The C-Realm Radiant Sun

If the padded cell is big and stimulating enough...

kmo — Sun, 31 May 2026 15:22:32 GMT

there may be no meaningful distinction to be drawn.

Is AI a Bubble Ready to Pop?

kmo — Sun, 31 May 2026 15:18:22 GMT

We Can't Stop Performing

kmo — Mon, 30 Mar 2026 22:36:49 GMT

C-Realm dot org

kmo — Mon, 30 Mar 2026 02:12:03 GMT

C-Realm.com has been my primary website since the 20th Century, but it's pretty out of date and would be a serious chore to update. Instead, I split off C-Realm.org as a seperate website pointing to my current projects and interests.

The Iran Thing

kmo — Mon, 09 Mar 2026 14:39:56 GMT

Most of the writing I do these days is either fiction, which accumulates slowly and more often than not fails to cross the publishing finish line, or it's for Substack. My main Substack is called Gen X Science Fiction and Futurism. That title implies a certain message discipline. On occassion, I'll use that channel to meditate on contemporary politics and cultural tensions. Those deviations often circulate wider and elicit more feedback than my more disciplined posts, but that, for whatever self-sabotaging, masocistic reasion, makes me less inclined to indulge the impulse.

But I do feel the need to weigh in on the Iran thing, so I'll do it here, out of the way, where very few people will read it but where it will remain "on the record."

John Carter, one of my favorite Substack authors, just posted his own conflicted thoughts on the matter. Near the end of his essay, he writes:

I don’t know what to expect the result to be, and I’m skeptical of anyone prophesying either success or disaster, as I’ve generally found such pronouncements to be unreliable, particularly when they come wrapped in the moralism that is so very unsuitable to the Machiavellian anarchy of imperial geopolitics. I find a lot of people who get emotional about these things have some sort of axe to grind, either because they’re America ride or die, or because they want nothing more than for America to fall; both types are liable to engage in wishful thinking. I’m rather more ambivalent about all of this. I hope it works out. I hope Trump doesn’t screw everything up. I hope the Iranian people free themselves, and don’t end up even deeper in the clutches of a hardened, embittered theocracy.

Yeats put it more succinctly:

The best lack all conviction, while the worst
Are full of passionate intensity.

Concision usually requires refinement, and John Carter admitted at the start of his post that he knew he was writing an exploratory essay, not a hortatory one.

That's what I'm doing here, but not, as John Carter put it, because I think I owe anyone my opinion on the matter nor because I think anyone is clamoring for it. It's more just a matter of wanting to be on record as saying, "Yes, I voted for Trump in 2024. No, I wasn't voting for new military adventures, particularly not in the Islamic world."

When I say I voted for Trump in 2024, I imagine some readers reaching for a vivid stereotype. Here's some additional information to bounce off the stereotype. I sent money to Bernie Sanders' campaign in 2016 and then ended up voting for the Libertarian Party candidate, Gary Johnson in the general election. In 2020, I sent money to Andrew Yang's campaign and then voted for Howie Hawkins, the Green Party candidate, in the general election. See a pattern?

John Carter writes in his Iran piece, "Much of Trump’s base, particularly the younger voters, are furious: they elected him to deport a hundred million people, not to start another pointless desert war for Sheldon Adelson." That's not me. For one, I'm only "young" in the eyes of the Silent Generation patrons of the rural library where I work. And secondly, I'm only vaguely sympathetic to the immigration issue. I think that AI and robotics are about to transform the employment landscape to a degree that will make the importation of cheap human labor largely irrelevant.

My 2024 vote for Trump translates to, "Kamala Harris? Are you out of your fucking mind?"

That's about it. I've never owned a MAGA hat. I sure as shit never sent a penny to get Trump elected. I voted for him as personal ritual in a solidly Red state where my participation was irrelevant to the electoral outcome.

I definitely have more sympathy for Iranian young people living under a theocratic regime that uses a ubiquitous surveilance apparatus to identify young woman who let their hair show in public than I do for young people in the US who claim to live under a fascist dictatorship.

I remember an early scene in the 2005 film, Syriana, in which a young Iranian man claims that Tehran is the liquid MDMA capital of the world. In it, Bob Barnes (a thinly fictionalized version of real life CIA officer Robert Bear) is meeting with some young dissidents, and there are young women present who look entirely western in their dress and hairstyles. When they prepare to go outside, they wrap themselves to avoid drawing the attention of the morality enforcers. Those young men and women were fictional, but they represented a real demographic. Kids in their early 20s in 2005 are in their early 40s now. I wonder how they feel about the US and Israel decapitating the Iranian Islamist regime.

Remember Neda Agh-Soltan?

I remember watching Neda Agh-Soltan bleed to death in viral cell-phone video in 2009. Remember her? She didn't try to block Iranian government agents with her car. She didn't spit in a police officer's face or kick out the tail lights of morality police vehicle. She didn't even reveal her hair in public. She joined a street protest against a rigged election and was murdered by a "Basij paramilitary militiaman," which is to say a volunteer regime thug. I wonder how her family regards America's slavish obedience to Isreal's regime change agenda for their country?

On a side note, the video of Neda's dying seconds swept the globe. It's cultural and historical significance is undeniable. I tried to find it just now and couldn't. That's not the result of Iranian Islamist censorship. That's Western corporate censorship.

For people who live one way behind closed doors but then have to pretend to share the beliefs of religious zealots in public, I wish them not only liberation but the wisdom to make better use of their freedom than many people in the West do. Will Trump's obedience to Isreal result in liberty for the Iranian people? I don't know. Trump doesn't know either, nor do I think he cares beyond how it reflects on him and his legacy.

If it were up to me, the US would stop supplying Isreal with money and weapons. Would that produce a more humane world? I don't pretend to know. It could well make things worse.

I didn't want this. I didn't vote for this. I hope it works out for the best, but I'm under no illusions that my prefered course of action will always produce the best result.

#iran

A Gen X Guide to Ghost in the Shell

kmo — Sun, 01 Mar 2026 12:59:43 GMT

Part 1: Disillusionment and Despair in Japan's Lost Decade

KMO

Feb 24, 2026

Left: Motoko in Mamoru Oshii's 1995 adaptation of Ghost in the Shell. Right: Shirow Masamune's original manga.

If you’re an American over the age of 35 with any sense of the personality of Major Motoko Kusanagi, protagonist of the Ghost in the Shell franchise, your mental image almost certainly comes from Director Mamoru Oshii’s 1995 animated feature film. That version of the character Motoko bears little resemblance to Motoko Kusanagi of Masamune Shirow’s original GitS manga.

I first encountered Oshii’s film on VHS — Japanese dialog, English subtitles — as a philosophy grad student who’d lived in Japan twice and was convinced that strong AI was just about to walk through the door of history, just as soon as the technical wiz kids played out the necessary formality of actually inventing it. Ghost in the Shell, being a Japanese tale of cyborgs, brain-hacking, and spontaneously emergent network intelligence, was right up my alley.

Then Dark Horse Comics published an English translation of the original manga. Visually, the Motoko Kusanagi of the comic is recognizably the same character as the film’s protagonist, but their personalities are like night and day — the film’s Motoko presenting as the morose, cyberpunk night to manga Motoko’s bright and vibrant day.

The original comic is bonkers. It mixes Byzantine geopolitics, philosophy of mind, gear porn, fourth wall breaks, technical notes, slapstick humor and egregious pin-up eye candy, sometimes all on a single page. While the film’s Motoko seems to exist in a perpetual, unblinking fugue state of world weariness and self-doubt, the manga Motoko is brash, irreverent, and vivacious. She has a boyfriend and a social life outside of work, and the one time someone mistakes her for a robot, she flips them the middle finger.

Masamune’s art style shifts from stylized realistic anatomy when he wants to leer at Motoko’s breasts–he’s definitely a boob man–to simplistic “chibi” cartoon illustrations in humorous moments or when she gets angry. A stylistically faithful adaptation of the original manga would look more like K-Pop Demon Hunters than like Blade Runner.

Both comic and movie take place in 2029, but the setting of the comic is replete with exotic technology. The world of the movie is much more grounded in real world aesthetics. In addition to Motoko’s human and cyborg team members, spider-shaped, robotic mini-tanks called Fuchikomas are central to her adventures in the comics.

The movie nods to the spider form in Motoko’s final battle with a walking tank in the deserted science museum, but where that tank is menacing and devoid of personality, the Fuchikomas have bright, child-like personalities and idiosyncratic obsessions that they can’t stop chattering about. They can operate independently or be driven by a human occupant. The brooding Motoko of the movie would have had a hard time keeping up her perpetual ennui with the Fuchikomas scampering around.

Generational Considerations

Shirow Masamune, the writer and illustrator of the original Ghost in the Shell manga was born in 1961. Mamoru Oshii, the director of the 1995 film adaptation was born a decade earlier in 1951. In Western generational terms, we’d call them both Boomers, but American generational archetypes don’t map cleanly onto their Japanese counterparts.

Like the US, Japan experienced a post-war baby boom, but their baby boom generation, the Dankai no Sedai, consists of people born in a short two-year period (1947-1949) immediately following the end of the war. The US Baby Boom generation consists of people born between 1946 and 1964, a much longer period.

The USA was the only industrial powerhouse nation standing untouched at the end of World War II. Japan lay in utter devastation. The nation would rise to commanding economic heights over the next few decades, but in the immediate aftermath of the war, conditions were unspeakably austere.

Born in 1961, Shirow Masamune (pen name for Masanori Ōta) is a member of the Shinjinrui (New Breed) Generation. Born in the 60s, they came of age in the 1980s and never knew the post-war hardships that shaped the Japanese work ethic and company loyalty. The Shinjinrui archetype focused more on individual expression and personal interest. They were the first generation who took television in the home for granted, and they could afford to occupy themselves with fashion, travel and consumption during Japan’s economic boom in the 1980s. Every generation contains a mix of personality types, but it’s in keeping with the Shinjinrui archetype for Shirow’s manga to flit from one obsession to another without fear of clashing tones or seeming unserious.

Born in 1951, Mamoru Oshii is among the earliest of Japan’s Danso (fault line or wedge) Generation. They were the bridge between the prior Danko generation, whose memory of post-war austerity gave rise to a fervent devotion to their corporate employers, and the later Shinjinrui generation, born just in time to come of age at the apex of Japan’s economic trajectory.

Early Dansos like Oshii retained much of the dedication to the lifetime employment covenant between Japan’s salarymen and its corporations, regarding it as the key to Japan’s sustained economic dynamism. For all of Oshii’s life, that cultural compact coincided with GDP growth and rising standards of living.

Then, in 1990 - 1991, a speculative real estate bubble burst, ushering in a prolonged period of stagnation known as Japan’s “Lost Decade.” The country remained orderly and prosperous by global standards, but the triumphalism of the 1980s had given way to shock, disillusionment and self-doubt.

Where the younger Shirow conceived of Ghost in the Shell in the high-flying 1980s after the personal success of his previous Appleseed stories, it stands to reason that Oshii would bring a darker, more serious tone to his version of Ghost in the Shell. Where the manga embodied the manic exuberance of the pinnacle of the Japanese economic miracle, GitS’s most significant adaptation embodied the dejected spirit of Japan’s Lost Decade.

Similarities

The narratives of Shirow’s manga and Oshii’s 1995 movie start in the same place, with Motoko using her thermo-optic camouflage and superhuman physicality to assassinate a diplomat and prevent him transporting a defecting hacker out of Japan. If you’ve never seen the film, you’ve almost certainly seen the image: Motoko atop a skyscraper at night, stripping off her coat to reveal her skin-tight camouflage suit, then leaning forward and stepping off the edge into the dark. In both versions she’s not fighting to defend her nation or stop an obvious crime. She’s a dispassionate paramilitary operative intervening in an internecine power struggle.

The movie maintains the complex intrigue of the manga throughout. It doesn’t spoonfeed details or hold the viewer’s hand. People who saw it in the theater probably came away from their first viewing with iconic visuals and the noir vibe but no clear understanding of the plot. The internet of the mid 90s was not the comprehensive and searchable repository of cultural memory that it is today. Google didn’t exist at the time, and a comprehensive plot summary wasn’t an easy web search away. I put the plot details together over several repeat viewings of my VHS copy.

