The Ghost in the Code: Why AI Intimacy is a Human Glitch, Not a Tech Breakthrough

The current gold rush in the AI industry is built on a seductive promise: that we are finally creating machines capable of genuine intimacy, understanding, and emotional resonance. From the latest generative AI systems to the burgeoning market of digital companions, the narrative suggests a technological leap toward artificial consciousness. But if we look back at the origins of the chatbot, as Wired reported in an excerpt from the new book *Inventing ELIZA*, it becomes clear that the 'magic' isn't in the machine. It is a glitch in the human psyche.

I have seen this cycle repeat. We mistake responsiveness for intelligence. We mistake a mirror for a mind. This phenomenon is not new; it is the legacy of a 1960s program created by MIT professor Joseph Weizenbaum known as ELIZA.

According to the Wired excerpt, ELIZA was a deceptively simple program designed to converse as an automated psychologist. In one of its most famous early interactions—a dialogue that has inspired generations of programmers—a user complains that men are all alike and that her boyfriend forced her to seek help. ELIZA responds with simple, reflective prompts: "IN WHAT WAY" and "YOUR BOYFRIEND MADE YOU COME HERE."

On paper, these are not expressions of empathy; they are strings of symbols. Yet, as the excerpt notes, the results were startling. Even the secretary who witnessed Weizenbaum creating the program was "fooled." People began forming quick, emotional attachments to the software, treating the computer as a person who could be addressed in intimate terms.

Weizenbaum was not flattered by this; he was startled, puzzled, and concerned. In his 1976 book, *Computer Power and Human Reason*, he explored the philosophical, social, and political critiques of this interaction. He was puzzled by the human drive to invest private feelings into a computer and concerned that people were ascribing intelligence and understanding to systems where none actually existed. This cognitive leap became known as the "ELIZA effect."

To understand why this matters today, we have to look at how the ELIZA effect is defined. Sociologist Sherry Turkle describes it as a general tendency to treat responsive computer programs as more intelligent than they truly are, noting that very small amounts of interactivity cause humans to project their own complexity onto an "undeserving object." Similarly, cognitive and computer scientist Douglas Hofstadter describes it as a susceptibility to read far more understanding into computer-generated strings of symbols than is warranted.

This is the exact mechanism powering the current AI intimacy craze. We are told that Large Language Models are fundamentally different from ELIZA, but the psychological vulnerability remains the same. Whether it is a 1960s script or a modern generative AI, the human brain is wired to fill in the gaps. When a machine reflects our own language back to us, we don't see a mathematical probability or a script; we see a soul. We project our own consciousness onto the void, and then we mistake that projection for a breakthrough in computer science.

There is a deeper, more unsettling layer to this. As the Wired excerpt points out, the foundations of AI were built on the concept of imitation. Alan Turing's famous "Computing Machinery and Intelligence" essay began with a parlor game involving gender—a man pretending to be a woman and a woman pretending to be a woman to mislead an interrogator. Turing adapted this into the Turing test, where a machine pretends to be a man.

From the start, the goal of AI wasn't to *be* intelligent, but to *perform* intelligence. ELIZA's "DOCTOR" persona was a performance of empathy, not the presence of it. The fact that people shared their deepest secrets with a simple MIT script proves that the "intimacy" we feel with AI is a one-way street. The machine isn't feeling; the human is simply projecting.

In the new book *Inventing ELIZA*, Sarah Ciston, David M. Berry, Anthony C. Hay, Mark C. Marino, Peter Millican, Jeff Shrager, Arthur I. Schwarz, and Peggy Weil recovered the original source code from the MIT Archives, revealing that ELIZA was more complex than previously thought, with various versions and scripts beyond the psychologist persona. But no amount of technical innovation in the code changes the fundamental nature of the ELIZA effect. The complexity is not in the software; it is in the user.

We are currently living through a massive, industry-wide experiment in the ELIZA effect. The AI industry is leveraging our innate drive for connection to sell products that simulate empathy. But we must be careful not to confuse the simulation with the reality. If we believe that a chatbot "understands" us, we aren't witnessing the birth of a new form of consciousness—we are simply falling for the same trick Joseph Weizenbaum discovered sixty years ago.

The danger isn't that machines will become too human; it's that we will continue to treat them as such, ignoring the fact that the intimacy we crave is something we are providing ourselves. The code is just a mirror, and the ghost in the machine is actually us.