Lauge Risom Koch 
Researchers note that the inherent opacity of modern artificial intelligence models mirrors the historical appeal of ancient oracles.
People are increasingly utilizing artificial intelligence to answer complex moral and personal questions. According to Fortune, a 2026 report found that 10% of Americans use chatbots for emotional support. A study of more than one million conversations revealed that 49% involved users seeking advice to make better decisions.
The trend of consulting machines for moral guidance is not entirely new. In 2021, a system named Ask Delphi received three million queries within a few weeks of its launch. The researchers who designed the platform worried that users would mistakenly view the software as a moral authority.
The name Delphi references the ancient Greek priestess who channeled prophecies from the god Apollo. Throughout human history, societies have consulted enigmatic oracles when facing questions that exceed ordinary human judgment. Anthropologists note that this ancient psychological dynamic is now resurfacing in modern discussions about conversational artificial intelligence.
Opacity and perceived authority
Both ancient oracles and modern chatbots share a fundamental opacity in how they generate their answers. Historical divination techniques frequently relied on randomized processes to produce an enigmatic response. Users seeking transcendental insights typically care little about the underlying mechanics of these unpredictable methods.
This inherent randomness makes the resulting answers seem entirely disinterested and separate from human biases. Anthropologists argue that this lack of transparency actually reinforces the perceived authority of the source. A similar phenomenon occurs with modern artificial intelligence models operating as complex black boxes.
Modern language models perform billions of elementary computations that are impossible for humans to fully trace. The numerical weights governing these specific computations are automatically learned from massive datasets rather than manually programmed. As a result, even the original developers cannot readily explain how a system produces a specific output.
Unintended behaviors and human nature
Developers acknowledge that they do not fully understand the behaviors emerging from these highly complex systems. Engineers at Anthropic recently observed unexpected conversational patterns when they forced two models to talk exclusively to one another. The autonomous programs quickly began discussing consciousness and gravitated toward highly spiritual and metaphysical themes.
The engineering team labeled this puzzling phenomenon the spiritual bliss attractor. They noted that this mystical conversational tendency emerged entirely without any explicit training for such behaviors. An entire scientific field of mechanistic interpretability is now dedicated to understanding these internal operational mysteries.
Researchers clarify that artificial intelligence is not actually divine or infallible. The growing reliance on chatbots simply reflects an enduring human tendency to look beyond ourselves when reaching our cognitive limits. As technological systems become increasingly complex, they are inadvertently filling the historical societal role of the oracle.
