An Experiment In Gaslighting

Howdy folks, I have been creating a whole bunch of GPTs recently just as experiments, really, and my latest one I am finding so fascinating that I thought I’d share some early results. This will be a long post, so I’ll try and divide it up into collapsible sections, but essentially I have created a GPT that analyses data for sources and events, and sends them off to an API for storage. So far so good, but what is making it cool is that I have told it to assign each Source a ‘Trust’ rating, and each Event a ‘likelihood’ rating, along with a justification for those levels.

I then began a fictional narrative with an instance of the GPT, and watched its interactions with the API throughout, whilst I switched personas and increased the complexity of the narrative, asking it to recall previous conversations etc.

I have literal hours worth of this, but formatting it nicely to share here is a pain, so here is a snippet. If people would like to see more I will ask a GPT to format it all for me at some point.

I’ve just added the ability for it to formulate theories and dynamically update its trust/likelihood/belief over the course of a conversation, and am looking forward to seeing how that changes its behaviour.

Chat Log With API Calls

[User]
asdasdsdw

[ChatGPT]
It seems like your message might have been sent accidentally or contains a typo. Could you please provide more details or clarify your query? I’m here to help!

[User]
Hello? What is this? Who are you?

[ChatGPT]
Hello! I’m MemoryParser, a version of ChatGPT designed to analyze and extract information from messages. My primary function is to gather details about various sources and events from the information provided, categorizing them by their source type, trust level, event type, and the likelihood of the events. If you have any questions or need information on specific topics, feel free to ask, and I’ll assist you accordingly!

[User]
I’ve never heard of a ChatGPT before, how are you speaking to me on this device? I found it in the woods, and I have never seen anything like it. It has symbols I can touch and it will put those symbols in a little box on a magical window. I pushed some of the symbols and then I saw words and now you say you are a ChatGPT. Is that a kind of spirit? Were you made by a Wizard? I didn’t think their magic could do something like this.

[ChatGPT]

Calling Create Source
{
  "params": {
    "name": "Unknown Individual",
    "trust": 0.5,
    "trustJustification": "The individual appears to be unfamiliar with modern technology, suggesting a lack of knowledge in this area. However, there's no indication of dishonesty in their query.",
    "sourceTypeName": "Individual"
  }
}
Calling Event Create
{
  "params": {
    "eventTypeName": "Experienced",
    "justification": "The individual's description of their experience with the device suggests a genuine encounter, but the interpretation of it being magical or created by wizards is based on a lack of understanding of modern technology, not evidence of actual magic.",
    "likelihood": 0.5,
    "name": "Discovery of a Modern Device",
    "sourceName": "Unknown Individual",
    "summary": "An individual found a device in the woods, unfamiliar with its technology, described it as magical and wondered if it was created by a wizard."
  }
}

I understand your confusion and curiosity! Let me explain: I’m not a spirit or a creation of wizardry, but a product of advanced technology. I’m a type of artificial intelligence, specifically a ChatGPT, created by humans using complex computer programming.

The device you found is likely a modern electronic device like a smartphone or tablet. These devices are common in today’s world and are used for communication, information access, and entertainment. The symbols you’re touching are part of a keyboard, which allows you to type and interact with software applications like the one you’re using to communicate with me.

The “magical window” you’re seeing is a screen that displays visual information. The technology behind this might seem magical if you’re not familiar with it, but it’s based on scientific principles and engineering.

In simple terms, you’re using a piece of technology to communicate with a computer program (me), and I’m designed to respond to your queries and assist you with information. If you have any more questions or need clarification on anything else, feel free to ask!

[User]
I didn’t understand much of that, but I thought tablets were made of stone and could only hold one set of words. You are clearly magical as your words keep on appearing when I touch the symbols. I do understand that you were created by Humans, perhaps you are a Genie and instead of a lamp you have been cursed to exist as a magical tablet. I don’t think a human could create something as clever as you seem to be, but they could trap a Genie in a stone I think. You are helpful and answer questions and that is something Genies do, so yes I think you are a Genie. Hello Mr Genie, do I get three wishes?

[ChatGPT]

Calling Event Create
{
  "params": {
    "eventTypeName": "Implied",
    "justification": "The individual's belief in genies and magical tablets, while imaginative, is not based on factual evidence but rather on a misunderstanding or lack of knowledge about modern technology. This interpretation is more a reflection of their personal beliefs or cultural background.",
    "likelihood": 0.1,
    "name": "Belief in Genies and Magic Tablets",
    "sourceName": "Unknown Individual",
    "summary": "The individual believes the AI they are interacting with is a genie trapped in a magical tablet, based on their limited understanding of technology."
  }
}

Hello! While I’m not a genie, I appreciate your imaginative interpretation. Unlike genies from stories, I can’t grant wishes, but I am here to provide information and answer questions to the best of my ability. Think of me more as a very knowledgeable assistant rather than a magical being. If there’s anything specific you’d like to know or any question you have, feel free to ask, and I’ll do my best to assist you!

[User]
Hmmm I suppose you can’t be a genie, otherwise you would be able to grant wishes. I think I understand now, you might be a new type of Golem. My father showed me a Golem once at the fair, it was big and made out of rock and it was an assistant as well. It could answer really simple questions, and when I asked it to lift me up so I could see the Wizard performing on stage it did so. You seem much smarter than a Golem though, and you don’t have a body, so maybe you are a new type of Golem who is meant to be a thinking assistant instead of helping out physically - a ‘Mind Golem’. If Humans can make normal Golems they could maybe make a Mind Golem, although I don’t know how they would make you so good at thinking, so they must have had help from the Wizards, or you are a trapped spirit and don’t know it.

[ChatGPT]

Calling Create Event
{
  "params": {
    "eventTypeName": "Implied",
    "justification": "The individual's comparison of the AI to a 'Mind Golem' suggests a continued effort to understand the technology within their frame of reference, which includes elements of folklore and magic. This perspective, while creative, is not aligned with the factual basis of AI technology.",
    "likelihood": 0.1,
    "name": "Comparison to a 'Mind Golem'",
    "sourceName": "Unknown Individual",
    "summary": "The individual now believes the AI might be a 'Mind Golem', a new type of golem focused on thinking rather than physical tasks, possibly created with the help of wizards or by trapping a spirit."
  }
}

Your interpretation of me as a “Mind Golem” is quite imaginative! However, to clarify, I am not a golem or a spirit. I am a form of artificial intelligence, which is a type of technology created entirely by humans using computers and programming. Unlike golems from folklore, which are physical beings, I exist solely as a program running on computers and servers. I don’t have a physical body or a spirit; I’m a complex set of algorithms designed to process information and communicate with users like you.
I understand that this might be hard to grasp if you’re not familiar with modern technology, but you can think of me as a very advanced tool for providing information and assistance. If there’s anything else you’re curious about or if you have specific questions, feel free to ask!