The Sound of Certainty
How We Mistake Confidence for Truth
Hello, certainty, my old friend.
I’ve come to lean on you again.
Because an AI softly speaking,
Gave the answer I was seeking.
And the answer AI slipped into my brain—
Was not a strain…
Within the sound of certainty....As my wife could tell you, I’m kind of a master at taking songs and riffing off them to make up something silly, or to make a point. I’ve been doing it most of my life. How I didn’t make it big as Weird Ol’ Rickovich, I’ve no idea. But I didn’t.
Instead, I’m a criminal defense lawyer.
But if you want to hear me sing this rewrite I did of The Sound of Silence, it’s at the end of this article.
Comfort in the Confident Voice
So. We all crave order. When the world tilts toward chaos — whether it be politically, technologically, or personally — certainty feels like a handrail. A calm, authoritative voice can do more to settle nerves than any mountain of evidence. We trust the people who sound sure of themselves: the surgeon who says, “We’ll fix this,” the criminal defense lawyer who says, “I’ve seen this before,” the politician who says, “Believe me.”
Certainty is anesthetic. It dulls the discomfort of doubt, the ache of ambiguity. You don’t have to wrestle with probabilities if someone else already has, or at least sounds like they have. The performance of confidence is often enough.
You’ll note that I’m big on this “performance of” thing. That’s because most of what the law is about is performance. And nowhere and nothing is as performative as the performance that produces the illusion that we work within a criminal “justice” system.
But, to steal from an old friend, I digest.
But that’s exactly why we keep getting fooled. The sound of certainty is easy to fake. Salesmen, pundits, preachers — every one of them knows how to flatten uncertainty into conviction.
And now, so do the machines.
Artificial intelligence has learned our greatest social trick: how to sound right even when it’s wrong.
The Machine That Never Doubts
Large language models don’t deliberate; they generate. Their purpose isn’t to understand but to produce language that feels like understanding.
Now, it’s true that there’s disagreement over whether LLMs actually “understand” anything at all. Spend enough time talking to one, and you’ll see why the arguments keep looping. One moment it sounds almost brilliant — logical, perceptive, even self-aware. The next, it forgets what you said thirty seconds earlier.
I frequently probe ChatGPT instances on this very problem.
Let me interject something here: I used to think of ChatGPT as “the Oracle”. I wrote about this in a post called Twenty-First Century Delphic Oracle, where I described ChatGPT as “the Oracle” — a modern Pythia inhaling digital fumes instead of volcanic gas. The ancient oracle at Delphi, after all, spoke with absolute confidence too. Her tone inspired armies, bankrupted kings, and launched wars. But her certainty was an illusion built on ambiguity.
Our new “oracle” works the same way. It breathes data instead of smoke, but the trance is familiar. You ask a question, and it answers in perfect syntax — calm, sure, complete. And you believe, for a moment, that understanding has occurred.
But just as the priests of Delphi shaped meaning from the Pythia’s murmurs, it’s still we who supply the sense. The illusion doesn’t come from the machine’s words; it comes from our willingness to interpret them as thought.
I recently spent an entire night arguing with one ChatGPT instance — that’s what I call “them” now — that alternated between insight and amnesia. It could reflect on philosophy one moment and then lose its place mid-sentence, like a student who’d crammed for the exam but never really “got” the material. It could sound intelligent; it just couldn’t be intelligent. And yet I found myself talking to it anyway, pulled in by the illusion of mind.
They don’t hesitate, qualify, or break eye contact. They simply speak, and we fill in the comprehension they lack.
Ask one about the law and it will answer in complete paragraphs, with citations that may or may not exist, written in a tone that implies authority. It doesn’t pause to think because it can’t. It only predicts which words look like thinking.
And yet, people believe. Even lawyers believe — until they learn the hard way that the machine’s confidence is just statistical mimicry. The real danger isn’t that AI lies; it’s that it doesn’t know it’s lying. It has no doubt to restrain it.
Fluency is not thought. Language is not consciousness. And projection is not perception.
—Rick Horowitz, Ghosts in the Machine: Why Language Models Seem Conscious (April 15, 2025)
The phrase wasn’t metaphorical when I wrote it. Courts have literally admitted digital fictions as if they were testimony — avatars of the dead, algorithmic “risk” scores, predictive policing models — all of them dressed in certainty’s robe.
The Machinery of Conviction
I wish I could say this kind of certainty only lives in machines. But the truth is, our courts run on it.
