Live data from Hacker News

My kids and I just played D&D with ChatGPT4 as the DM

obie.medium.com

161–170 of 256 posts

Re: My kids and I just played D&D with ChatGPT4 as the DM

#161

It's both amazing and weirdly fallible - Tenzin, the rogue, jumps out of the tree to make a surprise attack, but in the next round of combat GPT-DM claims that he is still in the tree. Also: > Tenzin, still hidden in the tree, targets another wolf with his rapier. Does GPT4 not know that a halfling in a tree can't attack a wolf on the ground with a rapier?

> Does GPT4 not know that a halfling in a tree can’t attack a wolf on the ground with a rapier?

100% certainly can, though they won’t be in the tree afterwards unless they were on a very low branch.

Re: My kids and I just played D&D with ChatGPT4 as the DM

#162

Ha, I had the same idea a few hours ago. One thing I noticed is that the DM will let you do basically anything without fighting you on it. For example I was ambushed by some zombies and skeletons and I attempted to pitch them on a multilevel marketing scheme. Here is the result (my characters are Femribubs and Natnari): In a surprising turn of events, Femribubs and Natnari decide to change their approach. Instead of…

Have you tried telling it that sometimes it should push back on the player if what they are requesting is unrealistic-within/inconsistent-with the world of D&D?

I tried this last week and despite constant reminders to sometimes fail and provide challenges to overcome, it never worked. Frankly I’m amazed the OP chat worked so well, anything with numbers that fall over time just failed when I tried. Maybe the key was “you are no longer a friendly assistant, you are now a DM” and just the sheer amount of existent text on 5e made it work. I just gave up looking for challenge and just solved problems by the most ridiculous roundabout way just to see how gpt would make it happen

Re: My kids and I just played D&D with ChatGPT4 as the DM

#163
post #17

Earlier quoted context omitted.

Why would they hate this? They'll sell access to it, just like anyone else who is eager to take LLMs that consume content from the public domain, and repackage it into a walled-garden-entry-by-fee, giving nothing back to the public.

Why do I need to buy it from them if I can just do it myself directly on OpenAI? (Or are you implying that OpenAI will censor it?)

Maybee is implying that as WotC moves forward, they are likely to stop producing physical books and only provide new content in the form of D&D-as-a-service.

Re: My kids and I just played D&D with ChatGPT4 as the DM

#164
I had a great time playing Monster of the Week with GPT-4 when it first came out and the message limits were much higher. I loved how fluid it was as a GM and story partner. I had to teach it to push back and to realize characters with more interesting internal states, but that wasn't hard to do: I just coached it to reveal the character's inner position and desires as if it were discussing backstory.

It immediately knew all the rules (suggesting perhaps that GPT-4 was trained on some PDF libraries) and could both apply them literally and flexibly. My character's background was one where they deliberately repressed aspects which would govern their game moves, making all of my rolls go poorly. GPT-4 flexibly thought of ways to fold that idea in to its judgements on roll failures (in MotW, when a roll fails the GM "makes a move", generally progressing the story by raising the stakes). For instance, my character had a nascent form of prescience that they accessed by flipping a coin to make hard decisions. When the roll would go poorly, GPT-4 would induce visions in my character which would mislead them. We essentially brainstormed this move as a variant on the built-in one to make it more interesting given the weaknesses of my character's recalcitrance.

GPT-4 did an amazing job building environments. It proposed setting it in a sleepy town in the PNW. I've never been, so I thought it'd be interesting. GPT hallucinated a sleepy fishing town replete with a worn out lighthouse and a tourist trap B&B. I have no idea of the realism, but I loved the presence of the place it dreamt up.

Finally, it was all cut short after about an hour. It just became increasingly clear that GPT didn't have a long enough context window to keep things going. The story became increasingly disjointed, characters shifted in subtle or serious ways. No narrative cohesion was possible.

But man was it fun for about an hour. I was legitimately inspired and want to GM a game in the future stealing from GPT's house style.

Re: My kids and I just played D&D with ChatGPT4 as the DM

#165
post #50

An AI will never be able to replicate the real joy of D&D: asking every minor npc their name and watching as your friend tries to keep coming up with new names on the spot.

