Morality™

Golden Rule:

Do unto others as you would have them do unto you.

Silver Rule:

Don't do unto others as you wouldn't have them do unto you.

Platinum Rule:

Do unto others as they would have you do unto them.

Bronze Rule:

Don't be an asshole.

I would never dream of using an agentic "AI" that does not understand & implement these rules.

We had rule-based expert systems back in the early-mid 1980s. Time for a revival?

So in addition to the CommonSense™ module, we also need the Morality™ module - & many, many others. LLMs are NOT the answer.

Am I setting too high a bar?

If I am granting a piece of code agency, you're goddam right I am setting the bar VERY high.

In 2007, with a few other folks, I got a patent on SmartDocs - documents with intelligence built in that would represent your interests. Prolly either HP or OpenText owns that patent now.

But, I remember thinking then, I have this smart document, with intelligence built in - aka, code - how do I debug it? I don't care who develops the code, human or "AI", it has to be debugged. & even if you do some pretty good debugging, do you still feel confident releasing it into the wild world? Where there are hackers waiting around every corner to subvert your efforts?

I think I decided: "no".

What percentage of Ethereum smart contracts wind up being hacked?

---

Here's the home page/directory for my posts on Bullshit.

Pick Your Poison

So which is worse?

1. That LLMs are a complete security nightmare? This is from Bruce Schneier, a leading security expert:

   We Are Still Unable to Secure LLMs from Malicious Inputs

   Schneier quotes from this Wired article, which explains how easily a poisoned document can contain invisible-to-humans text that an LLM will read, & follow its instructions to send security keys to the bad guys.

   A Single Poisoned Document Could Leak ‘Secret’ Data Via ChatGPT

   Schneier concludes:

   It’s an existential problem that, near as I can tell, most people developing these technologies are just pretending isn’t there.

2. That LLMs will literally help troubled people pick the best poison to kill themselves with. From 404 Media:

   ChatGPT Answered 'High Risk' Questions About Suicide, New Study Finds

   The parents of a child who suicided with help from ChatGPT are suing OpenAI. Best of luck to them.

I'm going to go with #2. Killing children is IMO the worst crime there is.

---

Here's the home page/directory for my posts on Bullshit.

The Rapture Effect

"The Rapture Effect", by Jeffrey Carver, 1987, 386 pages, 119k words.

I probably got this for $0.99 or $1.99 via BookBub. Note, Kobo has this book listed as 2015, his website says it is 1987. I will confirm, this definitely reads like the 1980s. I might not have bought it if I had known it was 1987. I have liked most of Carver's stuff, but not enough to go that far back in history.

This is a 1st contact story, & an AI story. The pacing early on is weird, it's kind of like the book starts on page 50? The back story on the aliens is different, and somewhat arbitrary. The fact that their species trauma seems like it can be cured by disco music seems ... pretty 1980s. The last 1/2 of the book moves along well, & mercifully we have a happy ending.

Overall, an odd book, if you're looking to read an odd book, give it a try. Else, not.

Chatbot in a Writing Class

Via Kottke. Real data from the field. Interesting.

What Happened When I Tried to Replace Myself with ChatGPT in My English Classroom

If this were shorter, it would go in my table of names for LLMs.

One ... compares AI chatbots to “a very smart assistant who has a dozen Ph.D.s but is also high on ketamine like 30 percent of the time.”

This is almost ho-hum at this point, from 404 Media:

AI at the World’s Biggest Games Event Booked Random Meetings for Attendees

---

Here's the home page/directory for my posts on Bullshit.

Sycophants, Of Course

Here's an article from today on LLMs via Project Syndicate:

AI Is a Terrible Therapist

Investors are betting that AI chatbots can treat people who are struggling with mental-health issues and unable to access professional care. But the way large language models are currently trained makes them sycophantic, and thus fundamentally incapable of providing effective treatment.

So yet another job LLMs will royally screw up.

Why was the design decision made to make LLMs sycophantic? I'm guessing, because most people enjoy having some smoke blown up their ass. So they will come back for more. And pay more $$$ to the AI company.

Yet again, capitalism ruins everything.

---

Currently in my ongoing Bullshit project, I am reading & annotating "On Truth", Harry Frankfurt's follow-up to "On Bullshit". Mercifully short, but still a slog.

I realized while walking the other day that one of the root causes of bullshit fatigue is that all of this philosophical reading is just words, words, and more words. Very few pictures. No graphs or charts. No tables. Very few diagrams. I even wouldn't mind a code sample.

And I think that bullshit fatigue leads to bullshit contagion. Bullshit wears you out, you lapse into writing bullshit.

And I think that the fact that LLMs are mostly creatures of nothing but words is why they are a complete dead end on the road to AGI. I don't think you can ever get CommonSense™ from words alone.

---

Here's the home page/directory for my posts on Bullshit.

Don't Be Lonely, Lone Ranger

"Don't Be Lonely, Lone Ranger", by Joe Taylor, 2025, 115 pages.

My younger brother Mark, the author, sent me a trade paper copy of this book. The author of this book, Joe Taylor, "has been the director of Livingston Press at the University of West Alabama ... forever." Livingston Press published 1 of Mark's novels.

Mark said Joe sends him copies of all his stuff. Mark thought I would enjoy this novel so he passed it on to me.

Thanks Mark! I did enjoy this novel!

I would call the genre Magical Realism. The story is set in Lexington KY, where I have lived since 1981. It takes place Februarary 8-14, 2021, between the 1st & 2nd covid vaccinations. Our protagonist is laid off from his marketing job at Toyota in Georgetown KY - just north of Lexington, it is the source of all Camrys.

After days of doom scrolling bad news on his phone, he decides that he will revive the persona of The Lone Ranger, & travel the land trying to right wrongs & help people.

His spirit quest is immediately disrupted by the appearance of Coyote, the trickster god. He meets twin sisters, 1 of whom agrees to be Tonto, who join him on his quest.

The story moves along well, There are lots of wry comments about the state of things. It has a happy ending, yay!

1 thing I identified with, Taylor makes creative use of typography, particularly strikethrough. I love using strikethrough! It's like saying "here's what I really mean" followed by "here's what I am forced to say".

Plus there was a lot of riffing on exclamation points and ellipses.

I have only one complaint, and it's not the book's fault: it did have references to recent political event, i.e., the Orange Turd. I have been having nightmares about the Orange Turd and his Fascist Flunkies [worst band name ever] every 3rd night or so. I had a doozy the night after I finished this book. But, again, not the book's fault, it's hard not to get triggered, given the current Fascist state of affairs.

I'm dropping the copy Mark sent me at the Lexington Public Library to add to their collection - seems appropriate.

I'm having a copy shipped to my daughter Erica because she knew 1 of the side characters in the book! Lexington is not a big city.

Thanks again, Mark, and Joe! An enjoyable & quick read!

On Lying

"On Lying", by Augustine of Hippo, (St. Augustine to catholics), 395 a.d., 19k words. It is divided into 43 sections.

In "On Bullshit", the 1st source that Harry Frankfurt cites extensively is "The Prevalance of Humbug", by Max Black.

The 2nd source he cites extensively is "On Lying". Here's his summary:

In his essay “Lying,” Saint Augustine distinguishes lies of eight types, which he classifies according to the characteristic intent or justification with which a lie is told. ... It is only the remaining category that contains what he identifies as “the lie which is told solely for the pleasure of lying and deceiving, that is, the real lie.”

Man, I am sorely tempted to just leave it at that. Augustine is really, really wordy. Plus, the writing reminds me Adam Smith - no paragraphs. Each of the 43 sections is a single paragraph, some of them very long, very hard to read.

I have often viewed the review/summaries that I write as refactoring the text of the originals: trying to find places to create numbered or bulleted lists, break big pieces into smaller, more manageable pieces. I'm tempted to make a copy of "On Lying" and to go through & refactor it, see if I can make any more sense out of it then.

No, I think I'll just go through each of the 43 sections, & write as little as possible about each one.

1. Augustine is afraid of bullshit contagion. He says you have to read the whole thing before it makes sense.

2. Jokes are not lies.

3. One who says a false thing, but "believes or opines that to be true which he says", is not lying.

4. This section is a paragraph 1345 words long. OK, that's it, I'm done ... Reviewus Interruptus. Having read it a couple of times, there is not enough useful information in this tome to justify wading through this writing style to harvest bullet points from this word salad.

I did not identify Frankfurt's 8 types of lies. Augustine seemed to recognize the concept of "white lies". He talked about how to manage the situation where you know the person you are talking to is going to disbelieve what you tell them. Do you then tell them the opposite of what you believe to be true?

He also gets the concept of "sin" mixed in: some lies are sins, some are not, in his opinion.

If you are interested in more detail, you're on your own. It is not that long a read, but, damn, it reminds me of LLM-generated bullshit: words, words, and more words. Glib beyond glib. LLMs, though, have much better understanding & use of paragraphs.

You know, I think this does point out a general characteristic of many flavors of bullshit: it uses 10 words where 2-3 would do. The target is flooded with unnecessary words; they'll agree to the bullshit just to get it to stop!

Overall, though, I am tentatively concluding that discussions on lying may not be that informative on the nature of bullshit.

---

Here's the home page/directory for my posts on Bullshit.

Dead Country

"Dead Country", by Max Gladstone, 2023, 276 pages, 82k words. Book 1 of the Craft Wars.

I enjoyed the 6 books of the original Craft series, 2012-2018. I wasn't that crazy about his award-winning 2019 collaboration "This is How you Lose the Time War".

This book goes back to the original protagonist of Book 1 of the Craft Series, Tara Abernathy. She is now an accomplished CraftsWoman, who goes home for her father's funeral. Wouldn't you know it, all hell, literally, breaks out! Plus, even worse, an old boyfriend!

I thought this book might have 1 or 2 more major conflicts in it, but, no. 1 major conflict, a followup thereafter, then home.

Great writing. Totally a page turner. Great characters.

I'm Sure ChatGPT-5 Will Be Much, Much Better - Not!

It is increasingly inconceivable to me that this bullshit is what is currently propping up the US economy. It just get worse and worse, the chatbot responses are snarkier and more glib. This is what comes of being trained by the internet, with no idea of meaning, of morality, of common sense, with nothing but statistics to guide you.

Plus, increasingly, I think we are getting into what was described as "Degenerative AI", the downward spiral that occurs when LLMs start eating their own dogfood.

1st up, 404 Media has a post about ChaptGPT telling a guy to poison himself:

Guy Gives Himself 19th Century Psychiatric Illness After Consulting With ChatGPT

In this case, a man showed up in an ER experiencing auditory and visual hallucinations and claiming that his neighbor was poisoning him. After attempting to escape and being treated for dehydration with fluids and electrolytes, the study reports, he was able to explain that he had put himself on a super-restrictive diet in which he attempted to completely eliminate salt. He had been replacing all the salt in his food with sodium bromide, a controlled substance that is often used as a dog anticonvulsant.

He said that this was based on information gathered from ChatGPT.

Well, makes sense, I guess; both are halogens, 2nd column from the right in the periodic table. The disorder is called bromism.

Next up, "AI Weirdness", who had not posted in a while, had 2 interesting posts:

1. ChatGPT will apologize for anything

LOL, after being told how badly its recommendations worked out, ChatGPT apologizes profusely for:

1. Recommending the hire of a giraffe as CEO;
2. Trading the family cow for 3 beans;
3. Unleashing an army of cloned T. Rexes in Central Park.

The answers are so glib. Words, words, & more words. At one point it says it's going to show up in person with coffee. Increasingly, I like my characterization of LLMs as "deranged zen poets spouting bullshit". We are being smothered by this bullshit.

Shane concludes:

I wish more people understood that any "apology" or "self-reflection" from chatbots are meaningless - they're just continuing with your improv session.

A bonus post for subscribers!

Revised instructions for troubleshooting the warp confabulator

It's "bullshit all the way down".

---

Here's the home page/directory for my posts on Bullshit.

Evidence for Kardashev Type III Civilizations! Part 2

I published this post - Evidence for Kardashev Type III Civilizations! - for fun, and for the great pictures of galaxies, 2013-09-18. I love pictures of galaxies!

Wow, how time do fly! Particularly when the event is on the other side of the 1.5 missing years of COVID.

Recently, 2025-07-12, APOD published another candidate for this rogue's gallery of galaxies: NGC 2685, the Helix Galaxy.

I can see this as a galaxy entering a tunnel, maybe a warp drive, to transport the entire galaxy elsewhere? Not sure what the outer rings are doing in this scenario - maybe a cradle to hold the tunnel?

APOD, of course, provides a more reasonable explanation, but note the words: "bizarre", "peculiar", "strange", "old and stable".

Explanation: What is going on with this galaxy? NGC 2685 is a confirmed polar ring galaxy - a rare type of galaxy with stars, gas and dust orbiting in rings perpendicular to the plane of a flat galactic disk. The bizarre configuration could be caused by the chance capture of material from another galaxy by a disk galaxy, with the captured debris strung out in a rotating ring. Still, observed properties of NGC 2685 suggest that the rotating helix structure is remarkably old and stable. In this sharp view of the peculiar system also known as Arp 336 or the Helix galaxy, the strange, perpendicular rings are easy to trace as they pass in front of the galactic disk, along with other disturbed outer structures. NGC 2685 is about 50,000 light-years across and 40 million light-years away in the constellation of the Great Bear (Ursa Major).

Slopocalypse Now

I really felt like "The (AI-driven) Bullshit Apocalypse", which I posted 2024-06-28, was a worthy meme. I apparently really suck as a memetic engineer, because none of the memes that I create that I think deserve life seem to make it.

Meanwhile the title of Gary Marcus's post today totally rocks:

Slopocalypse Now
The Horror
The End of Quality Content
The Imminent Enshittification of the Internet
LLMs are creating a huge sanitation problem that will probably never be solved

The bullshit being generated by LLMs that is overwhelming the world is now generally being called "AI Slop". So "Slopocalypse Now" is a great meme, a better & catchier encoding of my "Bullshit Apocalypse". But, I think totally the same concept.

Marcus is a fan of Doctorow's "enshittification" concept. He gives example after example of real content being crowded out by AI slop.

I will disagree with his use of the term "Imminent". I used to refer to the "coming Bullshit Apocalypse", but have recently concluded, it's no longer coming: it's here and now.

---

Meanwhile, Cory Doctorow revisits his discussion of centaurs vs. reverse-centaurs with regard to writing code:

AI software assistants make the hardest kinds of bugs to spot

I think I mentioned last post, that LLMs can be a great tool acting as an assistant to a creative type - in this case, someone writing code. Tedious tasks can be passed off to an LLM, with the results being carefully monitored until the coder gets of feel for where the LLM can be trusted & where it cannot be trusted. This is the centaur case.

When your boss orders you to use an AI code generator and then makes you responsible for reviewing & cleaning up the output, this is the reverse-centaur case. All coders dislike code review, and "automation blindness" is a known problem. It is easier to write the code yourself. The reverse-centaur, the "human in the loop", becomes the fall-guy, the "accountability sink".

[There was a period, maybe in the early-mid 00s, where pair programming was a thing. I wonder if that could be modified to have 1 of the pair be an LLM?]

Ugh, I am glad I am no longer coding, to where I would have to deal with this: coding LLMs hallucinate names for support libraries. [Note, increasingly, I think all LLM output should be considered hallucinations. LLMs have no connection to the real world, they are just hallucinating words about it. Statistically, they are mostly right - but, as is the case with statistics, not always.]

This creates a vulnerability for AI-assisted code, called "slopsquatting," whereby an attacker predicts the names of libraries AIs are apt to hallucinate and creates libraries with those names, libraries that do what you would expect they'd do, but also inject malicious code into every program that incorporates them:

https://www.theregister.com/2025/04/12/ai_code_suggestions_sabotage_supply_chain/

Meanwhile, the statistics on LLM coding are ... bullshit.

while usage of AI coding assistants is up (from 76% last year to 84% this year), trust in these tools is plummeting – 33%, with no bottom in sight. 45% of coders say that debugging AI code takes longer than writing the code without AI at all. Only 29% of coders believe that AI tools can solve complex code problems.

---

Here's the home page/directory for my posts on Bullshit.

Agents of Bullshit

I didn't think this article in yesterday's Herald-Leader was worth referencing, but then Gary Marcus weighted in today on LLM-based Agents, which were supposed to be a Big Thing this year. Not going too well so far.

Here's the Herald-Leader article:

Will AI help or wreck your summer vacation?

The author, visiting Taiwan, asked ChatGPT to recommend a noodle shop nearby that took credit cards, as she was low on local currency. She had some delicious noodles, but was "oh shit" when the proprietor demanded payment in cash.

But, she points out, at least the place she got sent to existed. LLMs are sending people to places that don't exist. They call this "hallucinations", but ALL LLMs DO IS HALLUCINATE, ALL THE TIME! They do not have a model of the real world, all they have are words they have read on the internet.

Here's Gary Marcus's post:

AI Agents have, so far, mostly been a dud

Marcus quotes numerous bad reviews of LLM based "Agents". He summarizes it as:

to a paraphrase an old saying that goes back to 1969, “To err is human, to really screw up takes an AI agent.”

LLM "Agents" have "failure rates of 70% on some tasks".

1 thing I don't think is getting enough attention: as with almost every computer system going back pretty much forever, security sux.

As a consequence of their superficial understanding of what they’re being asked, current agents are grossly vulnerable to cyberattacks.

I like this name for the downward spiral that occurs when LLMs start eating their own dogfood, in a quote from Rohan Paul:

The team calls this road Degenerative AI, where models feed on their own flawed output until the answers collapse. [My bold]

So LLM-based Agents are, as ever, just hype, just bullshit, just techbros wanting to maximize their stock options.

LLMs are really good with words. As an assistant to creative types, in words, maybe science, they can be really useful.

Anyone thinking they are ready to really start doing stuff on their own, a la "Agents", is delusional. They are way, way too untrustworthy. And the limitations of the LLMs - no world model, no math engine, no physics engine, no chemistry engine, no biology engine, no social or moral engine - to me mean that they are a total dead end on the road to any kind of AGI.

