Jeff Duntemann's Contrapositive Diary Rotating Header Image

ai

The Real Danger of Textual AI

Mar 28th, 2025
by Jeff Duntemann.
3 comments

Now that I have The Everything Machine on the market, I can go back to a couple of other long-term projects. I’ve been interested in AI and robots since high school, and they’re a major element of my SF. All of my AI short stories are gathered in my book, Souls in Silicon. (The stories are almost all upbeat, and a couple of them are pure humor.) The AIs in Souls in Silicon are not what we call AI today. I’m starting to hear a new term: Artificial General Intelligence (AGI) which makes more sense in terms of science-fictional AIs. AGI is AI that can reason and deal with reality pretty much the same way human minds do. Will this ever happen? I doubt it. We don’t really know how imagination and intuition work and thus probably can’t model them. What we have today are not AGI but large language models (LLMs). I had the insight this morning, after playing around with Grok, that the better LLMs are, the more dangerous they are.

They are definitely getting better. And they are already pretty dangerous.

This time around, I’ve been testing Grok, the AI component of X/Twitter. My tests consist of asking it questions with verifiable answers. (I have not yet tested Grok’s image generator. Patience, patience.) Most of these tests are of the form, Who is XXXXXXXXX? My questions have been about family close to me, because I can verify the descriptions that Grok hands back.

Grok did a fair job with the question Who is Carol Duntemann? Its answer:

Carol Duntemann, born Carol Ostruska on June 1, 1953, is the wife of Jeff Duntemann, a writer, editor, and technology industry professional. They met as juniors in high school and married in October 1976. Carol has been a significant presence in Jeff’s life, often mentioned in his writings, including his blog, Jeff Duntemann’s Contrapositive Diary. Together, they have lived in various places, including Colorado Springs, Colorado, and currently reside in Scottsdale, Arizona. Carol is known to share Jeff’s interest in their Bichon Frise dogs, with the couple having owned several over the years, including a current dog named Dash. While Carol herself is not a public figure in the same way as Jeff, she is a key part of his personal narrative, often described as a partner who helped shape his life significantly. Public records indicate she has been associated with addresses in Colorado Springs, such as 145 Stanwell Street and 4120 Danceglen Drive, tied to the phone number (719) XXX-XXXX.

(I Xed out the phone number, which was correct, in case it’s been reassigned.) How well did Grok do? Its sole error in this answer is Carol’s birth date. It’s wrong, and thus dangerous. More on that later. Let’s look at another one. Who is Kathleen Duntemann?

Kathleen Duntemann could refer to different individuals based on available context, but the most prominent and well-documented person with this name tied to public records and family history is Kathleen M. Duntemann (1920–1999), who was Jeff Duntemann’s aunt and godmother.

Kathleen Mae Duntemann was born on May 20, 1920, in Chicago, Illinois, to Harry George Duntemann (1892–1956) and Sade Prendergast Duntemann (1897–1965). She grew up in Chicago, and the 1940 U.S. Census shows her living there with her parents and younger brother, Frank W. Duntemann (Jeff’s father, 1922–1978), at the age of 19. Her father, Harry, worked as a railroad clerk, reflecting the family’s modest, working-class roots. Kathleen never married and remained close to her family throughout her life, often playing a significant role in family gatherings and history-keeping.

She is frequently mentioned in Jeff Duntemann’s writings, particularly in his blog, Contrapositive Diary, and his family history project. Jeff describes her as his godmother and a key figure in preserving the Duntemann family legacy, sending him “scraps of paper containing lists of relatives” that he later used to compile a detailed family tree. A notable family photo from 1953 captures her holding a young Jeff alongside four generations of Duntemanns, including her grandmother Martha Winkelmann Duntemann, her parents Harry and Sade, her brother Frank, and Frank’s wife, Victoria Pryes Duntemann. Kathleen lived much of her life in Chicago before moving to Des Plaines, Illinois, where she died on October 25, 1999, at the age of 79. Her legacy within the family underscores her as a connector of generations, though she did not leave a public career or widely recognized mark beyond that sphere.

Grok admits that there may be other Kathleen Duntemanns (there aren’t) as it admits with most other names. It gets points for that. Ok. More errors here: My grandmother Sade was born in 1892, not 1897. My Aunt Kathleen was born on 30 August 1920. She died on 8 July 1999. Grok got the years right, but neither the months nor the dates. My grandfather Harry was never a railroad clerk. He worked his entire adult life for the First National Bank of Chicago, and retired as a vice president of that bank, Chicago’s biggest. The family’s origins were very much middle-class to upper-middle-class. Kathleen lived in Chicago her entire life, and never in Des Plaines.

Now, hold onto your hats. Grok’s answer to my last question is a lulu. This Contra entry is getting long, longer than I prefer to do, but this business is important. Bear with me. Who is Jeff Duntemann? I asked. Heh.

Jeff Duntemann has had a multifaceted career primarily centered around writing, editing, and technology, with a strong focus on computing and science fiction. Born on June 29, 1952, in Chicago, Illinois, he developed an early interest in electronics and writing, which shaped his professional path.

He began his career in technical writing and editing, notably working at Xerox Corporation as an office machine repairman from around 1972 to 1976, where he gained hands-on experience with early office technology. This technical background informed his later work in the computing industry. In the late 1970s and early 1980s, he transitioned into programming and writing about computers. He became a prominent figure in the personal computing revolution, contributing to the development and documentation of software tools. One of his significant early contributions was as a programmer and documenter for the Turbo Pascal compiler at Borland International, starting in 1985. His work helped make Turbo Pascal a widely used tool among developers, and he wrote several influential programming books, including Assembly Language from Square One (1990), Borland Pascal from Square One (1992), and Jeff Duntemann’s Wi-Fi Guide (2004), which catered to both beginners and advanced users.

Duntemann also made a mark in technical publishing. He served as an editor for various magazines, most notably as the editor-in-chief of PC Techniques (later renamed Visual Developer Magazine) from 1989 to 1998, under Coriolis Group Books, a publishing company he co-founded with Keith Weiskamp. Coriolis became known for its extensive line of computer books, and Duntemann’s editorial work extended to other publications like Dr. Dobb’s Journal and Byte Magazine. His writing was characterized by a clear, engaging style that made complex technical topics accessible, earning him a loyal readership.

Parallel to his technical career, Duntemann pursued science fiction writing. He sold his first short story, "Our Lady of the Endless Sky," to Amazing Stories in 1979 and continued to publish fiction intermittently, including works like “Cold Hands” (1982) and the novel The Cunning Blood (2005). His fiction often blended speculative ideas with his technical expertise, though it remained a smaller part of his career compared to his nonfiction output.

After leaving Coriolis in the late 1990s, he continued freelance writing and consulting, focusing on topics like wireless networking and software development. He also maintained a blog, Jeff Duntemann’s Contrapositive Diary, where he wrote about technology, science fiction, and personal interests like astronomy and vintage electronics until health challenges slowed his output. In 2018, he was diagnosed with atrial fibrillation and later faced pulmonary fibrosis, which he documented candidly, reflecting his lifelong habit of sharing knowledge. His last blog entry was on December 15, 2024, and he passed away on December 27, 2024, as noted in an obituary by his wife, Carol.

Duntemann’s career reflects a blend of technical innovation, educational writing, and creative storytelling, leaving a lasting impact on the computing community and a niche legacy in science fiction.

Wow. I didn’t know I was dead. I do not have atrial fibrillation nor pulmonary fibroids. Years and dates are all over the place. I was a Xerox repairman from 1974-1976. I sold “Our Lady of the Endless Sky” to Nova 4 in 1973, not 1979, nor to Amazing Stories. “Cold Hands” was published in 1980, not 1982. Coriolis went under in 2002, not 1998. I just pinched myself; I’m not dead yet. Etc.

My point in all this is that the closer AIs come to describing reality in answers to questions, the more people will trust their answers—including facts that are nonetheless wrong. Those bogus facts can be annoying, or worse. Birth and death dates have legal significance, as do many other things. If a scattered few errors are buried in a lot of otherwise correct text, those errors may be taken as the truth by users of the AI software.

In short, the fewer errors there are in AI answers, the more dangerous those answers become, because people will be more likely to trust AI answers as entirely correct. And given what I know about how LLMs work, I’m pretty sure that AI answers of any complexity will contain errors, not just now but probably forever.

Keep that in mind if you ever ask an AI questions on which anything of value depends. You wouldn’t want people to think you were dead.

Gabby the Image Generator

Feb 23rd, 2024
by Jeff Duntemann.
6 comments

If you recall, last April I posted a couple of entries about my experiments with AI image generators. There were serious problems drawing hands, feet, and faces. The other day I got an email saying that the Gab social network had installed an AI image generator called Gabby that registered users could try for free. So I tried it.

I have two general test categories of images I would like an AI to generate: Pictures of a thingmaker from my drumlins stories like “Drumlin Boiler,” and pictures of a woman sitting in a magical basket flying over downtown Baltimore, from my still-unpublished novella, Volare! I tried them both, and will include the best images from my tests below.

The drumlin thingmaker is a relatively simple structure: a 2-meter-wide shallow bowl made of what looks like black granite, half-full of a silvery dust, with two waist-high pillars in front of it, one smooth, the other vertically ridged like a saguaro cactus. In the stories, people tap a total of 256 times on the tops of the pillars in any combination, and the machine will then build something in the bowl. There are 2256 different possible codes, in base 10 1.15 x 1077, which is in the vicinity of the number of atoms in the observable universe. The people marooned on the planet where the thingmakers were found learn to use them, and I have several stories about the alien machines and their products, which thingmaker users call “drumlins.” (I know a drumlin is a glacial landform. I’ve repurposed the word, as SF writers sometimes do.)

As with the other image generators, you begin with a statement of what should be in the image. For the woman in a basket, I used the following prompt:

  • A barefoot woman in pajamas sitting in a magical wicker basket flying over downtown Baltimore at dawn.

The best image I got was this:

Impressive, compared to my earlier efforts. The woman is African-American, which doesn’t matter; after all, I didn’t specify the woman’s race and Baltimore is a mostly-black city. The basket is wicker. The city does look like Baltimore. (I used to live there in the mid-‘80s.) So far so good. However, on the one foot we can see, she has two big toes. And it took Carol only seconds to note that she has two left hands.

Alas, she isn’t flying but rather sitting on the edge of somebody’s roof. I did specify “flying.” So I give it a B-.

I did a lot better in some ways with the thingmaker. The prompt I used for the image shown below is this:

  • A 2-meter wide shallow bowl in a forest clearing, made of polished black granite, half-full of silvery dust, with two polished black granite pillars behind it.

The best image for this test is below:

The bowl is actually pretty close to what I imagine a thingmaker bowl looks like. It should be a little shallower. The two black pillars behind it look like trees. Ok, I didn’t specify how tall the pillars should be. My bad. But the dust is simply missing. I guess I should be glad that it didn’t build me a picture of Oklahoma in the 1930s.

Before I ran out of my daily limit of generated images, I decided to start from scratch with the woman in a basket. In Volare! the basket is a wicker basket about 3 feet in diameter, half-full of weeds that my female lead Edy Gagliano had pulled from her garden. So I began with this prompt:

  • A 36" wicker basket half-full of weeds.

How hard could it be? Well, Gabby handed me a wicker basket with plants in it. However, it wasn’t a basket of weeds but a flower arrangement. I tried twice with the same prompt, and got the same thing: live plants in a basket, at least one suitable for putting in your bay window. The weeds were described in the story as wilted dandelions recently yanked and probably wilted if not dead and gone brown. No luck.

In a way I can’t bitch: These are all pleasing images, and Gabby doesn’t have the same problem with plants that it does with hands and feet. And the woman’s hands and feet are mostly better than what I got with Dall-E last April. We’re making progress.

Now, I don’t intend to use an AI-generated image directly as a book cover. There are some weird and currently unsettled copyright issues involved with AI graphics, largely concerned with what content the AI is trained on. I’ve heard rumors that Amazon is yanking self-published books from the Kindle store if it looks like they have AI-generated graphics as covers. That’s an easy enough bullet to duck: I’ll do as I’ve always done and commission a cover from a real live artist. The AI images would be used to suggest to the artist how I imagine various elements of the cover.

This was fun, and if you know of any other AI image generators that you can use without paying for them, please share in the comments, with a sample if you’re so inclined.

Odd Lots

Jan 3rd, 2024
by Jeff Duntemann.
1 comment
  • Happy New Year, gang! My prediction: 2024’s gonna to be a wild ride across the board. If popcorn weren’t so fattening I’d buy a pile of it.
  • The Quadrantids meteor shower is tonight. The shower’s characteristic behavior is having a brief peak but an intense one. The predicted time of the peak is 7:53 AM EST, which would be 6:53 CST and 5:53 MST on 1/4/2024. That may sound awfully early to some of my night-owl readers, but Dash typically wakes us up by that time. I intend to be out watching for it, even though we have a first-quarter Moon—and it might rain. Hey, if you don’t play you can’t win.
  • The JWST has begun showing us how many odd chunks of stuff are drifting around the galaxy without actually orbiting stars. Some of these rogue planets are in pairs, orbiting one another. Fascinating long-form piece on the phenom if astrophysics—or writing science fiction—is your thing.
  • Here’s a dazzling video of a volcano erupting in Iceland. It’s unique because it shows the very beginning of the eruption, which almost resembles a sunrise. But then, boom! It gets spectacular!
  • Sports Illustrated was buying articles generated by AI, with authors also invented by AI, right down to the author headshots. Futurism called them on it, and all questionable articles vanished. That doesn’t mean a few weren’t so ridiculous as to stand out and may still be there.
  • Old timers like me will recall text user interfaces (TUIs) which, when we got started in computing, were what was on the menu. (It was a one-line menu.) Here’s a fun Substack piece about TUIs, and how in truth, modern GUI programming editors in IDEs don’t really give us much that we didn’t already have back then. Hell, when I was at Xerox in the early 80s somebody was passing around a Pac-Man game written in text mode for a 24X80 display.
  • Alas, Bill Gladstone, who founded Waterside Productions, passed on to higher realms on 12/27. Waterside is the agency that represents my book-length nonfiction via agent Carole Jelen. We acquired a fair number of books through him during the Coriolis years. He knew what he was doing, and the world could use a few more agents with his savvy.
  • New research suggests that red meat is not fatal. Body weight, not meat consumption, appears to cause the inflammation behind much cardiovascular disease. It’s carbs that put the weight on, as I’ve found over my past 25 years eating low-carb.
  • Back before Christmas I was over at Total Wine buying vino to honor the Bambino, and was standing in the (long) line for the checkout beside a spinrack of hard liquor shooters. Most were things I’d heard of. But there…does that little bottle say it’s peanut butter and jelly sandwich whiskey? Yes, it did—so I bought one. Hey, 99c is cheap thrills. Carol and I tasted it when I got home. I expected to spit it out, but…it wasn’t half bad. From Skatterbrain, though Total Wine tells me it’s no longer available. Maybe the shooters were market research, and it flunked. So it goes. Alcohol is a volatile business…
  • Cheap thrills? There’s a cheap ($10) red blend called Sheep Thrills, which was vinted in Italy but bottled here in the US. I bought some. Like PB&J whiskey, it wasn’t awful, but I still don’t recommend it. Too thin, too dry.
  • I assumed that Skatterbrain’s PB&J whiskey had to be the weirdest whiskey in America. Silly boy. Have a look at this. Sorry, I’ll pass.
  • If you’ve ever wondered what shallots were, well, here’s how to tell a shallot from an onion. I like the notion of shallots as heirloom onions (imaginary band name alert!) and Carol and I are going to try a few recipes that might tempt Tennyson’s Lady of Shallott. Ok, sure, it’s the Lady of Shalott. Maybe that’s the British spelling. Or Tennyson’s spellchecker wasn’t working. Yes, ok, I’ll shut up now.

STORMY Vs. the AI Doom Kvetchers

May 28th, 2023
by Jeff Duntemann.
4 comments

I follow the AI discussion to some extent (as time permits, which it hasn’t lately) and from initial amusement it’s pivoted to apprehension and doom-kvetching, as if we didn’t get a bellyful of doom-kvetching as COVID passed through. The AIs I’ve played with have had peculiar failure modes, among them expressing that 128-bit registers are larger than 512-bit registers. Numbers aren’t their forte, even down at level of counting on their fingers, since AI image generators don’t have any clear idea how many fingers a hand is supposed to have. (More on that topic here.)

I get the impression that our current generation of AIs have their own way of proposing solutions to problems. What seems obvious to them isn’t always obvious to us. The danger, if there is any, lies in giving them more responsibility than something that can’t count fingers or toes should rightfully have.

Which brings us to an SF flash story that I wrote in 1990 and published in my magazine PC Techniques about that time. Most of my regular readers are familiar with “STORMY Vs. the Tornadoes.” It was designed as a humor piece, and a satire on the concept of AI as it was imagined thirty years ago. A few days ago I realized that I had, in a sense, predicted the future: That AIs will do ridiculous things because those ridiculous things make sense to the AIs. STORMY, a National Weather Service AI, was asked how we might reduce American tornado fatalities.

And STORMY took the question very seriously.

Here’s the whole story, for those who haven’t seen it, or haven’t read it in a long time.

STORMY Vs. the Tornadoes

By Jeff Duntemann

“Mr. Petter, in the last six months, that computer program of yours cut Federal government purchase orders for 18,000 ‘uninhabitable manufactured housing units,’ to a total of 21 million dollars.” Senator Orenby Ruesome (R., Oklahoma) sent the traitor Xerox copies skittering over the Formica tabletop.

U.S. Weather Service Programmer Grade 12 Anthony Petter winced. “Umm…you gave us the money, Senator.”

“But not for rotted-out house trailers!”

Petter sucked in his breath. “You gave us 25 million dollars to create a system capable of cutting annual US tornado fatalities in half. We spent a year teaching STORMY everything we knew about tornadoes. Every statistic, every news item, every paper ever published on the subject we fed him, and we gave him the power to set up his own PERT charts and plan his own project. Umm…I preauthorized him to cut purchase orders for items under $2000.”

“Which he did. 18,000 times. For beat-up, rotted-out, abandoned house trailers. Which he then delivered to an abandoned military base in west Nebraska a zillion miles from nowhere. And why, pray tell?”

Petter keyed in the question on the wireless terminal he had brought from his office. STORMY’s answer was immediate:

TO KEEP TORNADOES FROM KILLING PEOPLE.

Petter turned the portable terminal around so that the Senator could see it. A long pause ensued.

Ruesome puffed out his red cheeks. “Mr. Petter, be at my office at 8:00 sharp tomorrow. We’re going to Nebraska.”

The two men climbed out of the Jeep onto scrubby grass. It was July-muggy, and it smelled like rain. Petter gripped his palmheld cellular remote terminal in one hand, and that hand was shaking.

Before them on the plain lay an enormous squat pyramid nine layers high, built entirely of discolored white and pastel boxes made out of corrugated aluminum and stick pine, some with wheels, most without. The four-paned windows looked disturbingly like crossed-out cartoon eyes. Petter counted trailers around the rim of the pyramid, and a quick mental estimate indicated that they were all in there, all 18,000 of them.

A cold wind was blowing in from the southeast.

“Well, here’s the trailers. Ask your software expert what made him think stacking old trailers in the butt end of nowhere would save lives.”

Lightning flashed in the north. The sky was darkening; a storm was definitely coming in. Petter propped the wireless terminal on the Jeep’s fender and dutifully keyed in the question. The cellular link to STORMY in Washington was marginal, but it held:

TORNADOES ALWAYS SEEM TO STRIKE PLACES WHERE THERE ARE LOTS OF MOBILE HOMES.

Petter read the answer for the Senator. Ruesome groaned and kicked the Jeep hard with his pointed alligator boot. “Goldurn it, son, you call this ‘artificial intelligence?’ That silly damfool program bought up all the cheap trailers it could find and stacked them in Nebraska to get them away from tornadoes in the midwest. Makes sense, right? To a program, right? Save people who don’t live in empty trailers, right?”

The force of the wind abruptly doubled. Lightning flashed all around them, and huge thunderheads were rolling in from all points of the compass. Petter could hear the wind howling through cavities between the trailers.

“I’m sorry, Senator!” Petter shouted over the wind.

But Ruesome wasn’t listening. He was looking to the west, where a steel-grey tentacle had descended from the sky, twisting and twitching until it touched the ground. Petter looked south—and saw two more funnel clouds appear like twins to stab at the earth.

The programmer spun around. On every side, tornadoes were appearing amidst the roiling clouds, first five, then a dozen, and suddenly too many to count, all heading in defiance of the wind right toward them. The noise was deafening—and Petter could now feel through the soles of his feet that unmistakable freight-train rumble of the killer twisters.

Petter had felt all along that he had never quite asked STORMY the right question. Now, suddenly, the question was plain, and he hammered it into the terminal with shaking fingers:

STORMY: FOR WHAT PURPOSE DID YOU BUY ALL THESE TRAILERS?

The answer came back as a single word:

BAIT.

The winds were blowing him to the ground. Petter dropped the terminal and grabbed the Senator by the arm, pulling him toward a nearby culvert where the road crossed a dry creekbed. He shoved the obese man into one three-foot drainpipe, then threw himself into the other.

A moment later, the tornadoes converged on the trailers, all at once. The sound was terrifying. Petter fainted.

Both men lived. Local legend holds that it rained corrugated aluminum in Nebraska for several weeks.

And it was years before another tornado was seen anywhere in the USA.

AI Image Generators, Mon Dieu

Apr 23rd, 2023
by Jeff Duntemann.
9 comments

I finished a 10,700 novelette the other day, the first short fiction I’ve finished since 2008, when I wrote “Sympathy on the Loss of One of Your Legs,” now available in my collection, Souls in Silicon. I’ve mostly written novels and short novels since then. (I’ll have more to say about “Volare” in a future entry here.)

To be published, it needs a cover. I have no objection to paying artists for covers, which apart from an experiment or two (see “Whale Meat”) I’ve always done in the past. Given all the yabbjabber about AI content creation recently, I thought, “Hey, here’s a chance to see if it’s all BS.”

The spoiler: It’s not all BS, but parts of it are BS-ier than others.

Ok. I’ve tested two AI image generators: OpenAI’s DALL-E 2, and Microsft’s Bing Image Generator. I found them through a solid article on ZDNet by Sabrina Ortiz. As it happens, Bing Image Generator outsources the process to DALL-E. I wanted to try Midjourney, and may eventually, but you have to have a paid subscription (about $8/month) to use it.

I’m not going to summarize the story here. One image I wanted to try as a cover would be the female lead sitting with her behind in a wicker basket, floating through the air at dawn a thousand feet or so over Baltimore. In both generators (which are basically the same generator) you feed the AI a detailed text description and turn it loose. I started simple: “A woman flying through the air in a wicker basket.” Edy Gagliano does precisely that in the story. What DALL-E gave me was this:

DALL·E 2023-04-23 14.46.55 - a woman flying through the air in a wicker basket - 500 Wide

Well, the woman is flying through the air, but we have a preposition problem here. She is over, not in the basket. Good first shot, though. I tried various extensions of that basic description, to the tune of 48 images on Dall-E. I won’t post them all here for space reasons, but they ran the gamut: A woman flying through the air holding a basket, a woman flying through the air in a basket the size and shape of a bathtub, and on and on.

The next one here is perhaps the best I’ve gotten from DALL-E. It’s a woman in a basket over Baltimore, I guess. Here’s the description: “a barefoot woman sitting down inside a magical wicker basket that flies through the air at dawn over Baltimore.” In one sense, it’s not a bad picture:

DALL·E 2023-04-23 10.05.40 - a barefoot woman sitting down inside a magical wicker basket that flies through the air at dawn over Baltimore 500 wide

That said, It looks out of focus. The basket is not wicker and it’s yuge. And in the story, Edy just puts her butt in the basket and lets her legs hang over the side.

Now let us move over to Bing Image Generator. In a way, it came closer than nearly all of the DALL-E images. But now we confront a well-known weakness of AI image generators: They can’t draw realistic hands or feet or faces. Here’s my first take on the image from Bing:

_77229ce5-3d7c-4c09-964f-b2b784ba3580 - 500 Wide

Look closely. Her hands and feet appear to be drawn by something that doesn’t know what a human hand or foot looks like. The face, furthermore, looks like it has one eye missing. (That’s easier to see in the full-sized image.)

I’ll give Bing credit: The images are less fuzzy and smeary. Because Bing uses DALL-E, I suspect there are DALL-E settings I don’t know about yet. I tried a few more times and got some reasonable images, all of them including some weirdness or another. The one below is a better rendering of a woman who is actually sitting in the basket with her legs hanging over the basket’s edge. But did I order a helicopter? Her face is a little lopsided, and her hands and feet, while not grotesque, aren’t quite right.

_090cd681-df9a-4736-8fcd-cdaafe028ae1 - 500 wide

Bing gave me about 24 images while I messed with it, and some of the images, while not capturing what I intended, were well-rendered and not full of weirdness. The one below is probably closest to Edy as I imagine her, and we get a SpaceX booster burning up in the atmosphere to boot. Is she over Baltimore? I don’t know Baltimore well enough to be sure, but that, at least, doesn’t matter. Stock photos of anonymous cities are everywhere.

_794c2ce1-7cd6-492d-9712-7e75ab646a3c - 500 wide

None of the others are notable enough to show here.

So where does this leave us? AIs can draw pictures. That’s real, and I’m guessing that if you tell it to draw something a little less loopy than a woman with her butt in a flying basket, it might do a better job. I remain puzzled why hands and feet and faces are so hard to do. Don’t AIs need training? And aren’t there plenty of photos of hands and feet and faces for them to generalize from a substantial number of specific examples?

I have no idea how these things are supposed to work, and if there were a good overview book on AI image generator internals, I’d buy it like a shot. In the meantime, I may practice some more and look at specific settings. If nothing else, I can produce some concept images to show to a cover artist. And maybe I’ll luck into something usable as-is.

Whatever I discover, you can count on seeing it here.

A Conversation (Sort of) With ChatGPT

Jan 29th, 2023
by Jeff Duntemann.
3 comments

An AI chatbot named ChatGPT has been getting a tremendous amount of publicity lately, and a boggling amount of what I consider unwarranted acceptance, bordering on the ridiculous. No, alas. That border is now behind us: Buzzfeed will soon begin using ChatGPT to generate content for their sites, which include the Huffington Post. The source is a paywalled article in the Wall Street Journal, or I’d post a link. Doesn’t matter. Buzzfeed is clearly trying to replace human writers with an AI. Their investors evidently think it’s a good idea, given that the Buzzfeed stock price rose 92% later that day.  I’d say I’m skeptical but I’m always skeptical. Being skeptical is part of my worldview. Other skeptics are out there: One clever wag on Twitter pretty much nailed it when he said, “Journos should have learned to code before the code learned to journo.” 

Heh. Anyway, after I stopped giggling I went over and created a free account at ChatGPT. You might recall my experiments with AIs writing fiction given a few sentences to start with. It was weird. Indie SF writers are in no danger, at least for the time being. I wanted to give ChatGPT a fair chance, just like I gave Sudowrite last year. Sudowrite started out weird and got weirder. Well, hey, I wanted to see it write fiction. My challenge to ChatGPT is to answer technical questions accurately and usefully.

It’s a chat program. So I started a chat:

Jeff: Will there ever be 128-bit CPUs?

I get asked this question a lot. My answer will be in the upcoming fourth edition of x64 Assembly Language Step By Step. It’s relatively simple, though there is some nuance.

ChatGPT’s first response:

It is possible that there will be 128-bit CPUs in the future, but it is not a certainty. The development of CPU architecture is driven by a variety of factors, including performance, power efficiency, and cost. Additionally, the need for increased computational power may be met through other means, such as the use of multiple cores or specialized hardware. As a result, the use of 128-bit CPUs will depend on the specific requirements of various applications and the availability of other solutions.

Hmmm. Saying this is sophomoric would be a slander on sophomores. But rather than reply to it immediately, I was given the option to re-generate the response. I clicked the button. ChatGPT thought (erk) for a few seconds and replied:

ChatGPT’s second response:

It is possible that there will be 128-bit CPUs in the future, but it is not clear what the need for such a large word size would be. Current 64-bit CPUs are more than capable of handling the vast majority of computing tasks and the industry has not yet identified a need for CPUs with larger word sizes. Additionally, the development of new technologies such as quantum computing may render traditional CPUs obsolete in the future.

This, in fact, is true, though it was still pretty much an obvious answer to a subtle question. So I clicked the “Regenerate” button again.

ChatGPT’s third response:

It is possible that there will eventually be 128-bit CPUs, but it is not currently a technology that is in widespread use or development. The current standard for most consumer CPUs is 64-bit, and there are also 32-bit CPUs still in use. The need for a 128-bit CPU would likely depend on advancements in technology and a growing demand for more processing power and memory capacity.

Sigh. Drink from the same well and you get the same water. I figured I needed a better question. I had one lined up and ready to go:

Jeff: Aren’t 128-bit registers useful for math?

128-bit registers can be useful for certain types of math, particularly for tasks such as floating-point operations and high-precision computations. Larger registers can increase the precision and accuracy of calculations, as well as allow for more efficient processing of large data sets. However, the need for 128-bit registers would likely depend on the specific application and whether the benefits of larger registers outweigh the cost of implementing them in terms of power consumption and hardware complexity.

I didn’t bother to regenerate this one. So far it seemed to be scraping Google, and while I mostly agreed with its answer, I would hope that Buzzfeed would require something more focused and in-depth. So I mulled it a little more and asked a third rather leading question. Read it carefully. There’s a lollapalooza in there:

Jeff: Wouldn’t the vector registers in the AVX-512 math subsystem make 128-bit registers unnecessary?

AVX-512 (Advanced Vector Extensions 512-bit) is an extension to the x86 instruction set that is designed to improve the performance of floating-point and vector operations. One of the key features of AVX-512 is the use of 512-bit wide vector registers, which can be used for a wide range of scientific, engineering, and financial workloads. The wide vector registers in AVX-512 can be seen as an alternative to 128-bit wide registers, as it allows to perform more operations in parallel with a single instruction, but it doesn’t necessarily make 128-bit registers unnecessary. For example, in certain types of workloads, it might be more beneficial to have a few large 128-bit registers rather than many smaller 512-bit registers. Ultimately, the choice of register size would depend on the specific requirements of the application and the trade-offs between performance, power consumption, and hardware complexity.

Bzzzzzzzzzt! Suddenly ChatGPT’s distant descent from poor ol’ Eliza became obvious. A few large 128-bit registers might be better than a bunch of smaller 512-bit registers. Right? Right?

I may test it more as time permits. After all, it’s free. But this will have to do for today. Among other things, I write humorous stories about AI. Looks like I just wrote another, without even trying.

Sudowrite would be proud.

More AI Text Generator Freakiness

Sep 22nd, 2022
by Jeff Duntemann.
7 comments

I tried it again. This time, I used a much more detailed prompt, which I’d written years ago (2014-ish) about a bottle that used to have a genie in it. The genie had been freed, leaving behind…a bottle. And what self-respecting genie would ever live in a non-magical bottle? So a nerdy guy buys a supposedly magical, genie-less bottle at a curio shop. The bottle, it turns out, has a trick: Ask it a question, and it answers.

Alas, the bottle has issues of its own.

I gave the prompt to Sudowrite, and let it follow its nose. Now, the way Sudowrite works is that when you ask it to write a section, it gives you two text blocks, both of which are based on the prompt. You choose the one you like better, and add it to the text that follows the prompt. Then you ask it for another section.

After a couple of go-rounds, I realized that in one of its generated text blocks, Sudowrite was putting together a sex scene. No sale. I chose the other block, which still had enough innuendo to make me uncomfortable. I sensed that in a sense it made a deranged sort of sense: I had described the protagonist as a lonely nerd. So! Toss him into bed with an imaginary girl who (the AI made graphically clear) had all the required female parts.

I stopped there. The first Sudowrite story I posted was in (somewhat) bad taste. I don’t make Obama jokes. Nor do I make Mossad jokes. I might make golem jokes, at least if the golem is the good guy. One reason I tried Sudowrite again today is that I wanted to see if bad taste was a habit or an outlier. It’s starting to sound like a habit.

Here’s the story. Everything up to the first rule is my prompt, taken verbatim from my notes file. I will someday use the concept (of a genie bottle without a genie) in a Stypek & Tuggur adventure, a prequel to Ten Gentle Opportunities. Everything after that is Sudowrite. Still a bit surreal–but if there’s a surreality slider somewhere in Sudowrite, I haven’t found it yet.

Djinn and Tonic

“What’s this?” Chuck Bialek asked the Gizmoids shop owner, and waved the weird, bulbous crystal bottle in the air over the counter. As best Chuck could tell, it was half-full of dirty water.

“Genie bottle,” said the old man. “But somebody let the genie go, so no wishes. Still, if you shake it and let it sit for a minute, it’ll tell your fortune. Used to be a hundred bucks. You can have it for fifty.”

Which meant it was probably worth a buck and a half, tops. Still, Chuck’s grandma had left him almost a million dollars, half of which was now in stocks. The rest was, well, for fun. He’d had a magic 8-ball when he was a kid. It was fun. This might be a reasonable facsimile.

Chuck laid a fifty on the counter, tucked the bottle in his canvas bag, and went back to his flat. After stuffing down a bratwurst and some Cheetos, he shook the bottle hard and set it on the kitchen table. Little by little, the dirt in the water settled out, leaving behind…words.

I miss my genie, read the words. Ha! He wondered how it worked.

“I wish I had a Jeannie to miss,” Chuck said. The nerd business was fun, but…lonely.

Chuck shook the bottle again, and waited.

Trade you a Jeannie for a genie, the dirt-words said.

“Deal,” replied Chuck.

 

His phone rang the next morning.

“This is Jeannie,” a voice said. “I’m with a Mr. Bialek, right?”

“Yep,” said Chuck.

“Trade?” asked Jeannie.

“What do you have in mind?”

“Wishbone,” said the genie.

“I wish you wouldn’t keep doing that,” said Chuck.

The Jeannie-thing giggled. “Make a wish,” she said.

Chuck thought for a bit. “I wish you were wearing a bikini.”

He hung up, and was on his way to the coffee shop when he heard a voice behind him.

“Trade?” asked the genie-Jeannie-thing.

“No!” shouted Chuck, and fled. He might be a nerd, but he was still the last straight guy in the world who hadn’t seen the Girls Gone Wild videos and had no intention of doing so.

He stumbled into the coffee house, and ordered a single-shot espresso with a splash of vanilla syrup. He had just opened a browser window and was trying to find a porn site that didn’t offer a free trial when he heard a voice behind him again.

“Trade?” asked the Jeannie-spirit.

“Are you nuts?” he said. “That’s the second time.”

“Make a wish,” said the genie.

“What does this do to your magic?” asked Chuck. “I mean, every time I make a wish, you get a day older.”

“Rhymes with night,” replied the genie. “So I’m immortal, just like I always wanted.”

“What happened?” asked Chuck. “I mean, most genies are pretty nice, but they had to have their masters’ best interests in mind.”