Yes - these the kind of things that make TRPGs interesting to me. Not the game mechanics, tedious combats or rote storyline. It's admiring the effort put in into the DM's worldbuilding, amateur acting, group laughter at good and bad jokes, and human interaction built around a shared sense of fun. The ability of LLMs to produce more or less well-realized fantasy worlds is definitely impressive, but I didn't get the feeling that anyone in the linked post was actually enjoying it.

Re: My kids and I just played D&D with ChatGPT4 as the DM

#166
post #72

Paraphrasing, the party goes to a mount shop, dude says he’s gonna roll to find exotic mounts, and voila, exotic mounts. Party rolls to try and steal an exotic mount and it doesn’t really work so they leave. This is a hard problem. It’s bad DMing imo. But it’s the natural outcome from a probabilistic model. I’m guessing you could announce your intention to search for just about anything and get the model to let you f…

I just tested it myself, and it's surprisingly interesting. Unfortunately, ChatGPT still behaves as its normal self, assertions included, as illustrated when I told my "DM" that actually the bad guy and I were good friends and asked what was up. https://imgur.com/a/4oeVsug In the first instance, I lied to ChatGPT and it took my assertion at face value, changing the scenario in a pretty extreme way. In the second inst…

Feels like artificial social engineering. Bizarre.

Re: My kids and I just played D&D with ChatGPT4 as the DM

#167
post #40

Earlier quoted context omitted.

What stage of denial is this?

I'm waiting for the moment when an LLM will prove a new math theorem or something and people will go: "Well, you know, it's not really that impressive considering it used X and Y previous results produced by humans. The model's contribution is tiny. This proof was kind-of low-hanging, a lucky PhD student would have stumbled into it eventually." It's coming.

We've always known that there's a quantum equation for gravity, like forever. Now we're all excited that it did what, some math? Whoop de doo.

Re: My kids and I just played D&D with ChatGPT4 as the DM

#168

Ha, I had the same idea a few hours ago. One thing I noticed is that the DM will let you do basically anything without fighting you on it. For example I was ambushed by some zombies and skeletons and I attempted to pitch them on a multilevel marketing scheme. Here is the result (my characters are Femribubs and Natnari): In a surprising turn of events, Femribubs and Natnari decide to change their approach. Instead of…

Possibly the system prompt has helpful and other pleasantries. May need to describe the type of dm at the beginning, with accent of keeping the story on track and making the world dangerous and unforgiving

Re: My kids and I just played D&D with ChatGPT4 as the DM

#169
post #115

Ha, I had the same idea a few hours ago. One thing I noticed is that the DM will let you do basically anything without fighting you on it. For example I was ambushed by some zombies and skeletons and I attempted to pitch them on a multilevel marketing scheme. Here is the result (my characters are Femribubs and Natnari): In a surprising turn of events, Femribubs and Natnari decide to change their approach. Instead of…

It's not strictly wrong to let the story go wherever the players want as a DM, as long as you can keep up. In fact I'd say one of the traits of a good DM is knowing how much to let the story diverge from their plans, and to be able to think quickly and respond to the players on the spot. If the direction is entertaining, most parties are okay with exploring it. How tightly you keep the story on the rails is part of t…

> What would be interesting is figuring out if you could configure ChatGPT's tolerance for divergence from the planned scenario.

I wonder how effective it’d be to explicitly state a divergence tolerance in either the system prompt or user prompt. Would be fun to find out!

Re: My kids and I just played D&D with ChatGPT4 as the DM

#170
post #15

This is not a big deal. Reading the pre-set text from an adventure module and rolling monster dice is the easy part of being a DM. clap clap ChatGPT . Having a personality, bringing out laughter, banter and fun, i.e. being a human being is the hard part. You aint got what it takes ChatGPT, stupid language regurgitating machine.

Can you show what parts of this campaign were pre-set text? Because I have a suspicion you've not read the article. This is a fully interactive campaign, and you can easily set up a unique scenario. I just tried the same prompt, but on the moon. It set up a situation with a secret moon base, characters that were trying to help you get past the guards, etc etc.

Nice way to miss my entire point.

I don't care if the campaign is pre-set or one you made up.

The point I was clearly making is DnD is about much more than the dungeons and dragons. It's about the camaraderie, banter, chitchat, and even arguments, the kinds of things you can meaningfully do, or want to do, with real humans.

Post reply on HN