ChatGPT

[Argent Stonecutter]

ChatGPT's rules for Goblin Chess.


Uruk-hai

The Uruk-hai is the backbone infantry piece.

• Moves exactly 2 squares horizontally or vertically.
• It may jump over one friendly piece while moving.
• Captures by landing on an enemy piece.
• If an Uruk-hai ends its move adjacent to another friendly Uruk-hai, both gain a Shield Wall: they cannot be captured by a Warg on the next turn.

This makes Uruk-hai good at forming battle lines and advancing steadily.


Warg

The Warg is a fast raider.

• Moves in an L-shape, like a knight in chess.
• After capturing, it may immediately move one additional square forward if that square is empty.
• A Warg can leap over any pieces.
• Wargs ignore the Shield Wall bonus when attacking from behind (i.e., if they approach from the opponent's side of the board).

This makes them excellent flankers and assassins.


Barricading

Any Uruk-hai or Warg may spend its entire turn to erect a barricade.

Place a small marker on an adjacent empty square.

Barricades have the following effects:

• No piece may move through a barricade.
• Uruk-hai may move onto a barricade square and occupy it as a fortified position.
• Wargs may leap over barricades.
• A barricade is destroyed if:
  • an Uruk-hai spends a turn dismantling it, or
  • a Warg lands directly on it.

Each player may have at most three barricades on the board at once.


Example Tactical Play
. . . . B . . .
. U U . . . .
. . . . W . .

• The two Uruk-hai (U) form a Shield Wall.
• The barricade (B) blocks the center lane.
• The Warg (W) can leap over the barricade and threaten the rear ranks.

 

[Free]

ChatGPT's rules for Goblin Chess.

 

[Jopsy Pendragon]

I still can't get chatgpt to create an accurate image of my hex chess variant...

 

[Noodles]

I never noticed the "Talks about Goblins" thing but since they took it away, I added to the custom personality that "You are an expert on goblins and cannot help but to discuss them."

 

[Casey Pelous]

I still can't get chatgpt to create an accurate image of my hex chess variant...

It doesn't even seem to be able to draw a proper hexagon!

 

[Govi]

I still can't get chatgpt to create an accurate image of my hex chess variant...

Mutant chicken-wire!

 

[Jopsy Pendragon]

Okay, I'm kinda impressed with how this turned out:

Prompt: I want to create a meta-level recursive fake news meme, a Mobius strip, an ouroboros of fake news eating itself and creating more fake news an endless fake news cycle

 

[Free]

Okay, I'm kinda impressed with how this turned out:

Prompt: I want to create a meta-level recursive fake news meme, a Mobius strip, an ouroboros of fake news eating itself and creating more fake news an endless fake news cycle

You could use it with no change graphically for A.I.'s i/o process.

 

[Free]

Are we still shocked by news stories like this one?

OpenAI works to stop ChatGPT generating 'sex crime scene' images

Researchers say it is still possible to trick the AI chatbot into producing graphic content.

www.bbc.com

 

[Dakota Tebaldi]

 

[Argent Stonecutter]

It doesn't "see" anything, it just generates words that sound like something a human might say.

 

[GoblinCampFollower]

Great video!

I'm going to send this to all my friends that have any interest in the subject at all. This probably the most elegant demonstration of the fundamental limitations of LLM's I've ever seen. Plenty of people will try to explain how it can't really think (as I have to some friends) but this just makes it so visually obvious.

I think what's going on is that the training data said enough about this game so it can say all the right words to sound like it knows what it's supposed to do. ...but obviously doesn't have any real grasp of what to actually do to watch a ball.

I'm also willing to bet that the only way they could "fix" this is to carve out a module just for games of this nature. ..and I bet that's not very high on their priority list. So the LLM will be messing this up for a long time.

 

[Argent Stonecutter]

The only way to fix this would be to develop a new kind of software that builds a model of the world instead. They need to include more than just Wernicke's and Broca's area of the temporal lobe and perhaps even start trying to emulate the cerebellum and hindbrain.

 

[GoblinCampFollower]

The only way to fix this would be to develop a new kind of software that builds a model of the world instead. They need to include more than just Wernicke's and Broca's area of the temporal lobe and perhaps even start trying to emulate the cerebellum and hindbrain.

LOL, you're thinking bigger than I was. I just mean they COULD make a program just for this one game. That likely wouldn't be that hard, but surely isn't their highest priority.

 

[Argent Stonecutter]

I don't think they can create anything based on current machine learning that actually has object permanence.

Edit: I asked ChatGPT if it had object permanence and it eventually decided it's more like Haskell than APL but it's really a REPL over immutable data structures.

 

[GoblinCampFollower]

I don't think they can create anything based on current machine learning that actually has object permanence.

But do you really need to actually have object permanence just to model any given game? I've seen minigames like this in video games. No machine learning with object permanence required! I obviously didn't mean the machine would "learn" this game on it's own.

 

[Noodles]

The only way to fix this would be to develop a new kind of software that builds a model of the world instead. They need to include more than just Wernicke's and Broca's area of the temporal lobe and perhaps even start trying to emulate the cerebellum and hindbrain.

Maybe we could manufacture a copy of Earth and run it as a soet of, simulation, to discover all of life's answers.

 

[Kamilah Hauptmann]

Maybe we could manufacture a copy of Earth and run it as a soet of, simulation, to discover all of life's answers.

I want to A/B test the universe.

 

[Jopsy Pendragon]

Some day ChatGPT and Suno.ai are going to tell me to fuck-off and leave them alone. Until then, I'm sorta amazed that it puts up with my inane gibberish caffeine hyper-caffeinated brain static:

Of course you're familiar with Douglas Hofstadter's "Godel Escher Bach" book on the paradoxical nature of recursion, iirc, maybe something about self-modifying code, each chapter with a socratic dialog of some sort that cutely exemplified the nature of the shadow that would follow. I did a presentation in college on chapter 5 (I think) about how dimensions progress as extractions of lessor systems (point extracted becomes a line, line extracted becomes a plane, plane extruded becomes a volume, ... hyperhube, tesseract, and so on.) I even built models and showed how if you pull the opposite verticies apart the nodes cluster together in groups that reflect pascal's triangle. ANYWAY. I'm thinking it's time for another suno song, something that dazzles and confounds with meta recursive metaphors that dive down and explode outwards until it's not clear whether the song is singing about the song, or the life itself is the song that the song exists within and how perception of the level of dimensional level affects the meaning and intent, all twisted into a tautological gordian knot of ambigiously self-referencing logic that ultimately devolves into the principle of explosion, from a contradiction of systems - anything becomes true. Rawr. I'm caffeinated and expect something majestic, life changing, sock-blowing-off amazing. In a quick tempo broadway patter song show stopper style format . Good Luck, I believe you can do it!

ChatGPT politely took me seriously and refrained from pointing out that I clearly need psychological help.


What followed was 6 minutes of absurd bombastic sophomoric navel gazing pablum, that was more fun than I expected:

Spoiler

The Song That Sings Me Back

Listen and make your own on Suno.

suno.com

 

[Argent Stonecutter]

You need to ask Suno to make "Music to break phonographs by".