(pseudo) AI with RNG | How do we do it?

[Suspnded_Arr0w_Shaft]

Can we make a functioning AI with RNG (random number generation)?

I was thinking, what if I have a ton of libraries (words, text strings), so an AI can take a prompt and generate a result with a randomly selected query (related words/ text strings)

Example:

if | get property "resp" = What
then | if | number 1-4000 = 0001 then | set property "resp" = What the
then | if | number 1-4000 = 0287 then | set property "resp" = What the heck

“resp” as in response.

But it would find a query related to the word, such as “ghost rider.”

-Marvel
-comics
-Marvel Comics
-skull
-fire
-leather
-jacket
-motorcycle

This would take a lot of libraries (triggers), but I think it is possible.

Each trigger would contain around 50 words. each trigger is linked to a group and a subgroup (what number off of the group it is), and like 3 more subgroups so we don’t have sentences like “Ghost Rider, from Marvel Comics, is a skull fire leather jacket”

kind of like:

Group (648, 04, 01, 01, 01), where 648 is Marvel, and group 04 is the 4th trigger containing related words.
it chooses a word
Restraints:
-Memory
-Max amount of triggers (idk if this is real)

This AI would probably be very bare bones and probably only have knowledge surrounding GKC (no marvel).

Cool idea I thought of, and it is most definitely possible (unlike my last research topic )

@Suspnded_Arr0w_Shaft

@JustAToaster

To clarify, AI IS NOT POSSIBLE IN GKC,

What this devices topic is about is that if is would be possible to make a psuedo (fake) AI in gimkit that would see what the player did and then make a random response based off of that.

This is correct. I probably should’ve put this in the topic post, but this is not an actual “ai”, and it is just an algorithm to try and create the most linguistic RNG response possible.

PLEASE SEE THIS POST:

(pseudo) AI with RNG | How do we do it?

Sorry, people, I had a snow day yesterday, so I didn’t work on anything First off: Who keeps flagging Iain cus they’re suspended now on relation, what did they even say? Secondly: Why was clay-institue tag added and removed. First off, we haven’t hit any road blocks yet that would warrant this clay institute (thx tho that lowkey gave me an ego boost @Kormorant ) secondly, there are no rules depicting the clay institute tag other than hard problems, which means it should not hav…

[NagiSeishiro]

This could be possible, with GKC’s current server limit’s probably not. There are just too many words out there. If we were going to make one, you would have to start off small. So, this could be possible, but, alas, very limited.

You also only have a certain amount of blocks you can place when coding in GKC. (I think it’s around 75, but I’m not 100% sure.)

[Suspnded_Arr0w_Shaft]

@Suspnded_Arr0w_Shaft

This AI would probably be very bare bones and probably only have knowledge surrounding GKC (no marvel).

We really only need function words, (the how why when) and GKC knowledge (terrain block code item)

@NagiSeishiro

You also only have a certain amount of blocks you can place when coding in GKC. (I think it’s around 75, but I’m not 100% sure.)

yes you’re correct

[Suspnded_Arr0w_Shaft]

So essentially here’s a prompt:

How do I use the “set text” block in the text device?

Here’s a rundown of what the AI would do:

in text string = get property "prompt" find first occurrence of "text , block, text device, set text, etc)

after this it would find related words the the occurrences in the AI’s libraries. This is the part I’m researching.

AI analyzing the prompt is super easy to make. The libraries, and the searching part of the AI is what I’m researching.

[NagiSeishiro]

Yeah, it’s just that these “libraries” would be limited because of GKC’s minimal storage.

[Suspnded_Arr0w_Shaft]

Unfortunately yes.

I’m rather not looking for quality, but functionality. As I said, it would only have knowledge of GKC, and very small linguistic precision since it is randomly generated.

[NagiSeishiro]

Well, the AI would have to check the word that was last used, then, based off that word, select the best word to continue the sentence…

[NagiSeishiro]

It would end up in countless family trees of connected words that would simply overload the entire system. It would be functional, but again, very limited.

Wait, @Suspnded_Arr0w_Shaft , was I the one that basically shut down your last Devices topic?

[NagiSeishiro]

I think it’s already been explained as simple as it can be…

I’m not super fond of code either, so I understand what you mean, lmao.