I don't think agents can work without at least some rudimentary AGI. I think I discussed this in a little more detail in my post of 2025-02-17 "CommonSense™"

LLMs do have a great language engine, but that ain't enough.

---

Here's the home page/directory for my posts on Bullshit.

Saturation Point

"Saturation Point", by Adrian Tchaikovsky, 2024, 135 pages, 41k words.

This story is climate fiction with a vengeance. The earth now has, around the equator, the Hygrometric Dehabitation Region, or the “Zone”. Wet bulb temperature never goes below 95 degrees, so it is quickly fatal to humans - actually, to all warm-blooded species. It's not the heat, it's the humidity - which causes lots of equipment failures, & if your equipment fails, you die.

After being the only survivor of a research team 20 years ago, our protagonist signs on to go back into the zone. Deaths start almost immediately. Brutal.

Meanwhile, the Zone thrives with life: plants & cold-blooded animals. But it is death for warm-blooded animals. Some interesting biology here, about how much better adapted cold-blooded species are for this environment.

Tchaichovsky does a great job with this material. For a Brit, he understands American zeitgeist very well:

My family were never gun-havers. We were naturalised immigrants. We weren't ever white enough for our neighbours to be happy with us exercising our second-amendment rights.

I totally didn't expect the turns this plot would take, particularly the ending. Very well done.

Stochastic Parrots

[Updated 2025-11-29]

In my "Semiotic Zombies" post of 2025-07-12, I was pointed at "Language Machines" subtitled "Cultural AI and the End of Remainder Humanism". I bought a copy of the eBook, & have finished reading the introduction. This book is heavy going, I think I may read it & review/summarize it a chapter at a time. Review/summary of the intro will be forthcoming.

Meanwhile, several new names for LLMs. I particularly liked "Stochastic Parrots", attributed to "AI ethicists Timnit Gebru and Emily Bender" in a 2021 paper "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜".

The intro to the book also completely validates the "semiotic" part of my "semiotic zombies" moniker. And the "success" of LLMs is indeed the revenge of French philosophers: Chapter 1 of the book is titled "How the Humanities Lost Language: Syntax, Statistics, and Structure". LOL, the revenge of the humanities professors!

I am discovering so many names for LLMs, maybe all these names will help us understand what these things really are & how they should be used. So I'm going to start a list here, which I will continue to update.

I will order these in the order they appeared in this blog.

#	Name	Author	Date
1	Bullshit Generators	me	2024-06-28
2	Bullshit Machines	Bergstrom & West	2025
3	The Poetry Machine	Seth Godin	2024-07-11
4	Language Machines	Leif Weatherby	2025
5	Cultural AI	Leif Weatherby	2025
6	Semiotic Zombies	me	2025-07-11
7	The Talking Dog	Seth Godin	2025-07-17
8	Semiotic Operators	Leif Weatherby	2025
9	Stochastic Parrots	Gebru & Bender	2021
10	Culture Machines	Leif Weatherby	2025
11	Ideology Machines	Leif Weatherby	2025
12	Acme Self Writing Word Processor	me	2025-08-01
13	Autocomplete on Steroids	Gary Marcus	2021
14	Bullshit	Gary Marcus	2019
15	Deranged Zen Poets Spouting Bullshit	me	2025-07-14
16	Slot Machines	Cory Doctrow/ Deciphering Glyphs	2025-08-16
17	Superhuman Bullshit Generators	me	2025-11-09
18	Black-box Talking Dogs	me	2025-11-17
19	Clanker	William Tenn	1958
20	Autocomplete Robot	me	2026-03-12

Note, #2 above came from this post, which links to an impressive course titled "Modern Day Oracles or Bullshit Machines?" by Bergstrom & West, who are U of Washington professors. I just ordered their 2020 book "Calling Bullshit". These guys may be the bullshit experts we need!

I just came up with #12 this afternoon. I did a lot of work with early word processing & word processors - there were dedicated systems, Wang was big on that - in the late 70s, early 80s. DEC made 1, PDP-8 based, it had a custom keyboard with lots of different color keys for editing functions. There was a Canadian word-processing company who had a product Word-11, that came with decals that you put on the front of the keys of DEC VT-100 monitor keyboards. KMN!

LOL, I checked my Pocket Guides box in the basement. I have been trying to get someone to take these for years decades. Internet Archive is my next target.

I think this is another way of saying "Autocomplete on Steroids". Oops, that should be in the table. #13.

---

Here's the home page/directory for my posts on Bullshit.