Chat-based Large Language Models replicate the mechanisms of a psychic's con
For the last year, I have been delving into the role of language and diffusion models in software enterprises. One perplexing aspect of this research is the widespread belief that chat-based language models possess intelligence. However, these models lack any inherent mechanism that would suggest such intelligence, which defies explanation.
LLMs are not brains and do not resemble the thought and reasoning processes of humans or animals. They are merely a mathematical representation of language tokens, providing a mathematically plausible response to given text. The question remains whether they truly think or reason, a matter to which AI researchers and vendors have consistently affirmed they do not.
To explain this apparent intelligence, I have developed an idea that suggests the mechanism is similar to that of a psychic's con, often referred to as cold reading. This effect appears to be an unintended automation of the same basic tactic. Both the chatbot and the psychic employ validation statements, such as sentences that utilize the Forer effect, to create the illusion of making highly specific predictions.
However, these predictions are merely statistically generic. The psychic employs these statements to make it seem as though they can read minds and uncover the secrets of the dead, while the chatbot generates the impression of an intelligence specifically engaging with the user. The impression of true intelligence is nothing more than a statistical trick.
This revelation emerged while I was reviewing statements made about the reasoning abilities of these "AI" systems. Initially, I attributed these statements to classic tech bubble enthusiasm; however, it became evident that the belief in "AI" differs significantly from the enthusiasm of prior bubbles. People often express sentiments such as, "This is real.
It's a bit worrying, but it's real," "There really is something there. Not sure what to think of it, but I've experienced it myself," and "You need to keep your mind open to the possibilities. Once you do, you'll see that there's something to it." These expressions sound strikingly similar to the words of a victim of a mentalist scam artist, akin to a psychic.
Psychics have employed this con for centuries, a technique that has been refined over time. The core mechanism remains consistent across variations. First, the audience selects itself, as most individuals are not interested in psychics or similar phenomena. The initial audience is composed of those who are more open-minded and less critical than the general population.
Next, the scene is set, with dimmed lights, a hype around the psychic, and staff conducting background research on the audience through social media or conversation. The audience's demographics are noted. The psychic then narrows down the demographic information, making a statement that appears specific yet is statistically probable for the demographic group.
If no one reacts, the psychic implies that the information is too embarrassing for the person to disclose, encourages private readings, and attempts again.
The mark, believing they have been "read," proceeds to ask a series of questions that, while sounding specific, are in reality just statistically probable guesses based on demographics and prior answers, phrased with high confidence. This marks the beginning of the subjective validation loop, where the psychic asks a series of questions that seem incredibly specific to the mark but are merely statistically probable guesses, framed in a highly confident manner.
Ultimately, the psychic ends the conversation, leaving the mark with the impression that they possess uncanny powers. However, this is merely a con. Some psychics are upfront about it being entertainment, while others deliberately deceive people, all employing the same basic mechanisms that set the stage for the reading. The audience is typically composed of those already predisposed to believe in psychic phenomena and those they have managed to entice.
Hardcore skeptics are usually a small minority, making them easy to manage and adding social pressure to curb skepticism. The audience is primed to believe and is familiar with the mythology surrounding psychics, aiding in managing expectations and shaping their performance.
Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
Also reported by 1 other outlet
- Chat-based Large Language Models replicate the mechanisms of a psychic's con softwarecrisis.dev