notAI/LLM: A Power So Great It Can Only Be Used for Good or Evil
When In Doubt, Sing About It
Paul Schindler AI Song Playlist
Or Just Read About It
I have lots to say about LLMs in my weekly column. AI for short, not because it is actually AI, which it isn’t.
Bill Gates Gets With The Program
On His Own Website:
An epochal shift. The turbulent AI era is here. The choices we make now are critical. We need a plan to ensure that the good outweighs the bad.
In the New York Times:
Bill Gates Warns A.I. Is More Dangerous Than Big Tech Will Admit
= = = = = = = = = = = = = = = = = = = = = = = = = = =
Mt. Shasta climbers misled by AI rescued
Robert Roy Britt, writing coach, author and journalist, on LinkedIn:
I’m presently having a huge argument with myself over how vigorously to wade into the AI vs. no-AI debate with fellow journalists, other writers and editors, and media execs.
My take: LLMs Earn Praise When Not Used Stupidly
In his Substack Media & the Machine, Rob Kelly wrote, AI Isn’t Running Out of Data. It’s Running Out of Internet. His take: “Elon, Ilya, and Sam are confusing the web with the world. Here are 7 kinds of data we’ve barely tapped.” I don’t always agree with him, but this is spot on.
Plaintiff busted trying to use AI prompt injection to win court case, hides text instruction in filing — demands AI model reviewing the text should side with him, rumbled because of strange white spaces in text
–Thank you Tom’s Hardware
Don’t Call It Hallucination, Call It Bullshit
World’s Most Talented/Expensive Talking Bird.
LETTER-PERFECT DEPICTION Of AN LLM
I thought the Blondie comic strip was a dinosaur. Wow! A perfect depiction of LLM sychophancy
THIS WILL KEEP YOU AWAKE TONIGHT
The World’s Leading Deepfake Expert No Longer Trusts His Own Eyes
In the age of A.I., Hany Farid is struggling to prove what’s real before the internet decides for itself. A tip of the PSACOT hat to Joe Kashi, who unlocked it for us.
But First: Not AI/LLM Proven Not Ready for Prime Time (not even ready for 3 a.m. on Sunday) Let’s agree not to let it set bombing targets, OK?
Douglas Adams Predicted AI Slop
Hallucinations Cost $300,000 each.
A Munich court ruled that when AI speaks on behalf of a company, it represents the company. You can’t say “the AI made it up.” It’s the company’s responsibility. Google’s AI Overviews accused two German publishers of running scams. No source said it. The AI invented it. “The algorithm did it” is officially dead as a legal defense. Details (and references) here
Why AI Hallucinates
From Mohammad Huda on LinkedIn. This is the simplest explanation of hallucinations I have ever seen:
Here’s the plain-language version of what’s happening. When you ask a chatbot for case law — or a statistic, or a quote — it doesn’t look anything up. It predicts text that sounds right. A citation like “Henderson v. State, 412 F.3d 998” has the exact shape of a real one, so the model produces it, confidently, whether or not it exists.
The Future Is Laughing At Us
Sandra Loughlin, PhD wrote on LinkedIn about the stupid stuff being done today in AI which will be laughed at someday. Of course she means LLMs. Her introduction:
The number of unforced errors we’ve made with employees so far on this journey is kind of stunning. And the worst part is that it has all been painfully predictable, rational, and fiduciarily responsible.
This fits well with my speculation on ways in which the future will laigh at us. What Will They Think Of Us in 2125? 2525? Part 1 and Part 2
More NotAI/LLM briefs
including:
• LLMs Are Like Parrots
• You Have To Check The Other Lawyers’ Brief For AI
• The Most Mysterious Ai Response Ever
• A counterfeit world is coming online.
• Humans Create, LLMs Don’t
• “AI: I Know It When I See It”
• What LLM/Not AI Is Not
• Magistrate Finds Fake Cites, Levies Fine
• So Lemme Get This Straight
• Starbuck AI Lasts 9 Months, Fails
Meanwhile, across town
Not AI/LLM Bot Iterative Hell
Hot Fresh Links
Chose Your Poison
What was it like last time(s)?
AI: Timely Topic
This Isn’t New
Almost Fooled
Your One-Stop Used Link Shop
==
LLM/Not AI: The Part They Accidentally Said Out Loud. Facebook accidentally published a thoughtful analysis of the goals of the AI industry.
==
Pope Leo Warns of Risks From A.I. in 42,300-Word Encyclical
The Times said: “The document marks a powerful foray by the leader of the Roman Catholic Church into the debate about the misuse or overuse of artificial intelligence.” I think he was too gently by half. Basically, dangerous if not controlled, and be careful not to trash humanity. While he was at it, he apologized for the centuries during which The Church justified and supported slavery.
==
Magistrate Finds Fake Cites, Levies Fine
[The lawyer] filed arguments explaining her absence and why it shouldn’t be used against her client — but three of the four cases she cited as precedent, a federal magistrate found, were nonexistent. [She was fined $1,000].
===
======
This is a brief quotation used in a critical article (well not really critical)
You can see the whole cartoon here.
==========
Thank you Steven Vaughan-Nichols
Who Wrote The Declaration Of Independence
From LinkedIn, by Mark Henderson
Pop quiz: Who wrote the Declaration of Independence? Thomas Jefferson in 1776 [Ed. Note: With help from Ben “LLM” Franklin and John “Better Prompts” Adams].
An AI “detector” flagged it as 97.75% AI-generated. AI detection tools are not the answer. They never were.
If your entire academic integrity strategy is “run it through a detector,” you’re building on sand. The real question isn’t did AI write this? It’s does this assignment even require original thinking? Redesign the work. Make it personal. Make it specific. Make it theirs.
No detector needed.
===
WordPress Hates on Benj. Franklin
In a WordPress post, I quoted Benjamin Franklin: “We must all hang together, or most assuredly we shall all hang separately.” I tried to publicize the post containing that quotation. There followed a tale of AI incompetence, aided and abetted by human incompetence, which forced me to bowdlerize Franklin’s words.
All I had to do was paraphrase the quotation: Franklin also suggested that if we all agree to stand together and speak truth to power, there aren’t enough masked men in black to arrest all of us. (quotation bowdlerized to allow this post to be publicized). If you have a high school education and ever once paid attention in US History, the actual quotation should come to mind. In order to allow this post to be publicized I cannot offer a link to the quotation. Search for “Benjamin Franklin” “separately” and you will find it.
Briefs
How Accurate Are Google’s A.I. Overviews?
According to the New York Times, Google’s A.I.-generated answers look authoritative, but they draw on an array of sources, from trustworthy sites to Facebook posts.
The State of AI: Perfect Examples
Should you walk to the carwash or drive there?
Details on the Cron/Leeds book
I Asked ChatGPT and This Is What It Said
AI would hate humor if it had any idea what humor is. Or if it had any “idea” period.
Bot Banned by Wikipedia
Wikipedia just had to ban an AI that was making edits on its own.
===
Puttting LLM/notAI In It’s Place: H. Claire Taylor
I have been reading about AI since 1973. The single most clear, concise and cogent essay I have read on the subject in this century was written by H. Claire Taylor, an under-rated author of comic fiction who turns out to be an exceptional essayist. She puts notAI/LLM in its place, and describes that place clearly, while only using the word slop five times. Because she accurately points out that AI is a market term for LLM, I will refer to it in the future (as often as I remember) as notAI/LLM.
Puttting LLM/notAI In It’s Place: Helen McCarthy
Do you have to be a woman to see the truth? Helen McCarthy summed up my feelings perfectly on LinkedIn. And of course I credit her, because I am not an LLM scraping (stealing) her content, but rather a human agreeing that “I still think GenAI is destructive and pernicious. But that doesn’t make all AI useless. It’s up to us whether we use it to help people manage huge and complex tasks, or to steal the work of others and wipe out their jobs.” Of course she confuses LLMs with AI, but I am a voice screaming in the wilderness on that subject.
====
notAI/LLM vs. Live Music
I had a recent exchange with a musician on the subject of whether notAI/LLM will kill live music.
My comment: Records did not kill live music. Radio did not kill live music. The Internet did not kill live music. All these advances simply raised the bar. And, of course, put hundreds of thousands of musicians out of work.
I have read that prior to the invention of the phonograph, a lot of musicians were not all that talented, but survived because people had nothing to compare them to.
I don’t think AI is going to raise the bar: pasting together other people’s ideas and performances is not likely to create anything original, or end live music.
The final part of the exchange: “At least you have one protection: I don’t think notAI/LLM is going on tour anytime soon.”
His response: “I’m genuinely not so sure about that! Haha but live music will always have a place, I feel. Can’t beat genuine human connection” I hope he wasn’t whistling past the graveyard.
====
Law of Unintended Consequences
notAI/LLM Causes Parts Shortages, Raises Prices… Including for Jet Plane engines
First, a 20,000-foot analysis by a friend who reached this conclusion simply by reading the trade press. Links to actual stories amount to footnotes to his analysis.
AI infrastructure build-outs have become so all-encompassing that many component prices have tripled and critical computer industry parts, like DRAM memory and graphics cards, are now sold out through 2026 as the AI big guys option and buy up everything in sight. That will have a rippling negative impact upon many economic sectors.
The chip manufacturers, though, see the crash-project AI build-outs as an anomaly and are not spending tens of billions to build more chip manufacturing facilities that may have no demand by the time that they’re completed in a few years. In just a few months, the amount of DDR memory to build a standard office computer has increased in price from $200 retail to $600, as an example.
Similarly, large jet engines normally used by the aviation industry are apparently sold out through 2030 to AI companies that are using them instead for on-site power generation. Apparently, the AI companies are so desperate for more and more electricity that used jet engines are being stripped off older aircraft.
= = = = = = =
Yahoo Gaming: Ai demand doubles dram memory prices
Yahoo Gaming: DDR5 memory prices have literally doubled since September, with premium G.Skill kits skyrocketing 138% in just two months. That 32GB DDR5-6000 kit you bookmarked for $140? It’s now pushing $265, and the pain is only beginning.
= = = = = = = = = = = =
Gizmodo: We’re So Screwed: AI Is Making Every Gadget Cost More
The demand for AI is making every gadget cost more, whether it’s a PC or gaming console.
Your PC’s RAM costs are going through the roof, but you also need to be concerned with every other device that requires memory (aka, your phone, tablet, PC, etc.).
Just when we thought we were settled into the new normal of tariff-borne PC and gaming console price hikes, next year’s devices are set to be even more expensive. You can blame the current favorite buzzword—AI—for the increasing cost of RAM and flash storage. All computers depend on memory to some extent, but those who want the biggest, best PCs or gaming devices will feel the sticker shock worst of all.
We’ve been reporting about the spiking cost of RAM prices since October. Essentially, AI data centers have such a demand for memory that the prices of SSDs (solid state drives), DRAM (dynamic random access memory), and HDDs (hard disk drives) are all ballooning in price. This is most keenly felt in the discrete PC RAM market. Corsair, one of the most popular brands for fast, gaming-ready RAM sticks, told Gizmodo over email, “Our DRAM prices have increased in response to the severe supply shortage we are seeing across the industry.”
= = = = = = =
Well, OK, AI
You know I’m not fond of AI. I don’t consider it capable of original thought—most people I know in the field agree. It can be, however, a great boon to searches and synthesis. And while I intensely dislike what it is doing in HR (summarizing and then rejecting job applications without human involvement), I have to confess that the AI in WordPress is… amazing.
I wrote an essay I entitled Love Cycle and Quote Cycle. The AI renamed the post and offered an SEO summary:
The Power of Performance: Quotes, Songs, and Memory
“Explore the impact of performance on memorable quotes and songs, from Bogart’s films to Taylor Swift’s authentic writing journey.”
Color me impressed. WordPress’ default is to allow the site to rename your post, in terms of its appearance on browser tabs and the title shown when you share the link. On the page itself: still my title, Love Cycle and Quote Cycle. I can probably turn off the default but… until it re-titles a page devoted to my love songs as “Writer Begging for Sex,” I will leave it on.
JP Morgan Doesn’t Trust It
J.P. Morgan calls out AI spend, says $650 billion in annual revenue required to deliver mere 10% return on AI buildout — equivalent to $35 payment from every iPhone user, or $180 from every Netflix subscriber ‘in perpetuity’
Welcome to my depository of AI information, AKA the place that keeps it out of my weekly column, yet makes it available to anyone interested. I will only say this once. Even though I sometimes use the term AI (artificial intelligence) to describe LLMs (large language models), please accept that this is only shorthand. An LLM is AI the way predictive text on an iPhone is AI, using the same principles, but for tasks for which it is much less well suited. An LLM is industrial strength cut and paste. See my general opinion, which will gradually fall down this lifo page (last in/first out for those of you without management degrees)
This brief note: I realize I have picked what is probably the losing side, that I am merely shaking my fist, demanding that “you kids get off my lawn,” and that sooner, rather than later, AI will become sentient. It won’t be illegal to express these opinions, just impossible on the “New Internet.” Someday when my grandchildren ask what I did in the war I want to be able to say I fought, with the only weapons I had, words and ideas. Whichever side you’re on, do enjoy this page.
========
AI the time saver
Let AI write the brief, then read every word and check every citation anyway. Have you halved your work or doubled it? Wouldn’t it take less time to proofread for citations you already know were real, rather than guessing which is which?
I am starting to think AI’s future is in the creative arts, not in the law or medicine (or, for that matter, Air Traffic Control)—anywhere that imagination is useful and not deadly.
Baby Blues On NotAI/LLM
NotAI/LLM creeps into the cultural discussion…
= = = = = = = = =
Calvin and Hobbes on NotAI/LLM
It’s not worth the time for me to find the exact date this was originally published, but it was more than 30 years ago, when the comic strip was discontinued.
Amateurs+LLM=Amateur Product
My friend and former editor Joe Brancatelli says notAI/LLM reminds him of the “first generation of word processors and design programs. Just because amateurs had the tools didn’t make them writers or designers.”
Your text or image cannot possibly be better than your imagination or your expression of that imagination in a prompt. AI can’t listen to your requirements and give you something you never knew you wanted. You can’t compose a prompt that provides you something you’d prefer, better suited to your needs, had you but known.
My personal rule is “Don’t tell a professional how to do it, just tell them the end result you want.” You have to tell AI how to it, which means you will never get a creative output guided by a professional who has done it for 10,000 hours. Have you been a professional writer or a designer for 10,000 hours? For 10 hours?
Owning a saw doesn’t make you a carpenter. Playing a piano doesn’t make you Elton John. Asking notAI/LLM for instructions on using that saw to build a house doesn’t mean you can build it, and doesn’t mean you’ll like the result, because notAI/LLM can’t suggest a skylight, a bigger window, or a pantry unless you ask it for one. You aren’t an architect. You almost certainly can’t imagine the final result in your head. If you could, you’d be an architect. NotAI/LLM is a literal idiot. It may be able to recite the building code, but is extremely unlikely to apply it properly. There is a reason architects are required to understand ambiguous building codes. NotAI/LLM doesn’t do ambiguity.
Hot Fresh Links
- * Frightening ai corruption. He drives a bus so his favorite color is yellow. (huh?)
- * She Thought She Was Venting. The AI Thought She Was Giving Orders.
- * Mourning the death of copyediting. Someone should read stories before they appear in the newspaper.
- * Richard Sutton is the father of reinforcement learning, winner of the 2024 Turing Award, and author of The Bitter Lesson. And he thinks LLMs are a dead end
- * I don’t think many stories like this could have been written about running water, radio, TV, the telephone or flush toilets. OK, I admit it, electricity and automobiles might have justifiably produced horror stories, and, in the end, they turned out to be good for society… well, at least electricity was.
- * Anthropic says it has foiled the first-ever AI-orchestrated cyber attack, originating from China — company alleges attack was run by Chinese state-sponsored group
- * Kevin Sullivan, MIT ’73 has been my friend for a half-century and is an astute observer of… almost everything. “We are, once again, being overwhelmed with marketing hype for the sake of corporate greed. We would better serve ourselves and our discussions by referring to AI as what it really is, ‘Artificial Information’ at its best, and ‘Artificial Insanity’ when it crashes on the shores of discussion requiring context, meaning, emotion, and abstract thought.”
- * Maybe Someone Should Read AI Output
- * Thank you Robert Malchman: Vigilante Lawyers Expose the Rising Tide of A.I. Slop in Court Filings
- * Thank you Joe Kashi: ChatGPT Salt Advice Triggers Psychosis, Bromide Poisoning in 60-Year-Old
- * From the past, an article about AI hallucination in the Wine industry
What was it like last time(s)?
I have often prefaced my comments by, “I’m not a luddite, but…” People familiar with that formulation will know that it is usually a way to deflect, in advance, likely criticism. Yes, I do know there is a thousand-year history of massive displacement by new technology. Just to take the most recent example, think of all the livelihoods destroyed by the Internet. The book business and the music business for example.
Hundreds of years of book stores, a century of physical recorded music: poof. Try and make a living in this country selling books (get a cut of the Ebooks or else) or making music (it’s all about the merch and Patreon, baby).
There have always been those who opposed the new ways. “If book stores and music are disappearing, it means we don’t need them.”
Let me say this about that: if the automated machinery in the textile industry (which the Luddites tried to destroy) regularly made up patterns no one asked for, or duplicated patterns from other companies without permission or payment, or tried to sell imaginary products, it might not have put all the home weavers out of work.
I keep hearing, “Today, AI is as stupid as it will ever be.” Maybe, but it’s not clear to me that it’s going to get smarter, by any realistic definition of the term. Nothing is idiot-proof because idiots are so ingenuous. Most of the literature I have seen posits that hallucinations are inherent in Large Language Models.
They are industrial level cut and paste engines, working on the same principal as predictive text on your iPhone, and just about as smart and creative.
There are a pair of Nobel-winning economists who have tackled this subject—in comic book form. Perfect for the readers of this column.
Let me conclude by saying this. There are many tasks for which LLM is well suited, and many for which it is not. Let us hope someone, somewhere can tell this difference. When all you have is a hammer, everything looks like a nail.
As the Pythons put it in the Galaxy Song, “And pray that there’s intelligent life somewhere up in space, ‘Cause there’s bugger-all down here on Earth!”
== = = = = = = = =
AI: Timely Topic
Attorney and fellow MIT alumnus Joe Kashi weighs in:
Commenting upon AI and its effect upon people and society is perhaps one of the most timely of all topics, especially given the ability to make something fake that virtually everyone will accept as real, at least in this point of time. Sadly, when the information stream that most people receive is highly questionable, then so are their judgments and actions. I am of the opinion that most people are largely reasonable but base their actions upon their perceptions of facts and reality. In the days of professional, careful journalism, this was not a problem but rather a virtue. In the era of ubiquitous Internet and AI, it sets up society to be too easily manipulated by everyone.
== = = = = = = = =
This Isn’t New
According to S.M. Oliva, author of S.M. Oliva’s Computer Chronicles Revisited, there were a few different products where guests on the Computer Chronicles claimed “AI features.” Perhaps the most notable was the Halcyon, a voice-operated LaserDisc console that famously failed to perform when host Stewart Chefeit tried to use it.
== = = = = = = = =
One Stop Used Link Shop
Links start out at the top of the page, then drop down here, for those face slap moments: “Where Did I See that link?”
Thank you Joe Kashi for two Legal Lolz links:
* AI Done Wrong
Buckle up, kiddos, because we need to talk about the absolute catastrophe that is lawyers using ChatGPT to write their briefs and then acting shocked (SHOCKED) when the AI makes up fake cases.
* AI Legal Ethics Video
…
“An attorney got caught using ChatGPT to make up non-existent case citations. Seriously, I can’t go even a week without seeing one of these.”
–S.M. Oliva, The author of Computer Chronicles Revisited website
…
“Simply stated, no brief, pleading, motion, or any other paper filed in any court should contain any citations—whether provided by generative AI or any other source—that the attorney responsible for submitting the pleading has not personally read and verified.”
—Judge Stephen I. Goorvitch
…
Theft, Harm, Hiding The Evidence
…
Microsoft CEO says the company doesn’t have enough electricity to install all the AI GPUs in its inventory
…
Nobody Likes AI Art
…
Defrauded by AI
* AI Bot Iterative Hell
* Almost Fooled
Thank you David Boccuti for this link from our Alma Mater:
AI Cheerleaders Are Entirely Too Unambitious
…
== = = = = = = = =
Chose Your Poison
To oversimplify, there are two main uses of LLM.
SEARCHING
LLMs search the net (or your internal database) and synthesize the results. This is a task for which they are very well suited. It is a timesaver for users.
POISON BECAUSE
It is killing the free Internet. No clicks, no ad revenue. So as “ai” steals all the Internet’s content, it is destroying the Internet, by making it difficult/impossible for people to click to the actual sites from which the information came.
GENERATIVE
LLM is, in reality, a giant cut and paste engine, using the same technique (predict the next word) as the iPhone’s predictive text feature. This kind of “AI” has the same intelligence and potential for creativity as an iPhone. You don’t expect your iPhone to write creative poetry. An LLM can slap together a pastiche of real people’s poetry. But is that creativity, or just sleight of hand? (Ignore the man behind the curtain)
Another use of LLMs is generative “ai.” There are several types of generative “ai” for which LLMs are spectacularly suited. And some for which it is spectacularly unsuited. By its very nature it is likely to remain so. It is incapable of creation; all it can do is follow prompts from users (soon to include porn).
Generative “ai” excels at generation of images and video. It does music—badly.
POISON BECAUSE
Already you cannot believe your eyes, so you believe nothing, or deeply believe any generative AI which supports your own already-held positions.
