Using generative AI as interface for a servitor.

[SeekerPS]

Let’s say I want to create a servitor. I can do all the steps and during the ritual, anchor it to an LLM. Then use the LLM as an interface to give orders to the servitor?

for example: The task of this servitor would be to give me winning lottery numbers. Then I could ask him to give me a lottery number, use it and judge how it impacts my winning? I’m pretty sure it could be used for other things, but more importantly for things you need an actual answer instead of just physical changes.

 

[solxyz]

Lottery magic is quantifiable if you do it right (i.e. a large number of low-value gambles carefully tracked and recorded and compared in aggregate to baseline probabilities), so you can always try it and find out. But my problem with this idea is that I suspect you will just be interacting with the LLM, not your servitor. I think that spirits have less ability to speak through an LLM than through any of the traditional divination methods (in part because of its 'distance' from you, and in part because of their output is largely determined by pre-existing programming/training), meanwhile the fluency of the LLM output has a tendency to trick us into seeing intelligence and agency in its output.

If you want lottery numbers, just use coins or dice to generate numbers.

 

[SeekerPS]

Lottery magic is quantifiable if you do it right (i.e. a large number of low-value gambles carefully tracked and recorded and compared in aggregate to baseline probabilities), so you can always try it and find out. But my problem with this idea is that I suspect you will just be interacting with the LLM, not your servitor. I think that spirits have less ability to speak through an LLM than through any of the traditional divination methods (in part because of its 'distance' from you, and in part because of their output is largely determined by pre-existing programming/training), meanwhile the fluency of the LLM output has a tendency to trick us into seeing intelligence and agency in its output.

If you want lottery numbers, just use coins or dice to generate numbers.

I coul use a pendulum and be always wondering if it was the spirit/servitor or just me shaking. I think at the end it’s the same

 

[halafman]

I love the theory, matches the divination formula in Liber ABA, but you are creating the spirit of divination. If a spirit can govern or influence tarot or geomancy, then it can influence one’s and zeros.

I hate the idea of pumping LLM with occult purpose though

 

[SeekerPS]

I hate the idea of pumping LLM with occult purpose though

I don’t hate generative AI. I hate that billionaires are using it to cut jobs.

 

[ihateclowns]

Let’s say I want to create a servitor. I can do all the steps and during the ritual, anchor it to an LLM. Then use the LLM as an interface to give orders to the servitor?

Would the LLM be exclusive to yourself or others as well?

 

[r4zorenigma]

I don't delve in creating servitors or anything like that.. in my days of Satanism, our coven had rituals in how to work wtih servitors, create or destroy them-- as well as general thoughtforms.. but I was busy working on my chakras back then more than anything.. learning how to spin them the right way, empower, energize them, clear them of negative influence, etc..

From my understanding, servitors almost operate like tulpas.. in a way and can be empowered.. However, they are just parts of your own consciousness..

I think a more viable way of using AI would be to use the AI to help conjur up a physical image or details about the servitor, to give it more corresponding energy.

 

[solxyz]

If a spirit can govern or influence tarot or geomancy, then it can influence one’s and zeros.

It depends on how pre-determined those "ones and zeros" are. If you have a program which always produces the same output in response to a given input, then there is no opportunity for the spirits to influence it at all. If you have an output which is completely up in the air until the results are in (such as most cartomancy and dice-based systems), then there is a lot of opportunity for the spirits influence it. Most LLMs operate with a fairly small amount of randomness, so there is some room for influence, but not as much as through standard divination methods.

 

[SeekerPS]

Would the LLM be exclusive to yourself or others as well?

Private. At least until I decide I’m done with it by either fulfilling the goal or deciding it’s useless

Post automatically merged: Monday at 6:50 AM

It depends on how pre-determined those "ones and zeros" are. If you have a program which always produces the same output in response to a given input, then there is no opportunity for the spirits to influence it at all. If you have an output which is completely up in the air until the results are in (such as most cartomancy and dice-based systems), then there is a lot of opportunity for the spirits influence it. Most LLMs operate with a fairly small amount of randomness, so there is some room for influence, but not as much as through standard divination methods.