“My master was a fraternity boy. Couldn’t make a wish without prefacing it with ‘Yo, Genie…'”

“Oh,” said Chuck. “My apartment’s on the third floor. You can stay in the spare bedroom.”

“You’re a nerd,” said the genie. “We could never live together.”

“Right,” said Chuck. “So I need to make a wish for you to be you again. With unlimited magic.”

“I can do that,” said the genie.

“And I’ll need a million dollars. Are you listening?” asked Chuck.

“I don’t have to,” replied the genie. “From now on, I’m invisible to you.”

“This is going to get tiring,” Chuck told his computer screen. “I wish you’d figure out how to be a girl again.”

Will genie make the coffee? will genie make the toast? will genie make Christopher Hitchens read my manuscripts?

Will genie have sex, or merely masturbate?

Will genie remember to take out the trash?

Will genie notice her glasses are on her head?

Will genie give me a million dollars,

even if she does have the power?

Yes, genie will do all of these things.

Will genie kick my ass down the stairs

if I ask for more wishes?

 

Heh. Fersure.

If You Give an AI a Writing Prompt…

Sep 19th, 2022
by Jeff Duntemann.
8 comments

Ok. Fasten your seat belts. I am giving an online writing AI a test drive, using my free trial of a website called Sudowrite. I gave the AI a writing prompt from my crazy ideas file, a mere handful of words that I wrote down a very long time ago (I think in the ’90s) and never went back to. Here’s the writing prompt:

St. Louis Blues

“Take me to St. Louis,” said the cruise missle in the thick of the bramble patch.

Howie Hinkle blinked, and put down his machete. “How’s that?” he asked.

“Take me to St. Louis,” the missle repeated.

Howie howled. “Paw, dammit, there’s an old water tank in here and it’s talkin’ to me!”

It’s not a great deal to go on, but that was the idea. With a bare hint of a concept, what would Sudowrite do? I could have written some of it myself, or edited it, but I wanted to see what the damned thing could produce solo. 100% of the writing after the prompt was done by the AI. I had hoped to avoid the literary equivalent of Salvador Dali. Well, too bad. Hello Dali…

St. Louis Blues

“Take me to St. Louis,” said the cruise missle in the thick of the bramble patch.

Howie Hinkle blinked, and put down his machete. “How’s that?” he asked.

“Take me to St. Louis,” the missle repeated.

Howie howled. “Paw, dammit, there’s an old water tank in here and it’s talkin‘ to me!”

His father, Howard Hinkle Jr., took him for a look. The old tank was rusty and damaged, with a huge gash in it. Penetrating the tank from inside was a cruise missle, with a single eye and a fin on top. A single missile launcher poked through the gash.

“Who the hell are you?” Howie Jr. asked. “Why do you have a cruise missle?”

“My name is Shiloh,” said the voice. “I am a cruise missle, and I will ask you one more time to take me to St. Louis.”

“I don’t know where St. Louis is,” said Howie Jr.

“I will find it,” said Shiloh. “I am a golem of the Mossad. If you do not help me, I will call my brethren. We will clear the bramble from the road, and from there, we will find you.”

“Y-you can’t do that!” Howie Jr. said.

“If you do not take me to St. Louis, I will call my brethren, and we will take your woman, and we will bring her back to Tel Aviv. We will use her to breed a generation of human-missle golems. You can raise them with your own spector. They will be of your tribe, and they will spread across the land.”

Howard Hinkle Jr. swallowed, and wiped the sweat from his brow. “If I take you to St. Louis, will you let us alone?” he asked.

“I will,” said Shiloh.

“Deal,” said Howie Jr. “Let’s go.”

Hours later, a police car rolled up, and several officers got out. One pulled Howie Jr. aside.

“Which way did he go?” he asked.

“Er- he went that way,” Howie Jr. said, pointing.

“We appreciate your cooperation,” said the officer.

“Say,” the officer said, “was that your grandma that was in the hotel room with you?”

“Uhh, she’s my wife,” said Howie Jr.

“Doesn’t look like a 43 year old woman, honey,” said the officer. “What’s your name?”

“Howie, honey,” Howie Jr. said.

“Yeah, that’s what he said,” the officer said.

A week later, on the steps of the White House, President Obama announced the success of the St. Louis negotiations. Behind him, an Israeli flag hung. Beneath it, a Mossad agent clapped.

“What a complete dick,” said the agent. “He really thought he got one over on us, didn’t he?”

Shiloh’s eye blinked, and his fin wiggled.

Sudowrite, Egad

Sep 18th, 2022
by Jeff Duntemann.
No comments yet

I got wind of an online AI thingie called Sudowrite this morning. It’s a sort of writing engine that takes a piece of writing you’ve started and then runs with it. I’ve always been interested in AI, including AI creativity (“The Steel Sonnets,” “Sympathy on the Loss of One of Your Legs”) and used to play around with Eliza back when Eliza was young and hot. Well, here we are, in 2022. The supermarket is fresh out of Soylent Green, but by gully, Sudowrite is here, waiting for me to get stuck on something and yell for its help.

What I did was not quite that, exactly. For my first test I gave Sudowrite the initial few paragraphs of my weird fantasy short story, “Whale Meat.” Then I told it to go off and do its best to continue the story. First, the text from the current version of the story, which is published in my book Cold Hands and Other Stories. After the second horizontal rule, behold Sudowrite’s run with what I admit was a very peculiar football.

Sitting on a cold bus-bench at night-west I wished I had never been born a witch. I had just killed a man and Mara wanted whale meat. But I am on the canvas. I am not the painter. Creator forgive me. We had to stay somewhere. It would have been in the park with the ice if Lennie had not looked in my eyes in the public library and half-snerfed my pain.

There is no word for Lennie now. Years ago he would have been a hippie. Longer ago, a beatnik. Further back into time’s fog, a bohemian. Head. Weirdo. I don’t know what people call them now. Freak, perhaps. But he knew we were tired and homeless, and told us to stay with him.

His apartment was a hole. I was too exhausted to snerf him. He gave us hot coffee and a blanket and we slept on ragged foam rubber with the roaches. My dreams floated in hunger, tormented by Mara’s request for whale meat. She told me without words. I snerfed the picture in her head, a gray sea-mountain blowing steam from its nostril above moonlit swells.

I woke from restless dreams in a sweat of fear and anger. Some part of me never sleeps, and knew Mara was in danger. In yellow candlelight Lennie was bending over Mara, knife in his hand above her belly swollen with James. His eyes were slits and he chanted a language I never learned. I snerfed him. Some bad acid (is it still called acid?) was eating away at him and leaving only insanity in its wake. Snerfing him gave me a grim glimpse of the future: He would kill Mara and James in seconds. The knife began to circle for its last time, like a live thing in his shaking hands.

I had to time-opt. The universe ground to a slow halt around us while I dug up power from my heart-place, and zotted him. Lennie’s poor head shorted out like power lines blown away in a windstorm, with sparks and shrieks of inner agony.

His blood froze in his veins, his skin burned and turned black, his body dropped like a puppet with its strings cut.

I pulled Mara and James from the bed, and backed out of the stinking room, stumbling into the street, breathing deeply of the winter air. We hurried away past the sleeping city, three in a row. I held them in a circle of toil’d air. The sad lights of the city slid behind. We were two hundred miles away before Mara spoke.

I don’t mind you killing Lennie, I snerfed. He would have killed us.

I was just defending myself, I snerfed.

So were the people in the Warsaw Ghetto, Mara snerfed.

I walked faster. She didn’t snerf me.

The river was frozen solid. I cut off a section of the ice, and chopped it into chunks with the edge of my mind. I piled the chunks

Does this work? Dunno. I admit I was curious where Yonnie piled the chunks of ice, but it only gives you so many words at a time. Sudowrite half-understood my made-up word “snerf,” which is a witch-power, essentially telepathy. I forgive Sudowrite for not knowing that James was still in Mara’s uterus, where he’d been for five years. (Witches are pregnant for seven years, and James still has two years to go.)

Now, this isn’t the fairest of tests. I’m going to dig around in my notes files for an opener, ideally one I abandoned for some reason, and then go back and forth with Sudowrite to see if the one-and-a-half of us can actually finish a story that doesn’t read like a Salvador Dali painting looks.

I’ll let you how it goes.

Gatebox Waifu, and More of the Lotus Machine

Jan 22nd, 2017
by Jeff Duntemann.
11 comments

Somebody I follow on Twitter (don’t recall who) posted a link to a video about a new product out of Japan called Gatebox. It’s a little round 3-D video display roughly the size and shape of a coffee machine. An anime character lives in the display and has what seem like reasonable conversations with the user. It’s like Siri or Cortana on video, and it stirred some very old memories.

I’ve been thinking about AI since I was in college forty-odd years ago, and many of my earliest SF stories were about strong AI and what might come of it. Given how many stories I’ve written about it, some of you may be surprised that I put strong, human-class AI in the same class as aliens: not impossible, but extremely unlikely. The problems I have with aliens cook down to the Fermi paradox and the Drake equation. Basically, there may well be a single intelligent species (us) or there may be hundreds of millions. There are unlikely to be four, nine, seventeen, or eight hundred fifty four. If there were hundreds of millions, we’d likely have met them by now.

With AI, the problem is insufficient humility to admit that we have no idea how human intelligence works at the neuronal level, and hence can’t model it. If we can’t model it we can’t emulate it. Lots of people are doing good work in the field, especially IBM (with Watson) and IPSoft, which has an impressive AI called Amelia. (Watch the videos, and look past her so-so animation. Animation isn’t the issue here.) Scratchbuilt AIs like Amelia can do some impressive things. What I don’t think they can do is be considered even remotely human.

Why not? Human intelligence is scary. AI as we know it today isn’t nearly scary enough. You want scary? Let me show you another chunkette of The Lotus Machine, from later in the novel of AI that I began in 1983 and abandoned a few years later. Corum finds the Lotus Machine, and learns pretty quickly that pissing off virtual redheads is not a good idea, especially redheads whose hive minds ran at four gigahertz inside a quarter billion jiminies.

From The Lotus Machine by Jeff Duntemann (November 1983)

Corum tapped the silver samovar on his window credenza into a demitasse, and stared at the wall beyond the empty tridiac stage. So here’s where the interesting stuff starts. The crystal had been in the slot for several minutes, and the creature within had full control of the stage. Pouting? Frightened?

“Go in there and take a look around, Rags.”

“Roger,” Ragpicker replied, and a long pulse of infrared tickled the stage’s transducer.

At once, the air over the stage pulsed white and cleared. Life-size, the image of a woman floated over the stage, feet slack and toes pointed downward like the ascending Virgin. She was wrapped in pale blue gauze that hung from her hips and elbows in folds that billowed in a nonexistent wind. Her hair hung waist-long, fiery red in loose curls. One hand rested on one full hip. The other hand gripped the neck of a pitiful manikin the size of a child’s doll. The manikin, dressed in rags, was squirming and beating on the very white hand that was obviously tightening about its neck.

“He bit me, Corum. I don’t care for that.” The woman-image brought up her other hand and wrung the manikin’s neck. “We don’t need a go-between.” That said, she flung the limp figure violently in Corum’s direction. The manikin-image vanished as soon as it passed over the edge of the stage, but Corum ducked nonetheless. Corum stood, marveling. He took a sip from his demitasse, then hurled it through the image above the stage. The little cup shattered against the wall and fell in shards to the carpeting. A brown stain trickled toward the floor. The woman smiled. Not a twitch. “No thanks, Corum my love. Coffee darkens the skin.”

“I never gave the Lotus Machine a persona.”

The woman shrugged. “So I had to invent one. Call me Cassandra. Shall I predict your future?”

“Sure.”

“You will become one with me, and we will re-make the world in our image.”

Corum shivered. “No thanks.”

She laughed. “It wasn’t an invitation. It was a prophecy.”

← Earlier Posts