I know, right? You can train it within a session, but it loses all of that content when you end the session. I'm pretty sure that is an intentional limitation to manage the experiment. You can see how if it was able to keep learning from what you input, so that over time it absorbed your preferences (such as removing adjectives like "large") and your ideas about what made a good module, you could create modules in a very short period of time. Maybe even good ones. I bet you could get to "not making @bryce0lynch hate his life" pretty easily.Frankly I'm astonished at how much semantic content it was about to retain between questions, e.g. "not initially hostile" and "use lizardmen instead."
The computer's output is still very much dependent on human input, so at this point there is still value to human creativity. It is just another tool.This is astonishing that there were not many more verbal gaffs. It's almost like you are chatting with the computer on the Enterprise. Stunning.
I should love this more because, if fast enough, it could make it possible to generate the details of an entire world on-the-fly with an amazing level of detail.
Truth is, I see what this is going to do to content --- make it so ubiquitous and effortless that it will eventually lose all value.
And I paranoid? FWIW I never was much into random dungeon generation with dice and tables either. Even if indistinguishable, it means more to me knowing it coming from a human mind.
Personally I don't care whether it came from a human mind. I care whether it's self-consistent and can be interacted with in a way that feels real. My dissatisfaction with random dungeon generation centers on the lack of cohesion: monster A's presence in the ruins is utterly decoupled from monster B's and from the state of the ruins, BY DESIGN. I would enjoy a more sophisticated procedural generation method which seeds a region with resources and some monsters, then has monsters move towards resources they desire, form alliances with or drive out other monsters, and leave signs of these interactions for desecrators/delvers/adventurers to observe and infer things about which monsters are where and where treasures can be found.[Am] I paranoid? FWIW I never was much into random dungeon generation with dice and tables either. Even if indistinguishable, it means more to me knowing it coming from a human mind.
If it lacks those elements today...then that will be fixed someday for sure. All the arguments against AI generation based on content quality will eventually be addressed...and yet it will still sadden me to see the human's role in all this diminished. That's because I take so little pleasure in being a consumer and so much a creator. I fear it will hollow out the hobby as an active hobby (as opposed to a yet another passive-consumption activity).I didn't look closely enough at the chatGPT output to see if it has this kind of consistency, but I would be astonished if it did.
Eh, maybe, but frankly I'm astonished enough that a GPT-based engine was able to copy existing modules. I'm not going to hold my breath hoping for something more, especially when it would be so much easier to create the deeper modeling explicitly. It's like hoping for GPT to generate a working videogame for you--it could happen in theory with nonzero probability, but it's not something I expect GPT to be capable of and not what it's designed for.If it lacks those elements today...then that will be fixed someday for sure.
You're more optimistic about AI content generation than I am. I see potential there for non-artists to generate customized art that is good enough for some purposes ("robot turtle with a laser gun") but only for casual purposes, like a quick-and-dirty splash screen for a game prototype ("Megaturtles", the game I am trying to get ready before the end of this month). It's faster and easier and more legal than copy-and-pasting together photographs from old magazines, but it's recognizably samey. Getting good art from a human artist would be preferable if I knew how to draw or could hire someone else to do so.All the arguments against AI generation based on content quality will eventually be addressed...and yet it will still sadden me to see the human's role in all this diminished. That's because I take so little pleasure in being a consumer and so much a creator. I fear it will hollow out the hobby as an active hobby (as opposed to a yet another passive-consumption activity).
Morbid and morose, I know. I'd be better served to adopt Beoric's (and other's) viewpoint of AI content generation as a tool.
Jimmy Cricket! How do I get a private instance of this that doesn't lose its memory? I could feed it all my setting information and it would spot patterns and interconnections that I have forgotten about or failed to spot. (sigh). I plotted out the outline of a ridiculously large map for my Majestic Fantasy Realms. With something like this I could use it to track deep interconnections between regions and characters.I know, right? You can train it within a session, but it loses all of that content when you end the session.
The good news is that it is a major idiot when it comes to anything else other than pattern recognition and creation. Think of it as the ultimate random table. But what can do is learn the patterns you feed it and spot inter-connections that even you are not aware of.Why do I still want to run away screaming? I see how useful this could be --- just another search engine of sort.
And yet.....EEEEiiiiiiaaagaghhgh!!!! Monsters!
Kill it! Kill it with Fire! Quick, before it spawns!!
The future $65/mo subscription-only version (or $299/mo pro version).Jimmy Cricket! How do I get a private instance of this that doesn't lose its memory?
Interesting to see that even AI can write a better kitchen than most of the dreck Bryce reads.7. Kitchen: This is a large room with a fireplace and several cooking pots, where the goblin minions of the dungeon prepare their meals. There is a trap in the room that will release a burst of flames if triggered, which has a one in six chance of affecting any creatures in the room. There are currently two goblin minions (neutral, armed with short swords and shields) in the room, who are cooking a stew for dinner. They are not initially hostile, but will attack if provoked or if they feel threatened. They are motivated by greed and loyalty to their goblin boss. They want to have good food and be well fed. They have a fear of being caught and punished by their boss if they fail in their duties. They are currently cooking and chatting among themselves.