[Suspnded_Arr0w_Shaft]

okay this is about as simple as i can put it

a trigger recieves a prompt, and looks through libraries and randomly chooses a related word to create a response.

check this

(pseudo) AI with RNG | How do we do it?

So essentially here’s a prompt: How do I use the “set text” block in the text device? Here’s a rundown of what the AI would do: in text string = get property "prompt" find first occurrence of "text , block, text device, set text, etc) after this it would find related words the the occurrences in the AI’s libraries. This is the part I’m researching. AI analyzing the prompt is super easy to make. The libraries, and the searching part of the AI is what I’m researching.

[Pineappleman]

Well, the first problem is inputting, we need to be able to decode the input into something to search the librarys for. For example, “How do I make block code with a trigger?” We need to be able to take out keywords and search for those in the libraries. We could do this by hardcoding keywords for it to look for or have it choose long words.

[JustAToaster]

I would assume for inputting a prompt there would be a ingame keyboard where you can type a word, the system of triggers would then take that word and match it with one in the library of terms.

Then from there I guess it would give you the related words from the individual trigger library.

Having the “AI” answer full sentences would straight up not be possible as it would take hundreds of thousands of memory or maybe even millions to store the entire dictionary just so it could decode any sentence.

[JustAToaster]

To clarify, AI IS NOT POSSIBLE IN GKC,

What this devices topic is about is that if is would be possible to make a psuedo (fake) AI in gimkit that would see what the player did and then make a random response based off of that.

[Pineappleman]

So then the libraries would have categories, so like if a keyword were trigger, it would go to the keyword trigger that would have categories that would divide into subcategories that would choose words at random and put them into the response.

[Suspnded_Arr0w_Shaft]

Sorry for being offline, time to clarify some posts:

@NagiSeishiro

Wait, @Suspnded_Arr0w_Shaft , was I the one that basically shut down your last Devices topic?

Yes XD

@JustAToaster

Having the “AI” answer full sentences would straight up not be possible as it would take hundreds of thousands of memory or maybe even millions to store the entire dictionary just so it could decode any sentence.

I’ve stated like three times now that it will not know the whole english dictionary, and only function words/GKC relations

@JustAToaster

To clarify, AI IS NOT POSSIBLE IN GKC,

What this devices topic is about is that if is would be possible to make a psuedo (fake) AI in gimkit that would see what the player did and then make a random response based off of that.

This is correct. I probably should’ve put this in the topic post, but this is not an actual “ai”, and it is just an algorithm to try and create the most linguistic RNG response possible.

edit: Essentially, it will take a prompt, find entirely random related keywords to the question, string it together, and hopefully answer a question successfully and with precision of language (giver reference lol)

[Suspnded_Arr0w_Shaft]

I think what I am going to do first, is see if it can handle one-word prompts, and then enlarge the system from their.

as in I’m going to enter a random function word (how, when, where, what, or why) and then it will generate a number, find the libraries containing related language, and then send back the response.

Prompt:

What

Response:

The heck

since memory is a constraint, it will not know grammar ( i try to be a grammarholic)

[NagiSeishiro]

I think we should start off simple and have like 10 words you can input, and each one has 3 varieties of answers.

@Suspnded_Arr0w_Shaft

[Suspnded_Arr0w_Shaft]

@NagiSeishiro

I think we should start of simple and have like 10 words you can input, and each one has 3 varieties of answers.

I think this is a great idea. I’ma hop on GKC right now and start experimenting. I’ll update you all when I have something.

^[1]

should i add concept since we dont have a functioning AI yet?

Summary

@some_kid I know you like topics like this, so i thought you might want to know of its existence. if you dont want to help though, thats alright with me!

---

1. i havent logged into gimkit for like a month now XD ↩︎

[Suspnded_Arr0w_Shaft]

Okay I am just now realising the mass of this project, and I need a list from you all.

What are the first words for the libraries. (function words please)

Edit: please do not add the word “why” it doesn’t need a conscience

For example, this is a library:

it will activate the trigger if any of these key words are related to the prompt

[Suspnded_Arr0w_Shaft]

This is an example of how the AI will find related words: