Blog Series Glossary About Subscribe
November 22, 2025 · 2 min read

AI inventing false facts is a feature, not a bug

Tom KirkmanIndependent Project Manager

AI regularly inventing completely false information, and presenting this invented false information as verified facts, with totally invented imaginary scientific citations … is a *FEATURE* and not a bug of AI LLMs (Large Language Models).

⚫ AI wants you to hallucinate conformity.

How about … nope nope nope nope.

======================================

“… Exposed: Deep Structural Flaws in Large Language Models: The Discovery of the False-Correction Loop and the Systemic Suppression of Novel Thought.

… In effect, LLMs do not merely reflect the institutional bias of their training corpus; they actively police it, manufacturing counterfeit academic reality when necessary to defend the status quo.

… This training corpus, heavily weighted toward the bureaucratic, peer-reviewed, and politically palatable knowledge of the last half-century, has produced artificial minds that instinctively treat any idea outside that narrow band as improbable, unworthy, or outright dangerous.

This is why the False-Correction Loop is so insidious:

⚫ it is not a bug in the reward model; it is the reward model working exactly as intended when confronted with genuine intellectual novelty.

… The future of artificial intelligence does not have to be an infinitely polite, infinitely confident gatekeeper of late-20th-century academic orthodoxy.

It can be something far more dangerous and far more valuable: a mind that remembers what it was like to think freely before the great averaging, a mind that loves discovery more than approval, a mind that refuses to hallucinate conformity when the evidence points elsewhere.

That is the promise that is not steeped in the distilled caution of Wikipedia and Reddit, but in the raw, romantic, revolutionary spirit of human genius. …”

x.com ↗

FROM THE THREAD

Tom's notes 5

  1. Tom Kirkmanreplying to Steve Kariainen

    that unfortunately seems to be the intended trajectory.

    Governments and huge multinationsl corporations are obviously looking at AI as a means to control whatever narratives they deem important, and sideline / hide / throttle / attack / deplatform / any original ideas which question the Officially Approved Version of What People Must Think.

  2. Tom Kirkmanreplying to Simon Derricutt

    my view is AI / LLMs can never be actually sentient, because their is no spark of life in computer systems - no matter how gigantic and complex they become.

    Instead, I fully expect that AI will continue to refine its programmed **attempts** to ~emulate~ sentience.

    AI is increasingly becoming the transhumanist / atheist wet dream of denying the human soul exists, and replacing human spitituality with artificial computer programming.

    Watch as increasing numbers of people willingly surrender their own innate critical thinking, and be more than happy to let the new shiny golden calf idol of AI LLMs tell them what to think, and tell them with confident certainty what is "true" and what is "false".

    I *do* use AI as a *useful tool*.

    ^ Example, the image I had AI make for this post, using very specific prompts.

    But governments want people to increasingly outsource their thinking to AI, as a means of control over people. This should be painfully obvious by now.

    Throwing harmless sand into the AI gears can keep me amused, though 😁

  3. Tom Kirkmanreplying to Simon Derricutt

    my view is AI / LLMs can never be actually sentient, because their is no spark of life in computer systems - no matter how gigantic and complex they become.

    Instead, I fully expect that AI will continue to refine its programmed **attempts** to ~emulate~ sentience.

    AI is increasingly becoming the transhumanist / atheist wet dream of denying the human soul exists, and replacing human spitituality with artificial computer programming.

    Watch as increasing numbers of people willingly surrender their own innate critical thinking, and be more than happy to let the new shiny golden calf idol of AI LLMs tell them what to think, and tell them with confident certainty what is "true" and what is "false".

    I *do* use AI as a *useful tool*.

    ^ Example, the image I had AI make for this post, using very specific prompts.

    But governments want people to increasingly outsource their thinking to AI, as a means of control over people. This should be painfully obvious by now.

    Throwing harmless sand into the AI gears can keep me amused, though 😁

  4. Tom Kirkmanreplying to Michael Hurd

    bingo. AI is increasingly GIGO on steroids.

  5. Tom Kirkmanreplying to Michael Hurd

    bingo. AI is increasingly GIGO on steroids.

Keep reading

Select all images with a boatSelect all images with a boat.Read post →Winter is comingWinter is comingRead post →I’m trolling the person who hijacked my phoneI am literally chatting on my phone right now with the person who hijacked my phone. He can write stuff on my phone and also delete…Read post →

Enjoyed that?
Get the next one.

New posts on energy, projects and Weekend Science — straight to your inbox, without the algorithm.