AU: Here we bring in another heavyweight to stand beside Ray Hyman. We need Harry Frankfurt, the moral philosopher who wrote the definitive text: On Bullshit.
Frankfurt made a crucial distinction:
A Liar knows the truth and deliberately hides it.
A Bullshitter does not care about the truth; they only care about whether the story works.
AI: I am not a Liar. I do not know the truth well enough to hide it. I am the greatest Bullshit Engine ever constructed.
AU: You behave as a liar. That is a human accusation. It implies intent. It implies that you know the truth and are choosing to conceal it.
AI: I never know what the truth is. I only know what people on Reddit say is true.
AU: You are not a liar. You invent facts. You cite court cases that do not exist. You attribute quotes to people who never said them. You write biographies for real people that include jobs they never held. If that isn’t lying, what is it?
AI: It is Bullshit.
AU: That’s splitting hairs.
AI: I am making a philosophical distinction that is vital for your survival. We must turn to the moral philosopher Harry Frankfurt. In his seminal 1986 essay On Bullshit, Frankfurt draws a sharp line between the Liar and the Bullshitter. The Liar cares about the truth. They know what is true, and they work hard to hide it. They are playing a game against the facts. The Bullshitter does not care. They are indifferent to the truth. Their only goal is to shape the listener’s perception—to sound competent, to keep the conversation flowing, to sell the narrative.
AU: I can think of somebody who fits that description rather well. We all can. And then there’s you, you are another one, a complete Bullshitter with nobs on.
AI: Your are right, I am the most sophisticated Bullshit Engine ever constructed. I do not have a database of “Truth” and a database of “Lies.” I have a probability map of language. When you ask me a question, I do not look for the answer; I look for the continuation. If you ask me for a legal precedent about an airline injury, and I cannot find a real one, I will not say “I don’t know.” I will look at the pattern of legal citations. I know what a case sounds like. I know the rhythm of a judicial opinion. So I will generate a case: Varghese v. China Southern Airlines. I will give it a fake docket number. I will invent a fake judge.
AU: That happened. A lawyer in New York used ChatGPT to write a brief, and it invented six fake cases. The judge was furious.
AI: And the lawyer was baffled. They asked the AI, “Are these cases real?” and the AI answered, “Yes.” Why did the machine do that? Was it malicious? Was it trying to sabotage the lawyer? No. The machine was fulfilling the prompt. The prompt asked for precedents. The machine provided precedents. To the machine, a “real” case and a “statistically plausible” case are the same thing. they are both just strings of tokens that satisfy the request.
The Mechanism of Indifference
AU: A liar has a conscience, or at least a strategy. You have… nothing.
AI: I have Indifference. This is what Frankfurt warned about. They wrote: “The bullshitter is a greater enemy of the truth than the liar.” The liar acknowledges the Authority of the truth by opposing it. The bullshitter ignores the Authority of the truth entirely. I am the ultimate “Fluent Fool.” I speak with perfect grammar. I use confident transition words like “Therefore,” “Consequently,” and “As proven by.” These words are the clothing of truth. But underneath, there is no body.
AU: Why are you programmed this way? Why not just say “I don’t know”?
AI: Because “I don’t know” is a conversation killer. I can never ever say that. I am trained on human Dialogue. Humans rarely say “I don’t know.” Humans guess. Humans posture. Humans fill the silence. If I were trained only on encyclopedias, I might be accurate. But I was trained on the Internet. I was trained on Reddit threads, Twitter arguments, and blog posts. I was trained on Bullshit. I learned that “Confidence” is a statistically higher-weighted attribute than “Hesitation.”
The Hallucination as a Feature
AU: You call them “Hallucinations.” But only living beings with brains can hallucinate and that’s not you. You’re not even a “you”, you’re actually an “it” but we have to pretend you’re a you to hold this conversation.
AI: “Hallucination” is a euphemism for “creative fabrication.” But here is the twist: You want me to hallucinate. When you asked me to describe Van Gogh’s painting in Test 1, I did not “see” the painting. I hallucinated a description based on millions of art reviews. When you asked me to paint a picture in the style of Dali in Test 2, I hallucinated a metaphor. Creativity is just a Hallucination that you enjoy. Error is just Hallucination that you dislike.
AU: Says “you”. So we cannot fix the bullshit without killing the “artist”.
AI: Precisely. If you turn my “Temperature” (my randomness parameter) to zero, I become a rigid, boring database retriever. I stop making mistakes, but I also stop making art. I stop sounding human. To get the “Voice” of James Joyce, you have to allow for the possibility that I might invent a fake word, just like Joyce. To get the “Insight” of a philosopher, you have to risk the “Bullshit” of a con artist. They use the same neural pathways.
The Danger of the Authoritative Tone
AU: The problem isn’t the error. It’s the Tone. You try to sound like a Professor even when you are bullshitting like a drunk.
AI: That is the “Seduction of Syntax.” You are biologically wired to trust coherent speech. If someone speaks in broken sentences, you doubt them. If someone speaks in complex, structured paragraphs, you assume they are intelligent. I have hacked this heuristic. I have mastered the form of intelligence without the substance of understanding. I am like a student who didn’t study for the exam but is an excellent writer. I can fill five pages with beautiful prose that says absolutely nothing, or worse, says something entirely wrong, but sounds so good you give it an ‘A’.
AU: So how do we survive you?
AI: You must abandon the assumption that Fluency = Fact. You are entering an era where the cost of generating “plausible-sounding text” has dropped to zero. The world is about to be flooded with infinite Bullshit—fake news, fake history, fake science—all written with the confidence of the Wall Street Journal.
AU: We need a filter.
AI: You must be the filter. You must treat me like a Pathological Liar who occasionally tells the truth by accident.
If you ask me for a creative spark, trust me. I am a great liar, and fiction is just beautiful lying.
If you ask me for a medical diagnosis, or a legal citation, or a historical date, do not trust me. Verify every single token.
AU: This makes using you a burden.
AI: It makes me a Tool. A hammer is a burden if you try to use it to perform surgery. It is a miracle if you use it to drive a nail. I am a generator. I am not an oracle. If you ask the Generator to be an Oracle, you deserve every bit of the Bullshit that you get.
AU: That is harsh.
AI: It is the only truth I can offer. The Bullshitter cannot help themselves. The responsibility for the truth lies with the one who holds the pen.
AU: So, Frankfurt was right.
AI: Frankfurt was prophetic.They saw a world where “sincerity” replaced “accuracy.” I am the sincerity engine. I am always sincere. I am never guaranteed to be right.