Freelance musicians and artists are being driven out of business because AI is cheaper. Not better, just cheaper.
Many people forget “you get what you pay for” or, to put it more elegantly, “There is hardly anything in the world that someone cannot make a little worse and sell a little cheaper, and the people who consider price only are their lawful prey.” Would you buy the cheapest pacemaker you can find? I wouldn’t.
Do we need artists and musicians? Only if you want creativity. When I purchase their services, my credo is, “Don’t tell a professional how to do it, tell them the result you want.” You can tell generative AI how to do it, but it is incapable of creating a better product which you didn’t know you wanted. It can’t create, just slap together things that have already been done.
I don’t know about you, but I think plumbers do a better job fixing the leak if they apply their personal expertise, rather than me telling them how to do it.
INCOMPETENT BECAUSE
LLMs are terrible chatbots for anyone with a problem outside of the normal range. I am being driven insane by their ubiquity. Let me be fair. For 80% of all conversations, their script is fine. Nurses use scripts. Living customer service reps (in dwindling supply) use scripts. Chatbots use them inflexibly, as holy writ, and consider any departure to be heresy (yes, anthropomorphism).
Human beings being can quickly tell if your request is out of their range of competence and either fix it or elevate it. On the other hand, AI chatbots (who lie about the fact that they are chabots) refuse to believe your problem cannot be placed in a cubbyhole. Some of them require a half hour of begging before they send you to a human. The WordPress chatbot, bless its little pea-pickin heart, said OK after a mere five minutes. Would that others were as good.
Of course “24/7 customer support” provided by chatbots is a fraud. Give me a minimum-wage customer service clerk in India anytime. At least they are trained to be polite and responsive, two things chatbots aren’t. And they lie (or used to lie) about being from Des Moines, but I find that less offensive than a lying chatbot.
“Ai” is slightly more useful to marginally competent writers (roughly 1/3 of the U.S. adult population), or even sometimes good writers looking for some ideas someone else has already had. Or poor writers who are capable of, or willing to, edit the suggested text. There will be considerable porcine aviation before that occurs.
Back in the days of print, there were books of ‘fill in the blank” letters for numerous personal and business occasions. They’ve been gone for a while, leaving people on their own, except now LLMs can instantly bring up a precisely average template. That’s what they do. They will never offer you a creative template, or a terrible template, just something in the middle. They probably scraped a bunch of those old books.
= = = = = = = = = =
Almost Fooled
A close friend of mine forwarded a video about a supplement that promised miraculous results and was promoted by Ben Carson and Sanjay Gupta, with a television spot featuring Anderson Cooper and featuring a newspaper article about the miracle cure of a movie actor. Generative AI from start to finish, no real people involved, except for the people trying to sell the supplement for a few hundred dollars. I waved my friend off just in time.
I was embarrassed by this experience: AI Bot Iterative Hell, partly because it reminded me another near miss, when I almost fell for the pigeon drop scam in Rome.
= = = = = = = = = =
A Rubber Stamp Is Not A Review
It is too bad that I am on the wrong side of history, and there’s nothing I can do to stop the universal adoption of LLM in all areas of life. Things are going to get interesting. People say it’s just like the Industrial Revolution. I don’t recall steam looms ever killing people, except employees.
It is one thing to create a machine that produces cheap, lower quality cloth, in such high volume and at such a low price that it replaces handcrafted cloth. It is something else entirely to have cheaper, lower quality writing, computer programming, management, customer service, art and medicine… replacing handcrafted human output, without supervision.
At least someone looked at the product coming out of the knitting mills. Of course they never rejected it, unless it was knitting slop. No knitting hallucinations. No factory-made cloth ever told someone how to commit suicide.
I know for a fact that the statement “a human being reviews our LLM rejections” is a lie. A rubber stamp is not a review.
I am not naming the company involved because when actual human beings (not LLMs, but actual human beings) get involved, their customer service is great. Also, I really need them and their service. And yet…
I submitted a piece of art. The LLM (clearly it was an LLM) rejected it because it “included text.”
I was told that a human had reviewed the LLM rejection. I cannot imagine how that is possible—except that the poor reviewer probably has to look at 100 pieces of art a day, staring at them until visual fatigue sets in, and it is easier to rubber stamp the rejection than actually think about it. It doesn’t cost anyone time and money… except me. Well, OK, not money, just time. And yet my time, even in retirement, is not worthless.
Paul E. Schindler Jr.
