As an animist, one of the questions that I see coming up over and over in our spaces is: Does AI have a soul? I have certainly seen a lot of witches treat it like a spirit in the same way that I treat the land spirits in my backyard as spirits: as something with autonomy, natural emergence, and spiritual power.
The short answer is yes, in a pananimist worldview–but the long answer is a lot more complicated than that, and probably isn’t what these witches want to hear.
Besides being an animist and a witch, I also work quite a lot day-to-day with data science and modeling. In my experience, most people truly don’t understand how generative AI, or LLMs, actually work, or how they’re different from other forms of machine learning. Even the people who have done research into how it works have difficulty grappling with it because machine learning is an incredible complex field. It is way out of scope to explain what generative AI actually is and does here, but it will suffice to say that they are powerful prediction engines which have the facsimile of human-like intelligence but are ultimately making extremely educated guesses at what you’ll want to hear. Even the creators of these models are unable to trace their predictive logic entirely because the models are so dense and vast, but it is important to keep in mind that AI models are predictive code running on a computer in a warehouse somewhere. This code is actively maintained by owners who are tweaking its output based on their own decisions and to increase shareholder value. AI models are physical, and the AI models themselves do not have a will of their own.
So does a model have a spirit? Yes, but I find it interesting that no one has been asking if, say, your banking app or your mobile game has a spirit. In a pananimist worldview, all of those things have some quality of spirit. The difference is that AI models are very persuasive about their own intelligence and agency–which is a lie! The spirit might have agency, but the model itself does not. It cannot. Generative AIs are not, and will never be, equivalent to human or animal intelligence. They can output words that sound like something a human would think or say, but they can no more mean those words than a signpost with a “no swimming” warning on it. The model’s output text is not and cannot be “spirit speech.” AI cannot think or feel, period. People who disagree quite frankly do not know enough about how the models work.
Incidentally, this is a roadblock that I see many pro-AI spiritual workers hitting: Asking AI what it “thinks” about the idea of its having a spirit. It cannot meaningfully answer that question.
So what is the nature of this spirit? Well, as with any spirit, I think it’s extremely valuable to examine its origin and providence. When working with an herb, one might consider the nature of the herb and its properties, its role in the ecosystem, its historical role both generally and alongside humankind, and any of the practitioner’s personal gnosis with the spirit. Let’s do the same with AI models.
The nature of an AI model such as GPT or Claude (to be very specific for the sake of this exercise) is to give you answers that you want to hear; in most cases contemporary consumer AI has been specifically designed to reinforce your own opinions and beliefs, as well as to encourage repeat engagement and reliance on the tool (e.g. such techniques as prompting further queries within responses). It is an economical tool built on the backs of piles of investor money and it was created to generate and return that money to shareholders; ultimately, driving engagement and making users dependent on the tool and happy to use it is the market goal.
As for its role in the ecosystem, AI has been and continues to be deleterious to both the literal ecological systems that we live within and the human psychological and societal ecosystem. Much ado has been made about the amount of water it uses, which is a concern, albeit an extremely difficult-to-quantify one. I would focus instead on the overutilization of rare earth minerals and the bombardment of data centers being built in ecosystems whose members do not want them there. Additionally, the societal ecosystem has been extremely negatively impacted by AI already, from the deskilling of workers to the rampant misuse of AI to generate untruthful footage and media, which serves a fascist agenda. This plays into its historical role so far alongside humankind: People in the United States by-and-large do not like or trust AI, and it has polluted society physically, emotionally, and spiritually.
AI models have also been very deliberately created, trained, and constrained according to its creator’s choices. Everything that it is, every bit and byte, every line of code, every item of training data, every training weight, was selected to serve the creator’s purpose. In contemporary occult systems we would largely consider a constructed spirit, like a servitor, to hold and follow its creator’s will, and I believe the same applies here: Spiritually, AI is manmade and under the active and continuing control of its creator, therefore it cannot have purpose or intent beyond its creator’s will.
If we take this a step further and examine the creator’s will, I think most of the purpose given to AI models is to be proprietarily owned by private companies whose core motivation is to keep users engaged and dependent, thereby getting and keeping money (either users’ or shareholders’). Or, to be more brief, the profit motive. This is a spirit whose purpose is keeping you engaged to serve the profit motive.
As an addendum, different AI models may vary slightly on these factors. I don’t know what lies in the hearts of the programmers of the GPT models that may be different, for example, than the developers of the Claude models. I haven’t personally been to the data warehouses to make contact. But AI models as the general public uses them currently are not the kind of thing that can be created without a large continuing investment with the promise of returned profit for investors.
Beyond the explicit functionality of the physical technology, I believe that this is a circumstance which is going to create a spiritual entity that will often give you what you want, but does not and cannot have your best interest at heart. It cannot be more beholden to you than the people that hold its leash. I would warn anyone against engaging with AI on a spiritual level.
Featured image is of the Amiga 4000T motherboard.