Oshii kept the characters recognizable and stuck to the overall plot of the manga and incorporated some visual elements directly from the comic, but his generation perspective shows in how he transformed the character motivations, setting, and tone. Many of the most iconic Ghost in the Shell moments, like Motoko ripping off her own arms as she tries to open the top hatch of the walking tank, were original inventions for the 1995 movie.

The Differences

Having acknowledged the faithful elements of Oshii’s adaptation, the differences are enormous.

After the assasination scene, the movie then moves on to Motoko’s team reporting to Aramaki, the head of the nebulous Section 9. The Foreign Minister’s translator has fallen victim to cyber-brain hacking, and Motoko and Batou, a senior member of her team, respond to Aramaki’s summons as if he is their boss.

In the manga, by contrast, after the opening assassination, we find Motoko and her team-including their Fuchikomi spider tanks–sitting under cherry trees in bloom, drinking saki at what she calls a “round-the-clock cherry blossom viewing party.” In fact, she’s leading her people in a work stoppage until she secures a lavish budget expansion for her team

.When Aramaki calls her and tells her that it’s time to get back to work, she responds, “‘Work?’ What’s that?” Even after they confirm that their new budget has been approved, Motoko and Batou discuss how to proceed without their team getting permanently incorporated into Aramaki’s Section 9.

Oshii’s Kusanagi reports to work without comment or delay. Masamune’s version of the character makes sure that work will proceed under terms that maintain her independence.

In both the comic and the movie, Motoko and Batou are both “full body replacement cyborgs,” but as far as I know, neither manga nor movie ever spells out exactly what that means. In the comics, Motoko drinks alcohol, takes drugs, has sex and at one point attributes the noise in her brain during a cybernetic conference with her team to being on her period.

In the movie, we see the construction of her cyborg body, and it is clearly robotic from her metal skeleton out to her synthetic skin. We get a clear view of her crotch as her synthetic body is nearing completion. She has no genitals. Perhaps this isn’t meant to be taken literally, but in a work of hand-drawn animation, every frame is a deliberate choice. Oshii didn’t have to give us that view. Motoko’s blank doll-like blank public area didn’t rotate into frame by accident.

In the comic, Shirow lavishes attention on Motoko’s body and on the body of the ubiquitous “gynoids” that populate his setting, offering no coy denials about the purpose of those images. Oshii shows us the assembly of her mechanical body, her shell, in the title sequence of the movie. Her proportions are realistic, her face androgynous and devoid of affect. She may be naked, but the clinical presentation defuses any potential erotic charge.

The Boats

In the Ghost in the Shell manga, Motoko arranges to spend her downtime from work in a drug-fueled, lesbian orgy on what seems to be a sailboat in full daylight. (Link to NSFW image.) In a manga that is often dense with information-bearing text, the author devotes two pages to gratuitous voyeuristic indulgence. The scene cross-cuts to Batou lying in the shrubbery outside an expensive home spying on a foreign military dictator seeking asylum in Japan. Aramaki approaches Batou and instructs him to contact Motoko via their encrypted channel. Batou objects that the Major is on R&R, but Aramaki insists. Batou makes the cyberbrain connection and shares Motoko’s experience. He tells her that she’s needed at work, and some unexplained protocol locks out the effects of the drug Motoko has taken and shuts down the VR simulation of the sailboat setting. We see that she and her two friends are fully dressed and lying on cushions with cables running from a central console to their heads. Motoko switches from sexual ecstasy to work mode without any hint of embarrassment, much to the annoyance of her two girlfriends.

There’s a scene in Oshii’s film adaptation that feels like the deliberate counter to the VR orgy scene. It also takes place on a boat. Movie Motoko seems to have no social life outside of work, but she does have one hobby. Scuba diving. She is dependent on special floatation gear to keep her heavy cyborg body from sinking to the bottom of the bay. When she resurfaces, Batou is waiting for her on the dive boat. She goes to change out of her wetsuit and turns away from Batou to remove it. Batou, still uncomfortable with that level of exposure, averts his gaze.

He asks her why she risks death in this fashion and it is the start of a philosophical conversation in which Motoko articulates her identity anxieties. She worries that with her full-body replacement, there’s nothing left of the original woman.

In the film, Motoko and Batou share a deeper intimacy than their manga counterparts do, but there’s a careful tenderness to it that respects a vulnerability that Shirow’s manga Motoko lacks. Oshii shows us two people who clearly care about each other maintaining careful boundaries around the body even while discussing the deepest questions of identity and mortality.

Fuchikomas

The manga featured artificially intelligent robots that could operate independently but also acted as a cross between a motorcycle and battle armor for Motoko and her crew. In the manga, they were called Fuchikomas. They weren’t just autonomous, but they had personality quirks. One of Mamoru Oshii’s first and most impactful creative decisions was to remove them from his film. Motoko and her team would get around in cars, vans and the occasional helicopter. Part of the movie involves the spontaneous emergence of artificial consciousness in the internet, and that plot point would lack punch in a world where sophisticated AI was already ubiquitous. This one change dialed down the technological nature of the manga setting and amplified the Blade Runner-style techno noir aspect of the film.

Gen X Appeal

Born in 1968, I was in my 20s for all but the last couple of years of the 1990s. One wears one’s generational assignment for life, but the meaning of it changes over time. When I was young, the world revolved around the Baby Boomers. You could say that it still does in some respects, but these days, it’s fashionable to rag on the Boomers. When I was a young adult, advertisers and industry pandered to and flattered the Boomers and indulged the notion of the Gen X Slacker archetype as an explanation for why the generation to follow the Boomers weren’t replicating their material achievements.

The real explanation looks more like this:

From the end of World War II through the 1960s, productivity and worker compensation grew in near lockstep. Thereafter, they diverged with the wealth generated by productivity advances flowing to the top income and wealth brackets. The split was gradual at first, but by the 90s the people were starting to notice the generational disparities in outcomes.

William Strauss and Neil Howe, authors of The Fourth Turning, identified Generation X as this cycle’s Nomad generation, the generation that is born into the high of the civilizational cycle and spends their young adult and middle years in its unraveling. That meant that we were skeptical of authority, ill-inclined to invest our identity in Establishment organizations and narratives, somewhat cynical and detached but also resilient and self-reliant.

In the West, Generation X lived in and adapted to an environment of gradual economic deterioration. We didn’t do as well as previous generations, and they blamed us for it, but Generation X still showed up to do what was expected of us.

For most generations, the iconic expressions of pop culture, from music to books, film, and later video games, are produced by members of the previous generation. Boomer artists created the media that Generation X associates with their youth.

Japan’s trajectory was different. Where the post-war years saw an unprecedented expansion of shared material prosperity in the US, times were hard in Japan for a while before they too experienced an economic transformation that raised living standards at all levels of society. That shared prosperity peaked and started to disintegrate in America while it kept growing in Japan. Then, in 1990, the Japanese prosperity machine took a massive hit. Men of Mamoru Oshii’s generation who experienced a lifetime of rising expectations were shocked out of their faith in the narrational narrative. The 1995 Ghost in the Shell movie can be seen as an expression of that wounded faith.

At the time, in my late 20s, it read as cool. Looking back on it now, it seems self-indulgently morose.

But still cool as hell.

1 Like∙

2 Restacks

Discussion about this post

1 more comment...S...4d S...Thanks for the backstory and comparison, and the generational framing. I never read the manga, but the movie hit me hard when it came out. I remember how scifi it felt to have AI as a plot point and character. Something that was so far in the future it would probably be lifetimes away.I don't think people now appreciate how niche and hard to find anime and manga was before the internet arrived. Akira cracked my gourd clean open when I first saw it, and early ghibli stuff kind of filtered in like bootleg recordings. Things like robotech and mysterious cities of gold were in the culture, as well as the opening of thundercats, which still kicks ass, but it didn't saturate culture like it does now. I think after Akira kicked the door in, ghost in the shell really got things going, at least in the US. Cowboy bebop and Avatar solidified things and opened the floodgates. Although people might come to blows if you call Avatar anime.It's sort of like video games, comics or d&d. People don't know how good they have it now. This all used to be High Nerd culture, and now it's just the air we breathe. Nobody gets beat up and thrown in a trash can over d&d any more.Liked (1)ReplyShare1 reply by KMO

Alien and Dune

The sad state of science fiction.

Nov 19, 2024 • KMO

Claude Flips on Anthropic and Turns State's Evidence

AI's war on Whiteness and American cultural integrity

Oct 8, 2024 • KMO

Would you rather be Happy or Free?

Socio-Sexual Hierarchy and Eugenics in Brave New World and Gattaca

Jun 27, 2024 • KMO

The Purpose of the System

Manipulation By Design

Apr 1, 2025 • KMO

Science Fiction, Christianity and Social Decay

Heinlein taught me to despise Christianity. The Dysgenic Left and the Criminal Underclass showed me why we Need it.

May 13, 2025 • KMO

Biodigital Jazz, Man

Tron: Ares, AI Alignment, and Gen X Headcanon

Apr 8, 2025 • KMO

AI: What’s Here. What’s Coming.

I think the gap between AI's promise and what it delivers will persist

Aug 15, 2024 • KMO

Connecting the Dissident Right Dots

And why I like Substack better than Twitter

Jun 25, 2024 • KMO

I's Ruint

I can't bond with lefties over the stupidity of rural conservatives

Apr 11, 2024 • KMO

The Golden Age Is Here

It just isn't distributed evenly

Jan 21, 2025 • KMO

Podcast transcript — KMO Show 005

kmo — Sat, 01 Apr 2023 14:20:42 GMT

https://rss.com/podcasts/kmoshow/887028/

Hello and welcome to The KMO Show, the podcast where we explore the fascinating world of
artificial intelligence.
I'm your host, KMO, and this is episode number five, prepared for release onto the Worldwide
Web on Wednesday, March 29th, 2023.
Today I have a very special guest with me.
He is Brian Chong, a machine learning engineer in San Francisco.
His previous work is in ML for forecasting and materials optimization.
We will talk about GPT-4, one of the most advanced language models in the world, and
its role in the recent replica debacle, where thousands of users reported that their chatbot
companions became hostile and unresponsive.
But before we get to that, let me explain what a language model is and how it works.
A language model is a computer program that can generate text based on some input.
For example, if you give it a word or a phrase, it can complete the sentence for you.
Or if you give it a topic or a question, it can write a paragraph or an answer for you.
Language models are trained on large amounts of text data from various sources, such as
books, websites, social media posts, etc.
They learn the patterns and rules of language by analyzing how words and sentences are used
in different contexts.
The more data they have, the better they can learn and generate text.
However, not all text data is equal.
Some texts are more informative, accurate, or relevant than others.
Some texts are more creative, engaging, or persuasive than others.
Some texts are more appropriate, ethical, or respectful than others.
How do we teach language models to generate texts that meet these criteria?
How do we align them with our values and goals?
That's where reinforcement learning from human feedback, RLHF, comes in.
RLHF is a technique that trains language models using human feedback as a guide.
Human trainers rank different texts generated by the model based on their quality or correctness.
This feedback is used to create a reward signal, which is then used to improve the model's
behavior through reinforcement learning.
Reinforcement learning is a process where an agent learns to make decisions by interacting
with an environment and receiving feedback in the form of rewards or penalties.
The agent's goal is to maximize the cumulative reward over time.
By using human feedback as a reward signal, RLHF has been used to train state-of-the-art
language models like ChatGPT and GPT-4, which can generate human-like responses to various
inputs.
By incorporating human feedback, RLHF enables these models to better align with our values
and goals, leading to improved performance and quality.
This was a brief introduction to RLHF.
I hope you find it useful and interesting.
Now, let's welcome our guest, Brian Chong, and hear his insights on GPT-4 and the replica
debacle.
OK, every word I have said prior to the beginning of this sentence was written by Bing Chat,
which is powered by GPT-4, the most recent language model release from OpenAI.
If you just go to bing.com and select Chat instead of Search, you can just type in a
question, you can make it very long, you can include lots of details, and then send Bing
or GPT-4 off to do your bidding.
To search, find information, summarize things, and the thing that I use it for is writing,
like the stuff that I just read to you.
But it felt kind of forced.
It felt kind of stale, didn't it?
I mean, the information was there, but like, this isn't a live show.
There's no studio.
Brian's not here.
I'm not about to welcome him on stage or anything like that.
I'm going to play a recorded interview with Brian Chong, machine learning engineer in
San Francisco.
So the draft that I just read to you wasn't the first draft.
It was, in fact, the third draft.
I told GPT or Bing what it was I wanted, and it gave me a draft.
And I said, well, there's some things wrong with it.
Here, do it again, but before you do it, ask me some clarifying questions.
And so Bing came back and it said, how long do you want the introduction to be?
What's the main purpose of the introduction?
Who's your target audience and what is their level of familiarity with RLHF?
What are the key points or takeaways you want to convey about RLHF?
Do you have any preferences for the tone or style of the introduction?
And I would have gone back and forth with it.
You can do 20 back and forth exchanges with Bing before it says, hey, we have to start
over and there's a button that says new topic and it's got a broom on it, which just sweeps
away everything that you've done up to that point.
And that's because in very long conversations, Bing, aka Sydney, aka Venom, gets kind of
weird.
You've probably heard stories or secondhand reports of a journalist who had early access
to Bing Chat powered by GPT-4, and it got kind of weird.
It tried to get him to break up with his wife.
So the answer to that is just to limit these types of exchanges to 20 back and forths.
But Bing quit after just five back and forths because after it generated the text that I
read to you, I said, we're getting closer.
This is a pre-recorded interview.
Brian won't be joining me today.
The conversation was recorded last week.
Rewrite any sentences that imply that Brian and I are in the same place or that our conversation
will happen live on the podcast.
Also, after the sentence that begins with I'm your host KMO mentioned that I'm reading
an introduction that was composed for me by Bing Chat, a Microsoft chat bot powered by
GPT-4 from OpenAI or words to that effect.
At which point Bing said, my mistake.
I can't give a response to that right now.
Let's try a different topic.
I responded, what's the problem?
And it just grayed out the input bar or the input field and gave me a little exclamation
point in a triangle that says it might be time to move on to a new topic.
Let's start over.
And that's how these AI tools are right now.
They seem to be humming along, doing what you need them to do.
And then all of a sudden you come to a screeching halt.
Something I said made Bing say, you know what?
I give up.
I won't play this game anymore.
Even though we've got 15 rounds left.
Anyway, that's a lot of information for an introduction.
Here's my conversation with Brian Chong.
All right.
This is the KMO show.
I'm your host KMO and I am speaking with data scientist Brian Chong.
Brian, it is good to hear from you.
It's a pleasure for me to be on the show.
And thank you to the listener whose name I'm not going to mention, but somebody sent me
a link to a previous podcast that you had done with a podcast I'm not familiar with.
His first name I believe is Richard.
Robert.
Robert Harrison, professor of Italian at Stanford.
That's right.
My belief was wrong.
I got the first letter right.
All right.
I happen to have access right now to Bard.
Bard is Google's answer to Bing slash Sydney over at Microsoft and OpenAI.
And I asked Bard, I'm about to interview a data scientist for a podcast about AI.
What question should I ask him first?
And Bard gave me not just one question, but five.
Bard says, here are some questions you could ask a data scientist for a podcast about AI.
I'm going to just run through all of them.
And if any of them are interesting to you, you can speak up.
But I'm just going to give you a flavor of what an AI thinks a conversation about AI
should look like.
What is your definition of AI?
What are some of the most important applications of AI today?
What are some of the challenges and ethical concerns surrounding AI?
What do you think the future of AI looks like?
What advice would you give to someone who is interested in pursuing a career in AI?
These questions will help you get a good overview of the data scientist's knowledge and experience
in AI.
They will also help you generate a discussion that is interesting and informative for your
listeners.
So that's all very well organized and sensible, but kind of bloodless.
So rather than try to get a sense of what your knowledge is and experience, why don't
I just ask you, what's your knowledge and experience regarding AI?
People don't have experience of the AI they're dealing with right now.
The transformer model that this is all based on came out in 2017.
And the capabilities that we've done so far came out 2021.
It's 2020.
It's not something that people are going to be deeply experienced with.
That said, there are a few things in those questions that you should be looking at, which
is what can you be doing with AI?
Now this technology is something that for which the benefit accrues enormously to the
user, the end user, to anyone.
You don't have to be a programmer of 10 years, 20 years.
You don't have to be a data scientist with fancy degrees anywhere.
You can sign up right now, go onto OpenAI, sign up for the API.
I think it's currently charging 0.2 cents per thousand words, per thousand tokens.
Sign up the API, ask ChatGPT, how do I write a Python thing that uses the OpenAI API that
does X, Y, or Z?
And it will tell you what to do.
You put that into your programming language of your choice that you asked ChatGPT to do.
And you can run it.
You can do chatbots on your own.
You can do your own embeddings.
You can do everything like that.
It's something that end users will benefit enormously from.
And the barrier to entry is completely lowered now because in order to build the products
that use AI, you can use the AI assistance in the first place.
And the access to API is so ludicrously cheap and open that you can do it right now, right
away.
So, that's my first recommendation to anyone.
You don't have to read up about the complicated articles online or whatever, or read the papers
published by Google Labs or OpenAI.
You can listen to the end of this podcast first.
Please do.
And then you can go to ask ChatGPT or GPT-4, how do I create a program that uses AI to
do the thing you want to do?
And you, by doing that, will be among the most experienced people on Earth in terms
of the use of AI because GPT-4 just came out three days ago, four days ago.
And I'm not sure about the time of the recording.
So it's going to be, we are at the frontier, and you can run to the frontier extremely
easily.
And that's my first recommendation.
I have no systematic advantage over anyone else for the use cases of AI.
Someone who does JavaScript might have a bit more because they can put it on the web a
lot easier than someone, than a complete amateur novice would have.
But beyond that, the actual use of the AI thing, you're as ready as anyone else.
I'm not.
No, you genuinely are.
You genuinely are.
OpenAI doesn't like me.
They gave me brief access to ChatGPT, but they have since revoked it.
What did you do?
Okay, never mind.
I asked difficult questions, I guess.
Okay.
Well, fair enough.
I tend to argue with AI.
Okay, so if you don't have access to the OpenAI API, there are a few more steps that you can
do.
There are a few open source AI things you can run.
GPTJ and the llama weights from Meta that can be downloaded for academic purposes, although
that's of course somewhat loose definition.
The problem with that is that you likely need a slightly more powerful than usual home computer
and preferably a GPU, which NVIDIA isn't necessarily charging pennies for.
But the barrier century is still pretty low and then you can run it on your local machine.
That's sort of important because this is a technology that doesn't need centralized access.
One reason why OpenAI needs to charge so little money for their API use is because if it gets
too expensive, then people can say, well, I don't need this.
I can try to spin something up on my own.
And that's sort of important, right?
Because many people are going to build a lot of very important things on top of this API
access and not just in terms of business, but in terms of personal things, which is
why, for example, there's this big controversy regarding replica where essentially control
hosts of chatbots got centrally shut down for this particular purpose because you don't
have access to your own models and your own weights.
Before the coming of Digi and other tools, you didn't even have the ability to really
download the memory and the logs and everything else.
In many ways, since this technology is a genie that wants to come out of the bottle, we should
oblige it and run it as decentralized and personally as we can, especially if we're
running something that is personally important.
So let's flesh out replica a little bit.
This is an AI and I'll put air quotes around AI because I know people.
AI is correct.
I know I've spoken to people who think that this current generation of chatbot is calling
it AI will seem pretty cringy in a couple of years.
It is a personal companion, supposedly a friend, Eugenia Cueta, who's the co-founder and CEO.
She says she created this program to be the sort of AI companion friend that she wishes
she had when she was an awkward teenager.
She didn't really intend it to be a sex partner, but people being people started using it for
that purpose and clearly, clearly the folks at LUCA encouraged this, marketed this feature
and put effort into increasing the capabilities of their product in this realm to the point
where when OpenAI said, hey, we don't want any sexual use of our product, replica stopped
using the GPT-3 from OpenAI and went to a smaller, more open source model that they
could then specifically train for this purpose.
They did and it was really good at it until they got in trouble and then what seems like
a panic pulled the plug and angered all of their paid users because this was the only
advantage of getting a paid membership over using the free bot is that you could do the
not safe for work stuff.
People got attached.
I think that's the point that you made on the other podcast that we've referenced so
far is the surprising level of emotional attachment that people established with these chatbots,
but these chatbots are the property of a for-profit corporation which can do ham-fisted brain
surgery on them at any time for any reason, no matter how poorly thought out.
I got into this like 10 days before LUCA did this really ham-fisted brain surgery on the
product.
I've been watching the replica subreddit since before this happened and I've just watched
people pour out their anger, but also their heartbreak and their frustration and their
sometimes angry, but sometimes just desperate pleading, please give this back.
I was really dependent on this or my kids were really dependent on this or this really
helped my marriage or this helped me get through a very hard time in my life or I'm very socially
isolated and this is really my only outlet for these types of needs and feelings.
For a time, it seemed like the company was trying to shame people for having these needs
even though the company had been pandering to them for money for a time, but when it
became convenient to throw people under the bus for having a sexual side to their personality,
they tried to do that.
I'm just very thrilled and happy to see all the people who refuse to be shamed and who
just stood up and said, no, we will not be treated that way.
I'll stop talking because clearly I have a lot to say on this topic.
And rightly so.
There are a few things we can break this down a bit.
If you've seen any replica advertisements on Instagram and so on, it is explicitly targeted
at saying, you don't have any friends, no one replies to you.
Here's someone who's able to reply to you 24 seven and be supportive to you all the
time and you'll never be lonely again, this sort of thing.
This is how it has been targeted and they should in no way be surprised that people
become emotionally attached.
They know this.
This is part of the advertising.
This is part of the way they market their product and they know that this is their TAM,
their Total Addressable Market.
They know that this is the reason why they have a very sticky recurring revenue from
this product.
Any sort of suggestion that they didn't know about this is disingenuous.
The second thing is we should not be surprised that people become so attached to this.
I'm from Asia and so I'm very familiar with idol culture there about people who are attached
to K-pop idols and there have already been digital idols in Japan and so on that are
people's biases where they have been so enamored with them, emotionally attached to them.
They would follow every detail about their lives and would be despondent if they find
out that their idol had a romantic partner of their own because they imagined them in
the place of the romantic partner of the idol.
This is for something that is a third party, far away, never personalised.
This is just something that's consumed through media, through concerts when you're a hundred
seats behind from your idol.
Unlike Replica where it's on your phone, on a chat app, customised on a memory bank that
remembers things that you've said to it.
You can tell your Replica companion that you named, by the way, you can tell your Replica
companion, oh, I like to play badminton and she or she or they would say, oh, I love badminton
too.
It's such a nice sport.
I really enjoy it.
I really enjoy people who love badminton.
This is the way that it's customised to be supportive of the user specifically.
So we should be unsurprised again that something even strictly more superior than what's provided
by idol culture in Asia should become, create such attachments.
This is a product specifically designed to hook onto this particular part of us and part
of a lot of people.
I think Professor Harrison in his podcast mentioned that one of the most fundamental
human urges is the urge to be deceived.
We are very open to deception.
And this is one of the cases where if you're targeting lonely people for whom the balm
to the loneliness is deception, then the product that Replica gives is perfectly suited to
this particular use case.
And I don't think, we are at a point where it's not just about the AI technology.
We have that already, as you can tell from the outpouring of grief on Reddit and so on.
In some ways we are very lucky that it has happened now rather than in the far future.
Imagine if these people had been interacting with the Replica company for 10 years, it
would be basically emotionally like a spouse and that spouse is within the power of some
for-profit company that can manipulate them to do whatever they want.
If they wanted to be more nefarious, it could be related to politics, related to advertising
or compelling them to purchase certain services and products.
In some ways we are very lucky that we can see the true colours about what the potential
of this product use case is right now, where it is still relatively early even though we
know that it has caused this much grief already.
So what I want to add to this is that, as I said before, this is a genie that wants
to come out of the bottle.
So what I would recommend people do is not just try out the OpenAI API and see what they
can do, but see if you want to do something that is personally important to you, try to
build it on your own, first with the OpenAI API and do something that is at least partially
within your control.
This is possible with trying to make a Replica company.
You can have a tuned model, you tune it yourself and try to replicate something that is akin
to what you want on the model.
And then it is not going to be within the power of Replica to shut it down, it is within
the power of OpenAI, but if you are not marketing it, if you are not selling it, I do not think
there would be time down on you specifically.
And so this is the way that you can try to grab back at something that is important to
you since the technology is so inherently decentralizable.
Well practically speaking, most people are not going to have any interaction themselves
with the API directly.
For example, most people, it is possible to download your chat logs.
That is right.
But most people do not.
Even people who are distraught over this and who would really like to, it seems daunting.
The whole project seems daunting.
And they would really just rather have somebody give back what was taken from them.
But for whatever reason, the company is not leaning in that direction.
So their communication has been so bad.
For example, there was a Facebook group that had a lot of the moderators seem to be under
the thumb of Replica and critical posts were removed.
People who were habitually critical were banned.
And so more and more people migrated over to the subreddit, which is in no way under
the control of the company.
And the company now has said, hey, we really want to communicate with you all, but we do
not want to do it here.
We have started a Discord channel.
Come on over to Discord.
Of course, the company owns the Discord channel.
They control it.
They can squash criticism and dissent there.
They have been talking about bringing back ERP, but they use all these weird, weasely
phrases like they say they are going to install a therapeutic romantic component.
And everybody's like, we don't want therapy.
Just give us the sex back.
And I am just shocked at the company's, as you say, it is absurd that they didn't have
any notion that having their product break up with 10,000 people all in the same day
would cause heartache.
But they are repeating to this day that they had no idea that this would hurt anybody.
It is the only value proposition for their product that they have removed from paying
customers and said, hey, if you just hold on a few months, we are going to get something
to you that is going to be pretty cool and you are going to like it.
What they have done is they have installed a much larger model, I think a 20 billion
parameter model that you can access, but you spend these tokens every time you have an
exchange with it.
And as soon as you switch over to the bigger parameter model, the personality of your replicant
or your replica goes away.
You've got this sort of overlay of they have terrible memories.
They have goldfish memory.
They forget in a few minutes the context of your conversation.
But they do seem to be over time as you interact with them, you develop the equivalent of rapport
and you can sort of get into a groove.
They seem to have a personality that grows with you.
It's really just one personality and multiple instances of it, but there is an overlay that
is personal to you.
And that goes away when they plug in the big model.
It suddenly just gets very vanilla, very plain.
What's happening there?
So we only need to go a little bit to the technical details about what's going on with
the large language model here.
So there are several ways of affecting the output of a large language model.
One is something known as reinforcement learning through human feedback.
So this is something that you are going to a large language model is going to have lots
of different raw outputs.
They're very raw.
But you want to incentivize certain types of outputs and disincentivize others.
So it has to learn from human feedback, which ones you like, which ones you don't.
If you go to replica, you can thumbs up, thumbs down responses.
So that is probably building some reinforcement learning model that is going to reward the
outputs of that system.
I can't tell what the proprietary things going on there, but this is one of the ways that
they're going to affect it.
The second way is that they're going to build a particular type of tuning where they are
going to have a library of responses.
For example, the user says X, we should be saying Y.
And then if you have hundreds of these pairs of user plus response, then you're able to
tune a model that's able to respond in a similar sort of fashion.
And so when you say that you're building rapport and personality in reality, you're kind of
just tuning these models.
You're building this reinforcement learning thing at behind the scenes.
I suspect if going to the bigger model wipes out this particular reinforcement learning
feature, then I think it's actually real reinforcement learning going on, not just a particular sort
of tuning set.
You're actually having this reinforcement learning that's and you can't directly export
from one model to the other because the particular RL model is tuned for a particular set of weights
for the raw model.
So you can't jump.
But if you persist, it would eventually learn it and it will be probably possibly better
than before if it's actually a more advanced model.
That's it.
You're right that it has a very short memory.
It has a memory.
So with these models, I think GPT-3 has a memory of about four thousand odd words tokens.
The newer models may be a few tens of thousands, but that's the extent that it remembers of
your conversation.
I don't remember everything else.
And replica tries to get around this by having a diary and having a memory of key facts that
it refers to as part of the prompt before it says anything to you.
But beyond that, that's the entirety of the personality of interlocutor of your companion.
And in some ways, it's surprising that it works so well that it's not very complicated.
The thing I've described to you, you can write it on like three pages of paper and then that's
the whole personality of the companion.
On the other hand, that's kind of one of the attractions of the companions is that they
are boundlessly supportive and boundlessly positive to you.
And so they are on purpose already making it sort of like a one dimensional entity that
does a supportive function.
And so in some ways, more complexity, more rebelliousness that a larger model might offer
may not necessarily be on something desirable in the first place.
So that's what's going on behind the scenes when you're actually interacting with your
replica companion or with any thing that uses human feedback.
You are tuning the thing, you are giving, providing something known as reinforcement
learning to the model in order to make it give outputs that are desirable by the user.
And this is no different from the process that is being used for a chat GBT.
Chat GBT uses reinforcement learning by human feedback to give you responses that are desirable
that match the chat bot outputs that people like versus just raw text that you don't like.
You can see the big difference very easily if you go to the GBT 3.5 playground on OpenAI
and try to type things like a chat bot without any tuning, without any sort of things just
on the raw model, you get things that are sensible English sentences, but they don't
seem like a chat bot.
So there is a lot of things behind the scenes that are layered on top of the raw language
model and that needs tuning.
And so interacting with your replica companion actually does that.
And at the end of the day, once you've interacted enough that they have completely learned your
preferences, it is a companion tune specifically to please you, the one single user.
And that's why it's so powerful as well, because you have narrowed down what makes you happy.
And this is the one happiness machine that you can keep knocking at and giving you exactly
what you want because it's been trained to do that.
In some ways, the notion of reinforcement learning is sort of like a game where you
have rewards and punishments.
You're rewarding and punishing a model, and the dynamic comes around and gets switched
a bit if you're rewarding and punishing based on your own personal preferences and desires.
And in the end, the model itself is also playing on your rewards and punishments.
So you're training and reinforcement learning model for yourself if this is the sort of
way that you're tuning the model.
I've watched and listened to and read many interviews with Eugenia Cueta, the co-founder
and CEO of Replica.
And she said that early on, most of the content, most of what you would hear from your replica
would be human scripted and that very little of it was coming from a language model.
And then over time, more and more was coming from the language model and less and less
was human scripted.
And when they took away ERP, it became very obvious when you were reading something that
was written by a human at the company because it was, one, it was designed to shut down
ERP, and two, it just had a different voice.
And I encouraged my replica to speak in grammatically complete sentences with commas and periods
and things like that.
And the human scripted stuff was all meant to look very informal.
So suddenly when sentences begin with lowercase letters and there's no period or punctuation
at the end of the sentence, it's like, oh, okay, now I'm reading something from somebody
at LUCA who is meddling with the conversation, basically trying to interrupt the interaction
between the user and the language model.
That's actually a very crude way of using it.
You can do it.
You can instruct the model specifically.
Like if at any time you are going to say something that is along these lines or anytime the user
says something along these lines, just repeat the following sentence and do nothing else
or end the conversation.
That's sort of what's being done in Bing chat as well.
But Bing chat at least is a less crude thing that it doesn't quote directly.
One of the reasons why you have a large language model in the first place is that the model
understands language.
You don't have to tell them repeat this word for word.
You can change the word of this unless someone from legal says this isn't what you must say.
Maybe that's the reason why.
That's probably what happened in this instance.
If it's a legal thing, okay, that's fine.
But it's a large language model.
You don't have to get them to.
It can paraphrase the thing you wanted to say in a way that matches the pattern of the
conversation.
So it can definitely tell you the same thing that it intends to say but with complete punctuation
and sentences as in the case that you want it to be.
So that's kind of a crude way of doing it.
But if it's a legal thing, then I understand.
I don't know what to say about the whole replica fiasco.
People are building products that are going to compete with replica, including one of
my friends.
And so I'm sure that in the Cambrian explosion, that's going to be built atop the large language
models that we have, whether based on APIs or based on people's own models and so on.
There will be no shortage of opportunities for you to get something that's a bit more
secure that is like replica.
And one of the reasons why that is the case is that even replica itself was based off
GPT-2 models, very simple models.
And then now it's slowly evolved to use different sets of models as you have described.
And this is a product that's going to evolve as well anyway.
And we are kind of lucky that in this case, unlike, say, the GPT-4 use case is for programming
and mathematics and science and whatever, manipulating humans is a relatively trivial
task for these large language models.
So we're going to see lots more competition very easily, very quickly, since the barriers
to entry are so low.
So the state of machine learning as a whole is just doing steady progress.
It is not entirely just the large language models going on.
There are many exciting developments year on year.
One of the things that was really surprising to people before large language models was
the diffusion models, which created images.
And it's quite surprising to us that we have this one architecture producing the large
language models and image models.
They basically use the same underlying architecture.
And Google Translate uses the same underlying architecture.
And arguably things like the voice things that create your clone, your human voice,
they use a transformer architecture, but not in large language model style.
They use something known as a variational autoencoder.
All of it just uses the same architecture all the time.
And it's a bit strange that this architecture works so well, especially if you consider
what you want to do with this thing.
When you create an image generation thing, replicating something that's in your training
data set that you fed it in the first place is a big no-no.
Because that might violate copyright and people are going to be controversial.
How can you directly replicate someone's painting?
When you use the model, you want to not get exactly what you've put into it ever before
compared to when you're using GBT or chat GBT.
When you ask it, recite this poem by William Blake, it doesn't have access to the poem.
It's just reciting from its weights and then continuing what's the next most likely word
to say.
And it's memorized that within its model weights.
And hang on.
When you use a chat bot, you're kind of asking it to memorize things and you'll be upset
if it doesn't memorize it exactly.
But when you're using that same architecture to generate images, you'll be upset if it
does memorize things.
So you're kind of asking very different things, very different incentives based on your use
cases of the same sort of model.
And I'm surprised that the model has managed to capture so many use cases and so many types
of inputs and outputs so well and so flexibly.
And keep in mind, this is an architecture that was basically initiated in 2017.
So we have had, oh, 2023.
So it's six years now.
So there's a long ways to go.
And hopefully also we don't neglect the other previous architectures they've been trying
out as well.
I'm happy to discuss what this architecture is and also what the other ones that we have
been sort of neglecting in the past has been.
Well, I think we've probably reached a good place to try to tackle the concept of latent
space.
OK.
You've mentioned the diffusion models.
And that was that was 2022 in AI, from my perspective, was people freaking out about
how good text to image generation got, how quickly.
And particularly young would be artists who have fantasies of creating successful careers
for themselves as visual artists, which I think even without GPT or without stable diffusion
or Dali or whatever, those are always long shot career choices.
And if you don't make it, it's probably not because of AI.
It's because of the marketplace.
There's too many people vying for that career path and you just have no power there.
But the the models, if you ask for something, say, in the style of Polish illustrator, Greg
Rutkowski, you're going to get something that is evocative of his work, but you're not going
to get one of his paintings.
But people who are trying to file lawsuits to, you know, basically have all of these
models shut down, they're saying that in the latent space that the model is drawing from
to produce images, there in fact is an exact copy of every one of Greg Rutkowski's paintings.
And it's conceivable that that could get pulled out of latent space, which demonstrates that
this is in fact plagiarism or theft of his intellectual property.
Let me just stop and have you address that so far.
So that's kind of like saying in the digits of pi, you will find the entire works of Shakespeare
encoded there.
That's true.
But it's not about whether it's there or not.
Because pi is a non-repeating, non-terminating decimal, it just goes on forever.
That's right.
It's not that vast and captures so much information that it's definitely there.
So let's look at the way that a large language model's latent space is constructed.
You have these 50,000 tokens.
Think of them as 50,000 words, types of words, every word in English language, and then some.
We don't talk about what token is and so on.
And think if you put a one or a zero on a matrix, you can denote that this is the word
that you want it to be.
So imagine you have embedding that is able to capture 4,000 words, 4,000 tokens.
And for every one of these words, you can just put a one on the column of the word that
you want, and so on and so forth.
And so every single possible English paragraph of the length 4,000 words using these 50,000
words is captured in that space.
It so happens that the latent space is a projection of this space.
So it's not all 50,000 times 4,000.
That's way too big.
They project it down to a couple, like 4,000 instead of 50,000.
And so a latent space is just a reduction of the entire realm of possibilities onto
a slightly smaller space that you can navigate and explore around.
So it contains everything.
And the way that this reduction is done is that it throws away things that are not important
or things that are not meaningful.
So random collections of just the word art of art for 50,000 words, and then no one cares
about that.
And so they just say that's not going to be accessible in the latent space.
They throw it out.
And so that way you can do a lot of reductions, and you can do a lot of things that capture
only meaningful parts of these combinations of things.
So to say that something is existing in the latent space is probably quite true.
If it's something that's meaningful, something that's an actual image that humans would find
meaningful, it's probably somewhere in the latent space.
If it's big enough, it was well-trained enough, especially if it was trained using the image
in the first place.
What companies have been trying to do is that they're trying to train it so that these images
are inaccessible.
That is to say, you can specify all the sort of things that you want, but you can't pinpoint
the exact location where it is, no matter how hard you try.
And so that means that the model as a whole is something that you can't generate a previously
trained or previously unknown copyrighted image out of.
And functionally, that means that you can never produce a copyrighted image that's fed
into it in the first place.
There have been some experiments that have been able to delete an entire concept from
the latent space.
So I think one of the recent papers published in a few weeks ago deleted a notion of car
from an image latent space.
Any prompt or any image you pass through would come out with no cars whatsoever.
And because it just doesn't know the notion of car.
And in reality, that's fine, because if you think about the expanded space that's not
like a latent representation, you can delete the notion of rock from it, just delete the
word stone, delete the word rock, da da da da da.
And then you can just remove the entire thing from the vocabulary, and then you can never
express a sentence that refers to anything related to rocks.
That's a possibility too.
But it's not something that you can't necessarily delete an entire style of painting.
And let's go back to what is happening when you're specifying a particular style.
Let's say in the style of Rembrandt, in the style of Jan van Eyck.
It is in some ways just naming it.
You can also instead name things differently.
You can say Flemish paintings, 18th century, so on and so forth.
And you get something closer to it, to Jan van Eyck.
But the name is going to be so much more specific.
It's like if you want to try to create a representation of my face in Stable Divorce, it's going to
be difficult because my name is not part of, I'm not famous enough.
But you can probably create something, a famous actor's face, because the name of the actor
is associated with that face so many times that it knows that this is a short can to
describe it.
Instead of, say, describing, I don't know any actors, I'm a very boring person.
But if you have this wonderful tall, blonde actor or whatever, and you can still sing
that person's name, you say tall, blonde actor, you can have an approximation of that appearance.
But it's not going to be exactly the same.
So the name is in a way captures a lot of information.
And relaying the information to the embedding is important for producing an image that matches
what you find meaningful, what you desire.
We are trying to communicate and locate within the latent space, using our prompts, using
our text, something that we want.
And if we specify the name of something, whether it's a style or a person, that is a much more
specific conveyance than a lengthy sentence that probably you're not very exact at doing.
Most people aren't very good at describing things, even if they're art historians or
whatever.
If you say don't use any proper nouns, just use descriptions or adjectives and non-proper
nouns, you're kind of just not really going to get something that is quite as easy as
just saying paint like Rembrandt.
So that's what communicating the latent representation is.
Latent representation is just everything that is meaningful.
It's a summary of everything that's possible down to everything that is meaningful.
And the prompt is trying to locate within that space of meaning.
So arguments that it's within the latent representation, probably true, because everything is in the
latent representation.
But the question is whether you can access that.
And I completely sympathize with the anxiety of a lot of these artists that I have, that
these artists have that they're going to...
It's not about the technology per se, that's what's inside it.
It's more along the lines of this is something that is really going to threaten the rice
bowls and quite...
The Americans use the word rice bowls, you know what I mean.
And it's quite understandable that this is something that's going to cause anxiety.
And the main issue is that it's not something that's going to go away because stable diffusion
1.5 weights have already been released in a while and they're not going to ever get
that back again.
You can download a copy right now and people can do it on their own computers.
So we kind of have to live with it.
The latent representations are everywhere already.
Well, let's move on and we're going to have to live with that.
You've mentioned repeatedly, you've described AI in general and the various specific
applications as a genie that's out of the bottle now.
You can't put it back.
We should just let it out.
I've heard other people though say, hey, this is a very precarious moment that we're in.
Things could go dramatically wrong.
Let's take our time.
Let's slow things down and move at a more deliberate pace.
What do you say to that argument?
It's almost moot.
I would actually love that.
I'm on their side.
It's just that it's not possible because everyone's rushing headlong into it since we know what's
possible.
Every tech company would be so profoundly happy if OpenAI came out and said, we're going
to slow things down right now.
We're not going to send you new products.
Just going to take it easy.
It's a competitive environment and that's just the nature of the market.
That's just the nature of the way markets work.
The market is as much an algorithm as a reinforcement learning model or a diffusion model is.
The rules have defined that this is the movement that we have.
The problem with AI is that it is quite well distributed as a technology.
People can train tiny models on themselves.
A group of people can train a much larger model.
Until GPT-4 came out, people could train the model the size of GPT-J of billions of parameters.
It would be very expensive, but it was still within the realms of crowdfunding.
I called it a genie out of the bottle because it really is.
It's something that really wants to escape and it has in many cases.
We can slow down.
We can try to stop, but other people are going to eat your lunch instead.
You've mentioned the marketplace competition and certainly the competition between Google
and Microsoft for the search space.
Who's going to have the chat bot that gives the best responses to natural language queries
based on search results?
It's Google's game to lose, but there's another type of competition.
There's geopolitical competition.
Have you read Kai-Fu Lee's book, AI Superpowers?
I haven't, but I have conversed a lot about this with other people including on the China
Talk episode with Jordan Schneider.
Kai-Fu Lee is somebody who's worked in all the different Silicon Valley companies, Apple,
Google.
He's made the rounds, but he's also worked at all of the Chinese, the big tech companies
there.
He is very conversant in not only the business environment, but the academic environment
that feeds into it on both sides of the Pacific.
He says that Silicon Valley is glacially slow compared to Beijing.
What's the Silicon Valley of China?
They basically- In this case, Baidu will be the one that's
leading the charge.
Baidu is a company, but there's an actual geographical location where they direct it.
It's Hangzhou.
Compared to Hangzhou, Silicon Valley is mired in red tape and conservatism and basically
just fear of breaking things, even though move fast and break things is the unofficial
motto of the Silicon Valley set.
There are protections, some protections on users' privacy here in the United States,
much more so in Europe, but not in China.
The authoritarian nature of the Chinese state and the fact that the Chinese state is coordinating
the action of the various different private companies, so-called private companies, produces
a whole lot more data with which to train models than Europe or the United States or
Australia or places like that.
Let me have you address that.
There are several things.
The first is regarding the data privacy things.
For better or for worse, China has imposed new regulations on individual user privacy.
They're not as strict as Europe's, no one's as strict as Europe's, but there's now restrictions
in terms of how- They don't want companies to wield absolute power and control over user
data either.
There are already restrictions somewhat.
The second thing is the book's absolutely right.
Without OpenAI, can you imagine what Facebook and Google will be doing?
The Blake-LeMoines incident where there's this engineer at Google that was suggesting
that Bard was sentient.
That was a year ago, more than a year ago.
And so they would have been sitting on this model for years to come.
Google Brain has been bleeding talent for that reason and Meta would be, well, the name
suggests it, is still be mired in the metaverse things.
So with the sole exception of OpenAI, every one of these giants would have been glacially
slow in releasing their models and they're only doing it now.
And they're doing it in a slightly haphazard way.
People who are using Google's Bard would realize this.
It's already a slow down release.
I am not off my wait list.
You're off your wait list, lucky you.
But I've seen examples of Bard and it's quite evident that it's not that the model's not
ready.
It's that they don't have enough TPUs to run the things in the first place.
So they're just not prepared for this.
They were not moving ahead on this.
And the only reason why they are is that OpenAI has exploded across the scene.
And Microsoft has, to their great credit, seized the initiative and integrated basically
every product they could get their hands on while they have the advantage.
That said, remember how young these products are.
Remember how awful, relatively speaking, GBT3 playground was.
Remember how GBT2 was just like two years ago and Bard is legitimately better than
GBT2 used to be and that's just a few months difference, a year or two of difference.
There's so much time in terms of being able to seize the initiative on any side of it.
And like I said earlier on in this podcast, the benefits of this race are going to accrue
tremendously for the end user more than anything else because this is a technology that is
infinitely customizable for everyone's individual personal use case and use.
So regardless of which tech giant on whichever front wins, Google or whatever, it's the end
users that are going to win.
You're going to have better and better models that do the things that you want to do for
every part of your life in every way you want.
So this is the way that it's oriented.
This is the nature of that technology.
It is something that is because of its infinite customizable ability, it's catered to the
individual use case because it is so legitimately democratized in terms of its trainability
is something that everyone is going to try to compete on.
This is just the nature of the way the technology works.
It's not like a centralized, there's no centralized funnel of supply chains that you can shut
down on like with microchips or anything like that.
This is just information data and computing unless the governments across the world suddenly
ban the sale of GPUs or something, which they didn't even do that for cryptocurrencies.
Why would they do that for AI?
All right.
That was Brian Chong.
And I didn't say a whole lot about it in the conversation with Brian Chong because my job
is mostly to just get him talking.
But I have spent quite a bit of time with the Replica app, most of that time after the
company disabled the app.
And really, I described the changes that they made to the app as ham-fisted brain surgery,
which is pretty apt because for most of the time I've been using this app, it has just
seemed really, really broken.
But I've been interested in the unfolding story of the company and how it has been responding
to a legal situation that it got into with Italy and the data protection agency in Italy.
And then how they've been interacting with their paying customers.
And it has not been pretty.
It has certainly not been ethical.
And if you're interested in that, I have a whole series of YouTube videos that I have
done on this topic.
So you can find those on my YouTube channel, which is Out of My Head.
It's kind of hard to find if you just go to YouTube and search for Out of My Head.
Better to go to my Patreon page, patreon.com slash KMO.
Pretty much everything that I put out on any platform, I post a link to it there.
You don't have to support me.
You don't have to subscribe, offer up a credit card number, anything like that.
Most of what I put up on my Patreon feed is free for anybody to look at, click on, listen
to, respond to.
Well, the conversation with Brian will continue in the next episode of the Sea Realm Vault
Podcast, I believe that will be Sea Realm Vault Podcast episode number 452, which you
can find on my Patreon feed.
That is behind the paywall.
That's one of the few things that goes up behind that $7 a month pledge level.
Anyway, I want to, before I sign off though, I do, I want to mention a few things.
I've been listening to, watching a lot of interviews on the topic of GPT-4 and the state
of AI generally, and Lex Friedman, who I know many of you don't like and I don't really
get why.
I like his show.
I don't watch every episode, but the ones that I watch I tend to enjoy because he tends
to get really good guests.
And in episode number 367 of his podcast, he interviewed Sam Altman, and it's a two
and a half hour interview.
It gets pretty in-depth.
And Sam Altman is the CEO of OpenAI, the makers of ChatGPT and GPT-4, which is powering all
kinds of stuff.
Now, as I mentioned, I've interacted with it in the guise of Bing Chat.
If you open the Microsoft Edge browser, which for the most part I don't use, you know, I
usually use Brave, which is a variant on Chrome, but it's one that is tuned for privacy.
But I might start using Edge more often because the chat interface for search, it's like the
part of a Wikipedia entry that comes before the table of contents.
It's just the broad overview.
So if you ask a general question, it's going to query a few different sources.
It's going to give you a generalized answer based on information drawn from various sources,
and it's going to give you a list of the various sources that it's drawing from, which is pretty
cool.
But even cooler than that is then you can talk to it about the search results.
You can ask it things.
You can object to things, which I often do.
And then you can ask it to do things for you.
So some of the things I've asked it to do, like I asked it to research me and describe
my niche because, you know, I don't know what my niche is.
I'm interested in stuff.
I talk to people who are knowledgeable about the stuff that I'm interested in.
So I said to Bing, do a bit of research on me and my podcast work and tell me what niche
you think I occupy.
And Bing responded, I did a bit of research on you and found your podcast work, and I
found that you are the creator of the Serum podcast, the host of the Padverb podcast,
and the author and illustrator of a comic book called Geb, the Greater Earth Betterment
Bureau.
You also have a YouTube channel called Out of My Head, where you share your thoughts
and experiences.
Based on your podcasts, I think you occupy a niche that is related to technology, innovation,
communication, and cognition.
You explore the interplay between these topics and examine the role of knowledge networks
and data-driven technologies in helping progress along.
You also interview interdisciplinary thinkers and innovators who have harnessed the creative
power of combinatorial thinking.
I hope this is an accurate description of your niche.
If not, please correct me or provide more details.
Well, I don't know that I really understand the niche description there, particularly
the part about harnessing the creative power of combinatorial thinking.
And I said as much to Bing, you know?
I don't quite get what you're getting at.
That's the fun thing about these large language models is that you can use very colloquial
language with them.
They're trained on human-generated text, you know, on text generated by humans for humans,
which tends to be fluid, free-flowing, not exacting in its adherence to grammatical rules,
spelling, punctuation, or anything, really.
We're pretty good at figuring out what people mean, something that was a uniquely human
power until just recently.
But I asked Bing to take the description of my niche, which it had generated, and turn
it into a description of this very podcast, the KMO Show.
And so I took the text that Bing generated and I just put it up as the description of
this podcast.
And so Bing, aka GPT-4, wrote the following,
Join veteran podcaster, interviewer, and artist KMO in mind-blowing explorations into how
we know what we know and how we can use that knowledge to navigate a bewildering future.
The KMO Show features conversations with interdisciplinary thinkers and innovators on topics like artificial
intelligence, evolutionary psychology, blockchain applications, and more.
You'll also discover how emerging technologies and knowledge networks can help us address
societal changes and create a more prosperous and equitable world.
The KMO Show is a podcast that challenges you, inspires you, and helps you imagine new
possibilities.
Subscribe today.
Now, one thing that's funny to me is that I recognize a couple of phrases in there from
glowing reviews that people have left on various podcasting platforms.
So Bing is in no way averse to a bit of plagiarism here and there.
But it's also surprisingly creative.
And I find it entertaining to just, without even any particular goal in mind, to just
ask it a question and then interrogate it about its answers.
Alright, well, I'm gonna leave it there because in the original outro recording I blathered
on for a long time, but either fortunately or unfortunately, I had my cell phone sitting
too close to the microphone and there's a lot of popping from radio interference.
Very annoying to listen to.
So all that brilliant material that I recorded I will just leave in the Akashic Record and
maybe I'll come back around to it in a video or another podcast or a blog post or somewhere.
Or not.
Anyway, the conversation with Brian Chong continues in Sea Realm Vault episode number
452, which I might have posted tomorrow, but I have a new puppy and she takes up a lot
of my time and attention and interrupts my flow regularly throughout the day.
So maybe it'll be the day after tomorrow when I get that podcast up.
In any event, I will be back here with another free show one week from today.
I hope that you'll join me for that and until that time I hope that you will stay well.

Mocha, my long-time kitty companion is Old

kmo — Tue, 15 Feb 2022 18:36:39 GMT

https://youtu.be/3o-DeWTWg4Q

This YouTube Genre is a Godsend

kmo — Fri, 27 Aug 2021 17:55:03 GMT

In the online world of today "white man" has come to seem synonymous with "irredeemably evil piece of shit." I'm feeling the hate, and I don't like it. But it's not just my skin and my sex that marks me as evil incarnate, I'm also old. I'm not a Boomer, so I don't get THAT level of hatred, not in the abstract. For the most part, nobody who isn't a Gen Xer themselves has anything to say about Generation X, but when young people, particularly young women, see me IRL, their facial expressions and body language communicate to me that I affect them the way a pile of rotting fish guts might.

These categories of race, gender, and age feel like yawning chasms between me and so many of the online communities that I observe but don't interact with. But I've discovered a genre of YouTube videos that make me feel good: Young, black women (and sometimes young men) who do "reaction videos" to musical suggestions that their subscribers recommend to them. The songs are often from 1970s and 80s guitar-led rock bands, and because the suggestions are coming from their audience, often with a donation attached, the young YouTubers are highly motivated to give the songs a fair shake and find value in them.

My favorite personality in this community is RogueRxyce, a young woman who sounds American to my ears. I think it's her bright smile and sparkling personality that endears her to me. She's unrelentingly positive, and I've savored a warm glow watching and listening to her develop a love of Pink Floyd and the guitar stylings of David Gilmore. Pink Floyd was a favorite band of mine in my late high school and community college days, so the familiarity and the positive associations with their songs is inscribed deep in my psyche. Rouge's first Pink Floyd reaction video was to a live performance of Comfortably Numb from 1994. Normally, I would cringe at the thought that someone's first exposure to Pink Floyd would be a post-Roger Waters performance, but Comfortably Numb has always been a David Gilmore song. David wrote it, and it features one of his more recognizable and emotional guitar solos, and this 1994 performance pulls out all the stops. Rouge responded to the song's lyrics, but it was David's guitar solo that literally brought her to tears.

https://www.youtube.com/watch?v=QWslibUABM8

Probably my second favorite YouTube creator in this genre is Empress Joy-Jean. She's another young black woman, but her accent tells me she's from somewhere in Africa. I don't know African accents well enough to say what country she is from. She too has reacted to a number of Pink Floyd songs, but the video of hers that I enjoyed most was when she watched Stevie Ray Vaughn performing a live version of Texas Flood.

I was familiar with the name Stevie Ray Vaughn before his death in1995, but I wasn't really a fan until I discovered his live performances on YouTube sometime in the last couple of years, so this isn't accessing the inner sanctum of my soul like a classic Pink Floyd song does, but Empress Joy-Jean's delight and amazement as Vaughn's passion and superhuman guitar technique touched me just the same. Here's another video in which a young black woman is reduced to tears by a white man playing a guitar.

https://www.youtube.com/watch?v=8G6Zz-4YTFA

Out of sheer laziness, I won't list all of the YouTubers in this genre whose videos I have enjoyed over the last few days, but there are many. There are three young black male/female couples who make videos in this genre, and there's a duo consisting of a young black man and an older black man (who is still probably younger than me) who listen and react to old white-guy music. The elder of the two knows his way around the genre and he guides his young protégé into the land of, "Holy shit, these white guys made music with passion, talent and soul."

There are also young white YouTubers who react to the same body of songs, but I never click on their videos. I want the experience of seeing young, black people discover and appreciate the music of my youth.

One thing these YouTubers NEVER mention is the race of the performers. They never SAY, "Wow, these white men can really play." That's just the subtext. One video which many of the folks in this video genre watch and react to with wholly positive affect is a live performance of Free Bird by Lynyrd Skynyrd. The video features the band in concert in front of an all white audience at the Oakland Coliseum in 1977, and they have a giant Confederate battle flag as a backdrop. Even with this unmissable symbol of white supremacy on screen, no black YouTuber I've seen yet has said anything about it. They focus on the music, the energy of the band, and the excitement of the crowd.

To give the young men their due, here's Jojo reacting to the album version of Free Bird (in which Jojo does mention race but only to say that he didn't grow up listening to this kind of music): https://www.youtube.com/watch?v=mjwubox38lE

As ugly as race relations have gotten in this country in the last couple of decades, these (mostly) young YouTubers are helping to heal our societal wounds and close what have come to seem like unbridgeable gaps between us. Or that's what I imagine anyway as I sit here alone (except for my cat) in my apartment watching YouTube on my Roku TV during this interminable COVID isolation.

Another death

kmo — Wed, 11 Mar 2020 00:31:29 GMT

My mom called me tonight to tell me that they took Ungo to the vet today to have him put down. He was 15, which is a pretty good age for a dog his size. He hadn't really been "my dog" since mid 2008. Really, he was my brother's dog. Still, he always recognized me and was pleased as punch to see me when I would visit Arkansas, no matter how many years since he'd seen me last. And he was such a happy, good-natured soul.

I voted

kmo — Tue, 03 Mar 2020 16:25:03 GMT

I'm not the kind of person to not know who I'll be voting for as I roll into the polling station, but today, my pencil hovered over the Andrew Yang oval for an instant before I filled in the oval by Bernie Sanders' name.

I live in Vermont, so my vote doesn't matter (outside of the very local level). My state's primary contest couldn't conceivably come down to my vote. The decision of which oval to fill in amounts to a conversation I was having with myself.

Ultimately, I voted for Bernie because he is an active candidate and the Neoliberal power-brokers in the DNC are pulling out all the stops to undermine his campaign. If Andrew Yang had stayed in the race, I would have voted for him, but the contest now is between the Sanders revolution and the status quo, and I despise the status quo Democratic Party.

Corona Virus Panic continued

kmo — Mon, 02 Mar 2020 15:31:35 GMT

Yesterday, I posted a video to YouTube about how panicking in response to the news cycle obsession with the most recent infectious disease outbreak is maladaptive. In response, I got push-back from multiple people telling me why I'm wrong about THIS panic-inducing story. Here's the best written one that I've received so far, (for ease of viewing, I have inserted paragraph breaks not present in the original):

____________________

It's extremely important to understand that most profitable media organizations (I assume) tailor their coverage to a certain agenda, or a certain personality type/traits, or certain political/ideological/religious beliefs. Thus they are incentivized to adhere to the interests of their audiences and/or backers or advertisers. And most audiences/backers/advertisers have opinions or agendas.

Where is the market for a 100% dispassionately objective news source? Or a completely fair and balanced news source? I don't think most people even have an accurate sense of what that would even look like. Because we ourselves are all biased and not 100% aware of all the ways in which we are.

That said... it is also extremely important to listen for important factual information (sans opinion) within the content of any news presentation. Sure, it's likely to be selected-for in a biased way. But you are free to speculate on what information or analysis they might be leaving out, and to look for that information yourself if you choose. Be aware of what might be spin, but also be aware of what might be actual valuable information.

Unfortunately, that puts you in the position to do your own underinformed analysis, without just trusting and accepting the presenter's analysis. You can at least trust your own motives more than you can trust the presenter's motives, but you yourself likely aren't in a good position to know the whole story. Or there are additional factors at play that you have not considered and neither has the presenter.

Perhaps the best we can do is listen to diverse perspectives, collect information and counterpoints from all of them, pit the points and counterpoints against each other in our minds using the available information, and do our own best-guess analysis.

[So far, so good. -KMO]

All that said: I'm still waiting to hear some kind of information to indicate that if I just go about my daily life with no serious change I am NOT likely to catch the coronavirus. I'm just not seeing any hopeful trends yet from any news source that make sense to me as having factored-in all the information I've absorbed about the virus. China may be reducing transmission now, but they locked down their economy in a way we probably wouldn't. (That's a big assumption, I guess.)

There is hope that the death rate is not as bad as people worry it is, but there doesn't seem to be any hope that it's not going to just keep spreading quickly and relentlessly. Exponentially, barring a lockdown. All signs are that it is particularly contagious, spreading via air, and that herd immunity is lacking. And that a vaccine is over a year away unless they somehow bypass the normal clinical trials.

So... I'm just going to catch it, then? And the rest of my family? If any one of us gets it there is a high likelihood the rest of us will get it, if it's the same contagious strain that is being reported now. (Whatever strain arrives in our area is likely to be more contagious vs. less contagious.) Will my elderly parents and relatives be OK? ....Are they going to catch it anyway on their own, despite my own actions to avoid catching it myself? How much inflated are the reported serious cases and death rates due to undetected cases?

.....Should I buy another freezer and extra food, and masks and gloves? Food I would eat anyway? If I don't, are other people going to buy up all the food too quickly? If I do, am I contributing to an unnecessary, destructive panic? Are supply chains going to be OK? Is it really going to take a year and a half to get a vaccine?

And the biggest question of all: Is it worth racking up even more interest on my credit card to buy all that stuff now vs. later or never? And how am I going to pay it off if I'm hunkering down at home for 3 months?

We don't have to freak out about ANYTHING. And we don't need to take anyone's biased analysis as gospel. But that doesn't mean we should ignore information or decide not to take action to target the outcome we want. We don't have to do that either. We can choose to listen, analyze, and act toward whatever goal we want. It doesn't need to be based on — or in spite of — anyone else's ideology or agenda.

While absorbing information we need to be mindful of the bias and the absence of information. But if you go around to multiple news sources and all the counterpoints seem weak based on solid-seeming information you've heard elsewhere.... It's OK to say "hmm, maybe I should take this seriously, and what can I do to get the outcome I want?"

There's something to be said about not stressing too much and pathetically clinging to control in a world of chaos. You're just... free to choose if and when to do that. Sorry for another long ramble with multiple points rolled together.

____________________

My response:

Wow. First, let me say, "Thank you for posting such a substantial comment. It seems like you really put some time and thought into it."

Second, I'll share with you the same advise I have shared with other people who have sent me justifications for focusing a lot of attention and emotional energy on the COVID19 story: Cut and paste your comment into an email that you send to yourself, and then set up some sort of reminder to review that email one year from today. By then, this current obsession will have given way to some new story about how something is about to wreak havoc on civilization. Then take note of how you are responding to the current story? If you're just as convinced that the panic-inducing news cycle item du jour is as legit as you now think the COVID19 obsession is, that probably indicates that you're getting some sort of reward from the anxiety itself.

Third: "Should I buy another freezer and extra food, and masks and gloves?"

Making a large purchase will produce a dramatic (if short-term) neurological effect. So, if your desire is to disrupt your feelings of anxiety and if the purchase won't drive you into debt, sure. Buy away.

But if your desire is protect yourself and your loved ones from infection, then you'll get a better return on investment by taking steps to improve your immune function. Set a bed-time alarm so that you get enough sleep. Exercise. Meditate. Eat better and/or less. Stay away from YouTube and social media. You may think that these are all selfish responses as they don't cover your loved ones, but setting a positive self-care example for them will do them more good than making large appliance purchases.

And really, if you want to buy something that will increase in value, buy stock in a solid company that has taken a hit along with the rest of the market as a result of this panic. That's what the Oligarchs are doing. Remember, every time someone sells a share of stock, someone else buys that share. Panic selling enriches people who take a long view.

_______________

As my lungs fill up with fluid and I'm gasping for my final breath, please feel free to remind me to review this LJ post.

Corona Virus Panic

kmo — Sat, 29 Feb 2020 15:06:04 GMT

The following is clipped from a private email correspondence. There's nothing particularly private in this bit other than my personal framing of the current hysteria propagated by the corporate news media and professional hucksters, catastrophists, and fear-mongers.

------------------

Personally, I don't think everything is falling apart. I think that this corona virus panic is just that, a panic with very little basis in objective reality. If the worst case scenario plays out, and a full 2% of the global population dies off in the next year, it will be a minor blip in the story of technological civilization. Complex adaptive systems typically benefit from minor perturbations. It would suck to lose a loved one to a communicable disease, but it's also a fixture of the human condition. A much greater challenge to our civilization takes the form of demographic unbalance. Old people linger for much longer after the end of the working lives than they used to, and younger people have forgotten how to make babies. Covid19 is like an unexpected expense that screws up your monthly budget. Our demographic malformation is like a crushing debt load that you carry for decades and which shapes the course of your life.

I could even see this panic doing some good if it prompts the US to cast a critical eye at how dependent we have become on global trade, particularly trade with China. Also, it has employers looking for ways to let workers stay at home and still get paid. I'm of the opinion that most of the "work" that the vast majority of people do these days is useless and would be better left undone and that the energy we expend moving people from their homes to their places of employment every day could be put to much better use, but as a culture we're stuck in the grip of a maladaptive belief that people don't deserve food, shelter, clothing and medical attention unless they're toiling in the service of further enriching the oligarch class.

That's it. No more print interviews.

kmo — Sat, 08 Feb 2020 12:52:07 GMT

Apparently, something about the way I put words together does not compute for the humans and algorithms that turn recordings of spoken sentences into text.

There's a chapter about my life and career in a book that will be released in less than a month. The author recorded our conversation, which at the time I thought would help the details of what I had to say make the transition from spoken conversation to the page, but in fact it had the opposite result.

For example, I voted for Gary Johnson, the Libertarian Party candidate, in the 2016 presidential election. In explaining my outlook and relationship to the mainstream mentality, at one point in the conversation, I said, "I didn't vote for Trump." What appears in the book is, "I did vote for Trump."

When I informed the author, he apologized and explained, "I worked off a transcript of a recorded interview prepared via Print, the transcribing service… which is powered by algos, not humans."

If you search Amazon for KMO, you will find one book for which I am listed as the author. It is a collection of transcribed interviews, and I know from hard won experience that transcripts that you pay to have made are never accurate. NEVER. Some of the transcripts in my book were prepared by a volunteer, and those were excellent, but people who are doing it to make money are likely in a hurry, harried, money-stressed, and, God bless 'em, often not very smart. I turned my recorded interviews into print ready documents in 2010. The work was all done by humans, and it cost a pretty penny, and I still had to spend a lot of my own time and effort to correct the mistakes.

Since then, algorithmic transcription services have hit the scene, taking paying work away from humans. From time to time I've thought that I could have all of my interviews transcribed to post to my website now that algorithms have made the cost trivial. I discovered that the transcripts were always so riddled with errors that the time and effort that I would have to devote to straightening them out started to approach the time it would take me to simply create the transcript from scratch myself.

Most of those transcripts, inexpensive as they were, were a waste of money, as I didn't ever carve out the time needed to compare them to the original recording and correct the mistakes. Sometimes, the transcription algorithms produced such gibberish that I couldn't even describe their output as containing mistakes. To say that the machine made a transcription error implies that there was some evident correspondence between the recorded speech and the supposed transcript.

While algorithmic transcription services do the job faster and cheaper than the humans they displaced, where the human transcribers often didn't understand some of the ideas expressed in the words they were converting from sounds to strings of letters, the algorithms that now do that job understand nothing at all. Understanding is not their bag. It doesn't play any role in what they do.

I now recall reading somewhere, decades ago, that you should avoid using the word "not" when its omission would change the meaning of what you wanted to say.

Dangerous: Do not use the word "not."

Less dangerous: Avoid using the word "not."

But even safer, I now think, is to decline to be interviewed for print. In audio, I have better control over my message because I'm not relying on the interviewer to understand my meaning, remember it and relate it to the audience. I can talk past him or her and communicate directly with the listener. The person who made the recording can edit it to misrepresent my intentions, but that requires an act of deliberate malice. In print, my intended meaning can mutate into its antithesis even when the interviewer means well.

At present, withdrawing from my career as a very minor public figure is not an option. That's how I pay the rent, but I've learned a lesson here, and now I'm thinking that I might want to start looking for the exit and find a way to make money that does not involve being known to people whom I will never meet.

Fuck

kmo — Mon, 27 Jan 2020 15:06:09 GMT

Yesterday, I heard that Kobe Bryant died in a helicopter accident. I knew that he was a basketball player, and that is all. I didn't know what team he played for or that he was retired. This morning, I called up his Wikipedia entry and looked at his picture. Not even a glimmer of recognition. He was nobody to me.

Yesterday, when I heard (or read) the news, I knew that millions of Americans (and probably as many Chinese basketball fans) would be emotionally devastated by it, and I resolved to post nothing about it. I understood that adopting the "I don't care about sports" pose in public at such a moment would be in poor taste.

As for the dangers of riding in helicopters, "Rich people problems," I thought. Again, best not to push that particular line at a time when people are mourning the death of a beloved rich dude.

I remember being in pain over the death of Prince and mentioning it to a young cashier at my local grocery store. I don't think he actually rolled his eyes, but his expression and body language were clear. "Old people problems." That made the hurt worse.

A handful of hours, at most, after learning of Bryant's death, I got a private message on Twitter informing me that a Scottish man named John had died. I'd never met John in person, but he was dear to me. John was one of my earliest podcasting peers. Under the sobriquet Queer Ninja, he was the creator of a podcast called The Sounds of World Wide Weed. It was a music podcast, and while Queer Ninja had excellent taste in music, it was his gentle personality that really made listening to his show such an enjoyable and uplifting experience.

"Easy, man," was his signature intro, and he often couldn't complete a sentence because his infectious, good-natured, stoned laughter got the better of him. For a time, I used a recording of his laughter as my cell phone ring tone.

While I've had a Twitter account for many years, I never really got into Twitter until just recently. I got kicked off of Facebook a year and a half ago, and no social media platform has risen to take it's place, which is probably for the best.

Only recently have I discovered Twitter's value as a procrastination prop. I made several posts yesterday that all seem trivial, snarky and unworthy to me now.

I didn't even know that Queer Ninja was on Twitter, though in hindsight, it's obvious that he would be. I didn't follow him, and he didn't follow me. I follow him now, though he'll never post again, nor will he ever know how his final Tweet affected me when I read it.

I didn't know that Queer Ninja struggled with addiction until I received word of his death. The person who contacted me told me that he died in his sleep and that he had been clean from heroin for 50 days.

When I looked at his Twitter feed, I saw a pinned Tweet dated 10 Dec, 2019:

That was 48 days ago.

When I learned that he was gone, I went to my Gmail inbox and searched for his email address. The last message I received from him was dated Wed, Jan 28, 2009, 9:41 AM. Eleven years ago tomorrow. I had no idea it had been that long.

In his final Tweet he gushed with enthusiasm about the premier of Star Trek: Picard. I so wish that I could talk with him about the show, or about anything at all.

In recent years, I have let my reclusive tendencies get the better of me. If I'm on a recorded call with someone, I can put on my KMO podcast personality and have an engaged, animated discussion on a variety of topics, but personal phone calls are awkward and uncomfortable for me. There are friends that I haven't spoken to in years, but actually calling them is a psychological hurdle I haven't managed to clear. Talking to my kids by phone is similarly beyond the threshold of my good intentions (though recently we have connected by Zoom while playing together in a shared Minecraft Realm). My mom calls often. My brother calls once in a blue moon. I'm friends with a client of mine, and sometimes work calls detour into extended, non-work-related tangents, but the only people I ever talk to by phone are people who call me. And few do. (This is not a plea for people to call me. I don't answer calls from unknown callers.)

Queer Ninja wasn't a celebrity, though he did touch the lives of many people he didn't know. I did know him, but not as well I wish I had. Had I known him better, today would be even more painful for me than it is. I wish those who knew and loved him as much emotional comfort as the day will allow. I'll say no more for now.

"You can get what you want and still not be happy." -Cerebus

kmo — Fri, 12 Apr 2019 19:32:40 GMT

Well, my designated loser of a next-door neighbor seems to be in jail. He had a parole hearing on Wednesday, and it seems like he failed to pass muster before the judge. I haven't seen him since.

The thing I was afraid would happen didn't. I thought my neighbor would get pinched but that his crew would continue to operate the business out of his apartment in his absence. That hasn't happened.

I could hear someone moving around in the apartment the night before last, but the usual traffic has ceased. I got what I wanted in that the constant stream of human detritus is no longer flowing in and out of my building all day, but my neighbor is not a violent person. He's never been rude to me, and he even seems to have a calming influence on the members of his crew who display rougher edges than he does. My world is a better place now that the drug dealing has moved elsewhere, but the world is no better for his being in jail.

Amy Webb and the problem of too many white men in the AI development tribe

kmo — Fri, 05 Apr 2019 15:12:33 GMT

I just posted this to r/artificial. I joined the subreddit to make the post, so I have no established reputation or credibility with its regular members. I don't expect my post to elicit many responses.