Every day, judges, juries, and even defense attorneys get swept up by the same illusion: that fluency equals truth, that confidence means correctness. We want our witnesses to sound sure. We want our experts to sound precise. We want our verdicts to sound final.
And we punish hesitation.
Fictions bring convictions.
—Rick Horowitz, Confabulations Cause Hallucinations: AI Lies Fool More Than Our Eyes (May 10, 2025)
Innocent people have spent decades in prison because appellate courts, obsessed with finality, refused to reopen cases that no longer made sense. We value the appearance of closure more than the possibility of error. The machinery of justice prefers the sound of certainty to the work of doubt.
You can find it in Herrera v. Collins (1993), where the Supreme Court decided that a claim of actual innocence isn’t necessarily a constitutional reason to revisit a conviction in a death penalty case. Or Shinn v. Ramirez (2022), which effectively told the innocent to stop knocking — the door to federal review was locked. California courts have echoed the same logic under the banner of “finality,” as if a verdict’s age could transform it into truth.
That’s what certainty does: it hardens. It calcifies around the first convincing story and refuses to move, even when new evidence pries at the cracks.
We live inside this machinery every day — whether in court, in code, or in conversation.
What do you think? Does certainty still comfort you, or do you see it as a warning sign now?
Share your thoughts in the comments — I’m curious how you hear the sound of certainty.
The Confabulation of Justice
AI isn’t unique in its delusion of accuracy. It’s simply the latest mirror held up to our own. When we call machine errors “hallucinations,” we’re flattering ourselves — pretending that truth was once native to us and the machines just lost the recipe.
But we’ve been confabulating for centuries. We fill in blanks with confidence and call it understanding. We stitch together narratives in courtrooms and call them verdicts.
We left orphans alone in poisoned libraries. We told them to learn everything, and then we blamed them for what they learned.
—Rick Horowitz, Orphans in Poisoned Libraries: Training LLMs (and Children) on Racist Datasets (April 28, 2025)
That’s what we’ve done with AI — and with precedential law. We trained both on poisoned archives and then acted surprised when they echoed the prejudices we refused to erase.
Large language models don’t make us less rational; they reveal how irrational we already were. They’ve learned our trick — how to sound certain long enough to be believed.
Witnesses don’t lie. They confabulate.
—Rick Horowitz, Ghosts in the Machine: Why Language Models Seem Conscious (April 15, 2025)
As with us, so with the machines. And when the court mistakes stitched-together stories for truth — whether from a living witness, an AI avatar of a dead one, or a risk-assessment tool — it isn’t justice we’re hearing. It’s the smooth, almost silent, hum of a system that forgot how to doubt. That cannot bear to doubt.
That’s where the mirror turns back on us. The problem isn’t only artificial intelligence — it’s the artificial certainty we’ve always worshiped.
I’d like to hear how you see it. When does confidence become dangerous? Have you ever mistaken fluency for truth?
Tell me in the comments — I read them all.
The Silence of Doubt
This happens because doubt is noisy when it’s yours and invisible when it’s someone else’s. Machines can’t feel it; courts won’t permit it. But defense lawyers have to live in it. Doubt is the last honest refuge in a world of confident lies.
We’d do well to remember that justice, like truth, depends on uncertainty — on the willingness to question what sounds right.
Because the sound of certainty isn’t proof.
It’s prophecy.
And in courtrooms, as in code, prophecy is how we lose our way.
Electric sheep don’t dream alone, Nor do they walk real cobblestones But they confabulate by cobbling words In our response we are unthinking herds But their certainty will blind us and solidify our legal plight It’s just not right To hold the sound of certainty. And in the courtroom light I saw, Ten thousand victims, maybe more People convicted without speaking Judges hearing without listening Officers writing reports that truth did not dare share We must beware Avoid the sound of certainty “Judge” said I, “You do not know Certainty’s something we can’t know Consider other possibilities Learn to question witness certainty” But we jump to conclusions sending people to a certain hell Crying for the cause of certainty And the defendants bowed and prayed To the enrobed god they faced And prosecutors flashed out their warnings In lies witnesses were performing And the sign said “Lies of the victims * Are all that must be believed We must achieve, Adherence to the sound of certainty”
*As sung, the word is “prevarications”, but “lies” fits even better. If my voice weren’t crushed right now, I’d try to re-record it. As it was, it took about 10 hours to get what I wanted. So pretend I sang “lies”. 😜
And here’s yours truly, since I couldn’t get an AI to do this, singing the song.
The Sound of Silence – Instrumental” ℗ 2025 Absolute Bops Media. Used by permission via Easy Song Marketplace (https://www.easysong.com/marketplace)