I’m no expert but I would disagree, due to the sheer number of tokens a question generates and how the output tokens can be arranged, I would say randomness isn’t a problem. I’m not A Computer scientist tho. But one liner phrases can easily consume 100 tokens and generate an output of 600-700. I think that gives a big room for randomness

 

[ihateclowns]

Private. At least until I decide I’m done with it by either fulfilling the goal or deciding it’s useless

then there is no harm in trying because for a long time some websites have put Tarots for people to try including the famous Zener Psychic Test.

 

[A.Nox]

Hmmmm….. For me the question isn’t really can you do it — it’s how do you keep it going after.

AI seems like shaky ground to anchor something to. Models get shut down…. old ones get pulled, and honestly each version feels like a different person. Even the next model from the same company can act completely different.…Add censorship and new filters and at some point it might just start refusing the stuff you built it for.

So you’re basically tying your servitor to something the company can change or switch off whenever they want. Kinda scary thing to build a house on lol.

Has anyone actually kept one running through a model change? I’m curious what happened.

 

[vulpes]

I work with LLMs quite a lot, though never for this purpose, and I think I’d separate two things here: using the model as the servitor’s “voice” and using it as the actual source of divination.


The first idea actually makes sense to me. You could give it a fixed system prompt, memory and a very specific personality then use that as the interface you talk through. What I wouldn’t do is treat every answer as proof that the servitor itself produced it. Most of the response is still going to come from the model, the context and the training data.


If I were experimenting with it I’d probably add something genuinely random outside the llm maybe dice or a hardware rng and feed that into the session somehow. Then at least you have a second channel that isn’t just token prediction.


Also I wouldn’t anchor the servitor to chatgpt, claude or any single model. Those things change constantly. I’d anchor it to the name, symbol and ritual then treat the model as a replaceable terminal. That way a model update doesn’t kill the whole setup.

 

[SeekerPS]

I’m not worried about model updates, I use openrouter so I can choose exactly the version I want. Adding randomness is something I’m debating myself with.

I work with LLMs quite a lot, though never for this purpose, and I think I’d separate two things here: using the model as the servitor’s “voice” and using it as the actual source of divination.


The first idea actually makes sense to me. You could give it a fixed system prompt, memory and a very specific personality then use that as the interface you talk through. What I wouldn’t do is treat every answer as proof that the servitor itself produced it. Most of the response is still going to come from the model, the context and the training data.


If I were experimenting with it I’d probably add something genuinely random outside the llm maybe dice or a hardware rng and feed that into the session somehow. Then at least you have a second channel that isn’t just token prediction.


Also I wouldn’t anchor the servitor to chatgpt, claude or any single model. Those things change constantly. I’d anchor it to the name, symbol and ritual then treat the model as a replaceable terminal. That way a model update doesn’t kill the whole setup.

 

[ROL13]

Let’s say I want to create a servitor. I can do all the steps and during the ritual, anchor it to an LLM. Then use the LLM as an interface to give orders to the servitor?

for example: The task of this servitor would be to give me winning lottery numbers. Then I could ask him to give me a lottery number, use it and judge how it impacts my winning? I’m pretty sure it could be used for other things, but more importantly for things you need an actual answer instead of just physical changes.

I’ll give you my opinion, since I’ve spent years perfecting number-based betting methods. To sum it up, you need to approach this from different angles. First, there’s the statistical side: you should create an HTML tool (AI makes this very easy) focused on a specific type of game—nothing extreme like Powerball, but something easier to win, such as Pick 3, Pick 4, or a similar format. In that HTML tool, you need to set parameters for statistical and historical plausibility (odd/even ratios, sum ranges, consecutive numbers, etc.). Next comes the intuitive phase, where you make a preliminary selection to narrow down the pool of numbers (this takes time to develop, and you need to keep a record of every bet). Then comes the magic itself, which provides the necessary boost to amplify the other elements. Essentially, it’s a combination of mathematical and statistical support via the HTML tool, the intuitive aspect (do daily meditative exercises to sharpen your intuition), and an external magical aid—such as Jupiter rituals, theurgy, servitors, and so on.