• The VOIDRUNNER'S CODEX is coming! Explore new worlds, fight oppressive empires, fend off fearsome aliens, and wield deadly psionics with this comprehensive boxed set expansion for 5E and A5E!

D&D General The DM Shortage


log in or register to remove this ad

Celebrim

Legend
Prove humans can.

It's an open question as to whether they actually can, but we do know that the number of situations in which their algorithm can appear to reason is much greater than those of a computer.

Personally I am not a believer in general intelligence. I believe humans have a collection of sophisticated and specialized algorithms that work together to reason in a limited set of contexts. Computers can actually do math, but humans can't do math (actually they can, but only in a very small set of contexts). To "do math" a human has to hijack some of its other algorithms that have to do with symbol manipulation and painfully and inefficiently work out the answer. There are obvious huge gaps in human reasoning ability though, one of the more famous being that humans can't and I mean can't do statistics and as such have a really hard time even learning from statistics because the correct answer that humans get from statistical math by painstaking symbol manipulation runs counter to their built-in algorithms. They can't intuit when statistics are right or wrong, they can be easily deceived by them (including deceiving themselves), and in general they act like they are dumb with respect to them. Which, they probably are.

And on the other hand, humans are pretty darn efficient at walking, learning language, hitting things with a club, conceptualizing a spatial environment and throwing things. We don't know how they do it, but we do know that for the most part they aren't doing it in the high inefficient ways that we teach computers to do it by leveraging that they can do math and making them work out the physics.

There are definitely things that humans do that are intelligent that we can't teach computers to do.
 

Would your games be better or worse if you randomly determined DCs?

It ought to be pretty apparent that tying your shoes is typically easier than knowing trollish vulnerabilities. However, if you set the DC to a randomizer you could end up with a DC 25 to tie your shoes and 5 to know troll vulnerabilities. Because the AI can't reason out that tying your shoes ought to be easier than knowing about trolls. It has to rely on whatever data set and mathematical model it is using. And good luck finding a data set that covers every possible crazy question a player could come up with.
You don't need to do any of that stuff. You ask "does my character know X?". There are two possible answers: "yes" or "no". Without anything else to go on, the computer, or the human, just picks one at random. Since you can't see under the hood, how would you tell the difference between a random answer and a reasoned one?
 

Oofta

Legend
Honestly, the fact that a discussion on a possible DM shortage is bringing up AI DMs feels really icky to me. Isn't this supposed to be social game?
If I want to play a CRPG there are plenty of options. If you want a generated dungeon populated with monsters, you just need a few charts to do that, there are plenty of dungeon generators online if you wish.

But I agree, I play with a DM because I want more than just dungeons to explore and monsters to kill. Even if I did it with a group of friends.
 

Celebrim

Legend
Honestly, the fact that a discussion on a possible DM shortage is bringing up AI DMs feels really icky to me. Isn't this supposed to be social game?

It's said that porn drove the creation of the internet.

Well, games drove the creation of the personal computer. Bill Gates is famous for saying that no one would need more than a 640KB RAM, because you wouldn't need more than that for anything like word processing, spreadsheets, or other reasonable utilities.

Well, you need more than that for an RPG or to play Battletech the way the people who made the paper game really wanted. A lot of what we as gamers have built is out of frustration at not getting the social experience we wanted when we wanted it.
 

iserith

Magic Wordsmith
Then the human players always stick the AI character up front, he gets killed repeatedly, and the next thing we know something is lighting the candle on our nuclear arsenal.
DM: You can't do that - your character doesn't know that trolls are harmed by fire.
Player AI:

source nuke GIF
 

Fanaelialae

Legend
You don't need to do any of that stuff. You ask "does my character know X?". There are two possible answers: "yes" or "no". Without anything else to go on, the computer, or the human, just picks one at random. Since you can't see under the hood, how would you tell the difference between a random answer and a reasoned one?
If I ask whether I know how to tie my shoes and get a "no" response, and then ask if I know about troll vulnerabilities and get "yes", I'm going to quickly start to suspect that I'm getting randomized responses because that doesn't make rational sense.

Moreover, characters in many editions of D&D have proficiencies or skills that allow them know things. So a coin flip doesn't really work here.
 

Celebrim

Legend
You don't need to do any of that stuff. You ask "does my character know X?". There are two possible answers: "yes" or "no". Without anything else to go on, the computer, or the human, just picks one at random. Since you can't see under the hood, how would you tell the difference between a random answer and a reasoned one?

Talk to a chat bot for a while. After a while, you'll be able to tell whether it's giving you a random answer or a reasoned one. The more you know about how they work, the faster you can pin it down.

The same thing would happen even faster if you put a chat bot type "intelligence" in charge of your dungeon.
 

Oofta

Legend
Unless you're a certain ex-Google employee, no one believes that AI actually "thinks" in any way that we would consider it thinking. Without that it will never feel "real". AIs "learn" by feeding it a ton of material so that it can run algorithms to pick up patterns. Throw in various techniques to solve specific, narrow, problems and you have an AI. Some can do amazing things and using some randomization and evolutionary techniques even solve problems we have a hard time doing. For other things that we take for granted they struggle with or simply can't do it.

AI can emulate simple conversations because it has a databank of questions and responses gleaned from some source. An AI could be taught to learn the rules of D&D, just like it's been taught to learn chess. But holding a real conversation? You'll hit limits. Actually creating entire campaigns from scratch? Logically responding to what the PCs do with the flexibility that a DM can? Nope. I don't see it happening any time soon.
 


Remove ads

Top