The words on this blog are AI-free
2026-05-27 12:15 by Ian
TL;DR
- If it isn't worth my time to write, it isn't worth anyone's time to read. And I act reciprocally. Don't send me AI slop to read. I won't.
- LLMs seek and generate patterns. Never truth.
Seriously... LLMs are destroying the knowledge-value of the internet, even faster than advertisement. At least advertisement takes a few decades to irrecoverably corrupt the media it touches. But LLMs did it within a few years by being cost-free generators of inanity, rather than intelligence.
There is nothing intrinsically intelligent about an LLM. Nothing about them is truth-seeking (grammatically or otherwise). They are sophisticated pattern seeking systems that can be used to generate new patterns that are flavored similarly to those forming the statistical baselines of their training data.
Classic GIGO, but made even worse by the selective biases of the politically-motivated censors of both the garbage used as input, and the garbage given as output. See also: Mean-time-to-Hitler.
AI isn't making anyone a better writer. But it is doing these things...
- Allowing poor writers (language translations exempted) to imitate the sharp thought required to write incisively, thus polluting the internet with half-baked thoughts which no longer bear the grammatical flags of being such.
- Turning previously good writers into lazy bastards.
- Generating plausible language that is devoid of any thinking (thus, also devoid of intelligence).
All of this culminates in wasting the time of readers by way of raising the noise floor on all written material on the internet. Just as advertisement does, but better disguised.
Little of what I find in search results is worth reading in 2026. It is mostly AI slop that are the same trite half-answers emitted a thousand different ways; like a giant mechanical cow chewing its cud. I think we've reached (or are near) the entropic tipping-point where the last whisps of original human knowledge are being overwhelmed by the reshuffled and amplified tripe that costs practically nothing to write, and is just as effective at pulling ad revenue.
We need a humans-only internet if we want to continue to use the internet productively.
"The question of whether a computer can think is no more interesting than the question of
whether a submarine can swim." ---Edsger Dijkstra
I feel the need to reiterate...
I have been using LLMs for (limited) coding work for the past year or so, but the values in that case are somewhat different than communication with other electric meatballs.
Except where LLM output is the thing being discussed:
AI is never used to write any post on this domain, nor any of my other written communication.
It isn't even fed to a robot for proofreading, and I often choose to ignore my spellchecker.
I always disable autocorrect and autocomplete features in my text editors. Without exceptions.
I most especially do not ask robots if ideas make sense, or if statements are true (except possibly in the algebraic sense).
Were they shown such perversions of epistemology, I think 2026 would see Leibniz driven to homicidal rage, John Locke offering to supply him with ammunition, and St. Thomas Aquinas absolving them both of all sin.
Scattered on-point memes...
Previous: How to verify that Earth is not flat
Next: