed.
Anyway, this post was inspired by the joke on my SFW Patreon. That one turned out more ridiculous, but won't really be funny unless you read those books. I would have made AI art for this post as well, but "succubus" got flagged as a forbidden prompt and I decided that I shouldn't spend too much time on this. =P
Comments
ushiroace
I can't stop thinking about that line The idea of Simon, the living definition of a "man of action", drinking sullenly while he ponders "how messed up the world is" is S tier parody material.
Hotenhornay
"Disclaimer: techniques shown in this lecture series are intended for standard level seductions only. For subjects with more divinely entrenched inclinations, please refer to our advanced lecture series 'Putting the "Hole" in "Holy Magic": A Succubus' Guide to Seduction and Synods'."
Sierra Lee
Haha, please consider it canon that succubi have "How to properly tempt humans" lessons.
Runcible Technician
ChatGPT describing two succubi trying to temp a mortal is the cringiest thing I've ever read. That was on the level of a retail instructional video.
Markus S.
Well, I haven't tried ChatGPT directly, but the underlying OpenAI is also used for AIDungeon. From my experience there and looking at the image outputs, I can say: It's kinda entertaining, but it essentially cannot create some new. It's so good at remixing existing stuff, that you might mistake it as something new, but especially it's just a remix of what humans created. In that way btw. it will be a challenge to future copy right laws. I am telling you right now, government will probably need another few decades to realize. Also the AI has no real grasp of a situation. It can suddenly mix up properties of one character to another, to the point that the dick of the male one suddenly "appears" on the female one. It has no concept of "room", if you head somewhere, and turn 4 times right, it will not acknowledge to be on the same direction again. It has no concept of anatomy of a creature, it can just repeat similar descriptions, as it knows which stuff humans most when the keyword "from behind" comes up, but it could also result in something that's anatomically only possible from the front. Btw. also from the autonomous car driving stuff we realized that we have no idea on which actual concept deep learning actually interprets stuff it got deep learned. So when training traffic shields, you'd assume it trains on it's general shape and colours, but it actually tries to find a set common pixel in all training pictures. But then in real live a sticker on a hand rail can be mistaken a one-way street sign and suddenly turns 180°. Bottom line: Unless AI is trained in a way, humans think, instead of just mixing up stuff that has a x% of likeliness, the result will ultimately not result in anything new. So it can be used for content stretching (like making townspeople NPC talks in a human created game/story) or grabbing you some information as an AI search bot from a lexicon. And that's imho. what it can be used oh-so-cleverly unless the AI concept is evolved to a whole new level (and likely only runnable on a quantum computer).
Lord Forte
Conceptually, I think we'll just have to disagree there. To me, just the knowledge that it can only draw on and remix existing content essentially poisons the well, because I couldn't use it without wondering how much of what it created was essentially stolen, and from where. Also, I'd argue that the nature of LLMs is more prone to truly reusing generic idea than like art AIs, especially considering Sierra's notes about essentially being unable to get rid of many generic assumptions. Maybe as a mathematician by education, I'm just unable to trust something that can't or won't "show its work" as well. For example, I remember reading an article a while back talking about how an AI coded with reasonable directives essentially discovered a way to cheat to get to the desired result easier (it was supposed to reproduce images, iirc, so it determined the most efficient way yo do it was to hide a small copy of the original somewhere and work from that). To an extent that's human error, not AI error, but it's a type of human error that's a danger in making any set of rules that only needs to be followed literally, not in spirit. As a writer, I try to be very aware of my influences and deliberately moderate how much I let myself take from someone else--to ensure when imitating, as Petrarch said, that the resemblance is as son to father, not painting to subject. I have no reason to believe an AI black box cares about that distinction, which is both why I have no desire to use it, and why I think it should be banned. Most people don't even think about that distinction, much less have the ability to figure out if they're unknowingly borrowing too much when relying on AI output. This issue is more prevalent for art type things than LLMs, though. Moreover, whether such an "idea generator" has any real value to someone is highly subjective. To me, it has almost none because I'm good at generating ideas myself. If someone wants to use it for that purpose, I suppose I can't really complain too much, since they're still limited in how well they use the ideas. I personally think their time would be better spent cultivating their ability to come up with ideas themselves, so I would still only classify it as "fine." Honestly, while I'm sure some AI applications will stick around if only because they save corporations money in contexts where quality is of limited relevance, much the AI type feels like a solution in search of a problem like blockchain was, and I hope once the hype dies back, many AI things will met a similar fate. Anyway, I think I'm going to avoid going deeper on the topic than that, if only because this was a silly thread and I'm a terminally serious person who probably already brought too much of that seriousness with me XD
Drakenclaw
The thing is, the AI, at least the good ones, aren't just creating reused content. They are capable of creating new original scenes by combining existing ones in a way that no human would. Which also means they are going to be used as inspiration source for creative humans. One youtube channel for example used AI generated magic cards to make encounters for a RP dungeon. The AI creations were used as foundation for the encounter and then fleshed out by the artist. The new AI are an automation tool and they are going to do the same thing that all automation tools do. They are going to make things easier for some people, while also making some jobs redundant.
Lord Forte
I actually personally would be happy with a blanket ban on generative AI, haha. I don't really think it has any truly "good" applications. I think it would be "fine" replacing busywork, but I also don't think it adds any positive value in doing so for anyone but the CEOs. And the applications you mention are definitely negative. I just am glad, if we have to deal with it, that it feels like it's only any threat to things I perceive as meaningless work. I was discussing this with an artist friend earlier (we're both pretty anti-AI), and he pointed out, if anything, people relying on the AI crap for anything creative will likely accelerate the boredom with reused content we personally both see building in the public consciousness. Since AI is just a way of amalgamating and meta-izing existing trends, relying on it arguably makes actual originality a more valuable resource. But again, not really in favor of it, just looking for silver linings.
D
Yarra: Simon, stop reading Marx and let me seize your means of reproduction.
Sierra Lee
You guys are thinking about potential useful applications of the technology, but I'm thinking about the worst applications. There was already that admin using ChatGPT to write a condolence email, and there have been people experimenting with using chatbots to automate sending messages on dating apps. If either of those becomes widespread, the chatbot tone will be inflicted on everyone else. I've yet to see anything from ChatGPT-4 that strikes me as uncanny, but I'm not thinking about the type of things I'd want to create. There are plenty of projects where a character might be conceived of as "a tsundere" and ChatGPT will be able to write that archetype just fine.
Lord Forte
Yeah, I could see those uses, as long as you proofread/edited it. But I don't think replacing that kind of busywork is replacing anything meaningful XD
Drakenclaw
It certainly seems coherent enough to replace quite a few CEOs. :P Hasn't one japanese firm already done that? I did hear something in that direction, though that could just have been some funny rumor. I can see it or similar programs replacing "standard" texts. Like a generic introduction text of a new product. Or a press release about how your company has done the last quarter.
Lord Forte
If nothing else, the surface-level nature of this kind of output is enough to make me wonder how anyone can be worried about it replacing anything meaningful, honestly. I suppose there are plenty of people who write at a similarly superficial level themselves, but I feel like it's prose than only a corporate CEO could love XD Not to say that you're wrong that it will be enough for some people, I just don't think those people are worth caring about the opinions of, if they'd be content with this. I may be somewhat biased, though, since I have an almost uncanny degree of ability to derive nuance from text and read tone, and every AI-written text I've seen has certain telltales (which are rather similar to the useless business jargon CEO types favor).
Lord Forte
Yeah, I was thinking this too. TLS is the only world I'm familiar with where succubi are a "race" rather than a class of demons.
Sierra Lee
Glad someone else got a chuckle out of it. ^-^
Sierra Lee
It's difficult to forecast something like that, especially because I have no particular expertise. Right now I haven't seen anything that would make me believe ChatGPT will raise any standards.
Dubsington
Oh my god, Simon sounded like basically every college sophomore. That is fantastic.
Maz
Do you think involving ChatGPT into writing would lead to more bland games/stories? Or would it uplift the already bland games/stories to a higher standard?
Sundeigh
Very interesting to read since it's just the most bare-bones version of dialogue between these characters. Reminds me of other games that either have newer writers or are just a writer who isn't very motivated to make things unique.
Handsome Stalin
I fucking cried when Simon said "Yeah, just thinking about how messed up the world is." It's just so true, so raw, powerful, i cant On a more serious note, I'm very excited to hear you announce that TLS is getting a character class system, I can't wait to play as Simon the powerful mage!
Deloir
It's not a fantasy = demon connection but a succubus = demon connection.
FormerHater
Looking forward to it. I love the new twists and betrayal plots that a human mind couldn't possibly fathom. The title should be changed to better reflect the change in writer "The last man or woman who reigns within a geographical region".
