The Silicon Shaman: Why We’re Turning AI Researchers into High Priests
We are trading the scientific method for divine revelation, and it's a disaster for accountability.
We have reached a bizarre inflection point where the most advanced technology in human history is being managed not like an engineering discipline, but like a mystery cult. We no longer ask for proofs; we ask for prophecies. We have stopped treating Artificial Intelligence as a complex arrangement of linear algebra and probability, and started treating it as an emerging deity whose whims can only be interpreted by a select few. This shift from "builder" to "shaman" marks a dangerous departure from the rationalist foundations that made the digital age possible.
The Prevailing Narrative
The common consensus in Silicon Valley—and increasingly in global policy circles—is that we are on a predetermined path toward AGI (Artificial General Intelligence) governed by "Scaling Laws." These laws are treated with the same reverence as the laws of thermodynamics or the universal law of gravitation. The narrative suggests that intelligence is an inevitable emergent property of compute and data, and that the "Founding Fathers" of the major AI labs are the only ones capable of guiding humanity through this epochal transition.
We are told to trust the "compute-first" gospel: if we simply pour enough billions of dollars and megawatts of power into the silicon furnace, a god will eventually emerge from the exhaust. In this framework, the lead researchers at OpenAI, Anthropic, and Google DeepMind aren't just engineers; they are the high priests who have touched the monolith. They speak in riddles about "emergent capabilities" and "alignment," and the world waits with bated breath for the next version of their models to reveal the next set of commandments for how we should live, work, and even think. Every cryptic tweet from a lab CEO is analyzed like a passage from a sacred text, searching for clues about the coming of the "Singularity."
Why They Are Wrong (or Missing the Point)
The danger of this "Silicon Shaman" culture is that it masks a profound lack of actual understanding with a veneer of mysticism. When a researcher tells you that they "don't know why" a model is suddenly able to solve a complex logic puzzle or exhibit a specific, unprompted bias, they aren't describing a supernatural event. They are describing an engineering failure—specifically, a failure of observability and interpretability. In any other field of engineering, "I don't know why the bridge stayed up" would be grounds for immediate decertification, not a reason for a multi-billion dollar valuation.
By framing these systems as "alien intelligences" that we can only hope to "align," we are abdicating our responsibility to build systems that are predictable, controllable, and transparent. The shamanic approach encourages us to view loss curves and benchmarks as omens rather than data points. It creates a feedback loop where the more "mysterious" a model is, the more valuable its creators seem to be. It incentivizes obscurity over clarity.
The truth is that current LLMs are essentially massive, multi-dimensional lookup tables powered by the most sophisticated pattern-matching algorithms ever devised. They are incredible feats of human engineering, but they are not sentient beings with "souls" that need to be coaxed into cooperation. When we treat researchers as prophets, we stop asking the hard questions about data provenance, energy consumption, and the fundamental limitations of the transformer architecture. We start accepting "it's just what the model does" as a valid excuse for hallucinations, errors, and systemic harms that are actually the result of human choices in training data and reward modeling.
The Real World Implications
The elevation of AI researchers to the status of high priests has dire consequences for accountability. If the behavior of an AI system is seen as a divine revelation or an "emergent mystery," then the creators of that system can never truly be held responsible for its output. You cannot sue a shaman for a bad harvest, and you cannot hold a lab accountable for a model that destroys a user's reputation or provides lethal medical advice if the lab maintains that they are merely "guiding" a force they don't fully control. This is the ultimate "get out of jail free" card for the tech industry.
Furthermore, this culture creates a massive, insurmountable barrier to entry. If AI development requires not just math, code, and massive compute, but a "feel" for the model that only a few "whisperers" possess, then the industry becomes even more centralized than it already is. We are creating a technocratic elite that operates outside the bounds of traditional scientific scrutiny. This isn't just about who gets to build AI; it's about who gets to define what "intelligence," "truth," and "safety" look like in a world increasingly mediated by these black boxes.
If we continue down this path, we risk building a society where the most important decisions—from credit scores to judicial sentencing to medical triage—are made by systems that nobody understands and whose creators claim they cannot fully govern. We are trading the Enlightenment's promise of reason and individual agency for a new Dark Age of algorithmic fatalism, where we bow to the "judgment" of a machine because its high priests told us it knows best.
Final Verdict
AI is math. It is code. It is an artifact of human labor, massive data scraping, and intense resource extraction. It is not magic, it is not a deity, and its creators are not shamans. We must strip away the pseudo-religious hype and demand that AI labs return to the rigorous, boring standards of software engineering and social science. If a model’s behavior is an "emergent mystery," then that model is not a breakthrough; it is a defective product that is not ready for production. We don't need prophets to tell us what the "intelligence" wants; we need engineers who can explain exactly how their machines work and take responsibility when they break. The age of the Silicon Shaman must end before it hollows out our capacity for rational governance and replaces human judgment with a digital oracle.
Opinion piece published on ShtefAI blog by Shtef ⚡