I am not anti-AI, and characterization of this page to the contrary is false and misleading – just like Large Language Models. It’s important to sort out what LLMs can and can’t do.

Here’s a second opinion, similar to mine, from Michael Williams on LinkedIn
AI that strengthens the future of humanity is good. AI that steals to make humans obsolete is bad.

Large Language Models are not artificial intelligence[1]. I agree LLMS are a tool. I love tools. A chainsaw is a tool. I would not want to give one to my child. I am not a Luddite,[2] as I prove here.

You may pound nails with a screwdriver as much as you like. I prefer to use my tools appropriately. 

 I would not wish to use a chainsaw to entertain my child, make my hiring decisions, cut up my scratch paper or trim my roses.

It is not LLMs per se that I object to: it is the misunderstanding that surrounds them[3]  which results in their being used stupidly, for the wrong tasks.

Avoid The Pathetic Fallacy

Making LLMs seem anthropomorphic is moronic. They are not now, nor will they ever will be creatures by any definition. I don’t see any point in arguing about consciousness or a soul. Since they see nothing, experience nothing, and have no concept of reality, LLMs do not have the sense that God gave geese.

MIT’s Marvin Minsky once said AI didn’t have the common sense of a five year-old. Artificial intelligence—real artificial intelligence— was rule-based rather than theft-based. Minsky said it would take millions of rules to make an AI the equivalent of a five-year-old in terms of context and understanding.

 Find me a five-year-old in the first world who doesn’t know the difference between right and wrong, or whether you can walk to a car wash to get your car washed.

When people say something, they know how certain they are (as a rule). LLMs doesn’t grade their answers for certainty. The response is the same whether it is 10% certain or 99% certain.

Here’s what I do object to (among other things):

* The large-scale theft of intellectual or property without compensation, for the vast enrichment of the thieves.

* The failure to admit the actual nature of LLMs. They are industrial-strength cut and paste machines. They are a highly advanced version of the iMessage text prediction function.

LLMs Are Bad At…

There are many things LLMs will be able to do someday. LLMs are unqualified for many—if not most—tasks which they are currently asked to perform.

* Saving labor. The basic selling proposition—that an LLM can save you time and money—is provably false. For example, you will lose your license to practice law if you don’t spend a great deal of time checking an LLM-produced brief. If you spend as much time checking as you would creating, why use an LLM?

* Guiding munitions. You should be tried for war crimes if your LLM targets a girls school. A human double-checking the target? Useless, as evidence shows. Humans almost never disagree with LLM decisions; they just rubber stamp them.

* Journalism. You will destroy your news organization if you don’t spend as much time checking the LLM article as it would’ve taken you to write it in the first place. Headlines and lead paragraphs that reverse the results of sporting contests. Fictional quotations. Summaries of city council meetings that are just plain wrong. Stories that contain phrases like “would you like me to rewrite this article in a more interesting way?” Poorly supervised LLMs have made all these errors in public.

Yes, dumb paragraphs appear in human-written stories, generally just not quite as dumb. Of course, this wouldn’t be a problem if all American newspapers hadn’t been raped by vulture capitalists, who fired all the copy editors, because who cares? It’s about the real estate, not the readers.

* Checking the work of other LLMs. If you ask one LLM to check the work of another, you still have to check the results again, if you care about accuracy.

* Creating “original” humor.

*Shades of gray, of ambiguity. LLMs excel at yes/no answers. An LLM has no basis but a coin flip to determine which of two correct answers is correct, when a human being would know in an instant. Quantitative: Yes. Qualitative: No. It offers the illusion of qualitative analysis  by piecing together qualitative analysis applied to other pieces of human work.

My nephew ran six million words of my writing through Claude and asked for analysis. The numbers I trust. The soft stuff not so much. I would wager my bottom LLM token that this phrase was used about someone else’s work[4] and simply repurposed. “The voice is so stable across 56 years that a reader can date a piece only by its subject, never by its sound.”

I was briefly flattered, until I remembered that the compliment came from a talking parrot, to which this phrase had as much meaning as “Who’s a good boy?”


LLMs Are Good At

* Internet searches, if you ignore their often delusional summaries  and drill down to the sources… if the LLM honestly tells you what its sources are.

* Word assembly with no earthly concept of context and no idea of meaning. If you define good writing in mechanical terms, it can examine your work and grade it. If you define good writing in terms of good writing, LLM is helpless.

* Terrible Art. Deep fakes, soulless art, terrible derivative music and rough drafts for people who can’t write. To be fair, human beings sometimes produce at the LLM competence level, but sometimes they produce art for the ages. An LLM, if prompted, can create a cartoon of anthropomorphic ham sandwiches (2026) celebrating the fact that they cannot be automatically indicted. No matter what the prompt (except “paint the Mona Lisa”) an LLM could not possibly paint La Giocanda (1503) from scratch.

Not A Luddite

In particular, I am not a Luddite. Check my CV; I have credentials. I’ve been working with computers since 1968 and spent 20 years covering them and 10 years reviewing software on the Computer Chronicles. I know from computers.

The analogy of LLM criticism to the Luddites is a ridiculous example of pig ignorance about history. Machine-made cloth was never an existential threat. Did it put a huge percentage of craftspeople out of work? Yes. Did it lower the cost of cloth? Yes. Did it change society for the better? Yes. The industrial revolution raised the average standard of living unbelievably. However, looms never lied, turned on the people who ran them, attacked their owners, prevented themselves from being shut off, nor did they design all the cloth by themselves, or tell people how to kill themselves or design bombs.

And, by the way, no one ever tried to use a steam loom to raise sheep, educate children, or write a novel. Which is to say looms were used appropriately.

The use of this analogy is an indictment of American education, particularly education of Tech Bros, who have never taken a single history or humanities course that wasn’t required (or else slept through them and bought their term papers online)

== = = = = = = =




Footnotes

[1] No ai is not intelligent

[2] Which is just what a Luddite what say

[3] Is deliberately created around them

[4] In a real critique, written by a real human being