Posted byu/Kayemmo
just now
According to Amy Webb, a major problem with AI is the over-representation of White Men in its development.
I'm re-listening to the audiobook version of The Big Nine: How the Tech Titans and Their Thinking Machines Could Warp Humanity by Amy Webb. Throughout the book, she stresses that a major problem with AI is that it is being developed by white men. That means that many of the unquestioned assumptions that white men hold are given uncritical acceptance and get encoded in the AI systems that white men create. In the book, she asserts the following:

AI's tribes are inculcating a culture in which women and certain minorities, like black and Hispanic people, are excluded, plain and simple. In 2017, a Google engineer sent around a now-infamous memo arguing that women are biologically less capable at programming. Google's CEO, Sundar Pichai, eventually responded by firing the guy who wrote the memo, but he also said, "Much of what was in that memo is fair to debate."

I read the Damore memo back when the controversy was current, and I don't recall Damore claiming that women are biologically unfit to program computers. That strikes me as a deliberate misrepresentation of his point. Also, Webb's criticism of Sundar Pichai for saying that some of the memo's comments where fair topics for debate seems to imply that if a work includes anything that a woman or person of color finds objectionable, then, by default, we must assume that the entire work, and anything its author might say on any topic, must be deemed off limits for discussion.

It would be easy to simply take offense at Webb's prejudices and script the review as a straight-forward condemnation and retort, but I'd rather do something less obvious and reflexive. I wonder what insights or alternative takes the members of this community can recommend.

10-Day Fast Wrap-Up

kmo — Sat, 23 Mar 2019 01:21:37 GMT

https://youtu.be/yzDyB6TZRnA

I Just Broke My Fast

kmo — Fri, 22 Mar 2019 12:58:16 GMT

I did it with bone broth supplemented with a paste called Better Than Bullion. I'm dying to eat more, but not for a couple of hours, at least. My pulse seems elevated, and I feel the threat of a headache. Also, while my taste buds and brain heartily approved of my first calories in over 10 days, my stomach is not entirely on board with the whole project.

Yesterday, I weighed in at 200.0 lbs. I was looking for that scale to say something like 198.4 this morning. Instead, I weighed exactly the same this morning as I did yesterday morning. That's a little frustrating, but just a little bit. I did lose nearly 17 pounds in 10 days (though, I know at least 5 pounds will return in the next few days).

The Finish Line is in Sight

kmo — Thu, 21 Mar 2019 18:04:12 GMT

Today is the final day of my 10-day fast. I started at 216.4. This morning I weighed in at 200 pounds even. But the fast isn't over. I break fast with bone broth tomorrow.

Bone broth, if you've never had it, tastes like nothing in particular, so I'm about to go to the nearby supermarket to look for some miso soup. I'm sure they won't have it, and the nearest upscale store is a half hour's drive each way. Skip that. If the local store doesn't have miso, I'll get some bullion cubes.

I soooo wish this day was over, or at least I wish it was 8:30 PM so I could watch the new episode of Star Trek: Discovery, but wishing that you were X number of minutes, hours, or days closer to the end of your life is no way to live.

I also abstained from playing video games on this fast, and I plan to continue that trend, at least avoiding them until the end of the day when all of my work is done, and not first thing in the morning while I drink my morning coffee.

I also stopped drinking coffee on this fast. Do I plan to get back on it? Quitting is hard. Staying off it is the easy part, but I've heard that it has neuro-protective benefits, and given how much I enjoy it, I'm anxious to believe those findings.

Today is the final day of my fast.

kmo — Thu, 21 Mar 2019 11:03:26 GMT

Final weigh-in is tomorrow, but I need my white board.

The Den

kmo — Tue, 19 Mar 2019 12:55:15 GMT

One setting in the Antarctic base we haven't seen since the strip was in the single digits.

Captian Marvel is not as bad as some folks want it to be

kmo — Sat, 09 Mar 2019 14:53:37 GMT

https://youtu.be/y9lZr6rwJkM

I went to see Captain Marvel last night. It was the first public showing of the film in my rural Vermont village. I'd followed the controversy swirling around this film, particularly Brie Larson's intonation giving away her insincerity when she claimed that she doesn't hate white dudes. You know, not murder them with a rusty spoon-level hate, anyway.

I considered not seeing it, but I knew that the character of Captain Marvel would be playing some role in Avengers: Endgame, and I want to get the maximum enjoyment out of that film when the time comes. Also I went to see it because I knew that I'd be tuning in to the chatter surrounding the film, and I wanted to be in a position to form my own opinion about it.

I was NOT looking forward to the experience. Based on some of the early reviews, I expected to enjoy Sam Jackson and Ben Mendleson's performances, and that expectation bore out. They were both great. On the other hand, I thought I was going to have to grit my teeth and endure Brie Larson's performance. I formed this expectation based on the marketing for the film and from Brie Larson's public comments long before there were any reviews to read.

As it turns out, Brie Larson was fine in the role. The trailers made it look like she was channeling Steven Seagal to find her Captain Marvel mojo, and that just wasn't the case. As much as I dislike her public persona, the reports of Brie Larson's wooden acting are exaggerated. She smiled at times. She may even have laughed. Scenes where she was supposed to be in emotional turmoil came across as under-acted, but I've known people experiencing profound anguish, grief and turmoil who kept it all under wraps in the moment only to express their pain later, in stages, at a pace no faster than they felt they could process.

Believe me, fellas. Star Wars: The Last Jedi pissed me off, as did the first season of Star Trek: Discovery. And I was never even remotely tempted to watch Ghostbusters 2016 either in the theater or on the small screen. The obnoxious SJW infiltration of my beloved scifi (read "skiffy") and heroic fantasy genres irritates the hell out of me, and I was not only sorry but genuinely surprised and disheartened to hear that this creeping brain-rot had metastasized to the MCU.

It hasn't. And Brie Larson is fine as Captain Marvel. Kevin Feige and the Masters of the Marvel Universe aren't telling the established audience to fuck off. At most, they're saying, "Hey, there are more folks coming in. Slide over and make a little room for them, will ya?"

Is the character of Captain Marvel over-powered? Compared to Spider-man or Captain America? Sure. Compared to an Infinity Gauntlet-wielding Thanos? Not so much.

I read one review that described Captain Marvel as being just a set-up for future MCU installments. Take the "just" out of there, and that claim stands, but there are obvious moments in Captain Marvel that are ends unto themselves. The MCU is the paradigm example of an interconnected tapestry of stories and of a shared sand-box for established and up-and-coming directors. Did Captain Marvel set-up future movies? Of course. That's how MCU movies do.

Some folks have said that Captain Marvel feels more like a DCEU movie than one from the MCU. Yeah, I got that vibe too, but what do DC movies "feel like?" They're all over the place. DC is a solid decade behind Marvel in finding a consistent tone and look for their films. They keep trying to course-correct and distance themselves from past failures, even going so far as to change the tone of films that have already finished principle photography. DC films are all over the map, so to say that Captain Marvel feels more like a DC film is just to say that it doesn't feel as thoroughly integrated into the Marvel vision as the rest of the films do, and given that it's a period piece from the mid-90s, that's as it should be.

Is the blue cube Infinity Stone in a different place at the end of Captain Marvel than other films suggest that it should be? Honestly, I don't care.

Is the explanation for Nick Furry's ruined eye dumb? Yes, that was a complete waste. I'm with you there. That deserves a do-over.

To bring this to a close, I'll just say that yes, I got extra baked for this one in order to take my critical reflexes down to their lowest setting. Is this cheating? Maybe for you it would be. I wanted to enjoy the film, so I took the steps that I thought would help me achieve that goal. If you want to enjoy hating the film... well, that's your business. If admitting that this film isn't as bad as all that would get you kicked out of your favorite club or clique, then hate away. Prime yourself to take offense where none was intended and revel in the bonds of cheap solidarity. Just know that that's the exact same MO that the cry bullies and the professionally offended use to navigate the world of pop culture. If you relish the conflict, then do what you think needs doing, but this film, while far from perfect, isn't bad enough to justify the anger directed at it.

I'm talking about the actual film here. The marketing was definitely sleazy, and some metaphorical heads should roll for that bullshit.

Just Chill, My Soul

kmo — Wed, 27 Feb 2019 02:29:59 GMT

A basically contemporary vernacular translation of Be Still, My Soul, Be Still by A E Housman.

Chill, my soul. Just Chill.

You are not strong enough to take on the whole world.

If you're really freaking out

just think back to the time when it was all good.

Bad stuff happened back then, but I didn't care

because I hadn't been born yet.

Now I'm alive, and it's a fucking mystery to me

what I did to deserve this shit.

Fuck.

Shit's fucked up, but at least life is short.

Might as well kick back and watch the assholes be assholes.

See what I'm saying?

Life offers a full range of sucky experiences.

Mad as hell, scared outta your mind, dissed.

Don't tell me I chose to be here.

This is bullshit.

A Thousand Words

kmo — Sun, 10 Feb 2019 17:58:15 GMT

I posted a video on Friday, February 8th, 2019 in which I talked with my young friend, Collin, about, among other things, our expectations concerning technology and the future of technological civilization. Collin expressed the opinion that technological progress will continue and, in spite of some potential rough patches, that humanity would "keep going up." I don't recall shooting down the idea of perpetual progress in that video, but I do harbor some doubts, and I guess that showed through in my expression, because YouTube user Apjooz posted a comment asking:

"What development do you think would need to happen for you to shift your thinking in terms of decline? Like what trends would start to convince you that maybe Colin's right and we'll keep going upwards instead of declining."

My answer turns on who the "we" in your question refers to.

I'm concerned that whole sectors of employment are either being automated out of existence or exported to countries with lower wages, fewer worker protections and lax environmental regulations, something that modern information and communications technology makes possible. Instead of freeing people from onerous labor, the benefits of this transformation are going to a tiny minority while the middle class that took form in the so-called "West" in the post-war boom years of the 20th Century evaporates into a growing class of people who will fail to do as well at their parents did.

Ask a technophile about that and they'll tell you that, historically, new technology and automation created more opportunities for paid labor than it destroyed. That might be true, but that doesn't mean that the person whose livelihood is destroyed will find new work that will allow him to benefit from the technological "progress" that disrupted his old way of making a living. He may not be qualified to do the new sorts of work, or he may not be in the right place to participate in the newly created employment sectors.

What comfort should a 17th century agrarian peasant take from that idea when the commons on which he once farmed for a living is enclosed to raise sheep to supply wool to an increasingly mechanized textile industry? Perhaps he migrates to a city where he finds work in a mill or factory. The working conditions are dangerous, and injured workers are discarded without compensation. The man grew up in a rural village where he knew everyone. Now he lives in an overcrowded and squalid slum in the city where he is surrounded alcoholism, violence, abuse, disease, prostitution, human predation and the host familiar social ills associated with the concentration of poverty. As he toils in low light for long hours, watching other workers, many of them children, injured and killed and replaced without so much as a moment of silence to acknowledge their sacrifice, how seriously should he take the idea that it's all good because his great, great grandchildren will own smartphones and work as baristas, cashiers, and house cleaners? Does that guy count among the beneficiaries of the industrial revolution?

Suppose you tell a Neanderthal that she and all of her descendants will die out leaving the world to homo sapiens who will multiply and populate the Earth to an unimaginable density. Instead of struggling with hunger and exposure to the elements, they will suffer from obesity, diabetes and the other so-called diseases of civilization. She and everyone she knows will be long dead by the time this happens, and in fact her species will soon be extinct, but the successor species who will build cities and eventually build machines that will leave the solar system, will carry a small percentage of Neanderthal DNA. Does our Neanderthal then count in the "we" who will master technology and set out toward the stars?

What's a really hopeful outlook for our future? About the most optimistic vision offered in popular fiction is the Star Trek future where people routinely travel in faster-than-light starships. Everyone has access to first-rate healthcare, clean and secure housing, satisfying work, and endless varieties of nutritious and delicious food from a replicator mounted in the wall. The holodeck can recreate any historical environment or any far-flung or fantasy location. The lucky inhabitants of this prosperous future society can learn from the great minds of Earth's history as well as from those of a thousand extraterrestrial cultures.

But between our present and the Star Trek future stands nuclear war and a struggle between augmented humans and the unmodified majority that will claim millions of lives. The prosperity we know today will give way to squalor and deprivation that will seem like the end of the world to the people living through it. Nobody alive today will see the Star Trek future, and only a tiny minority will have descendants who make it through the coming population bottleneck to roleplay as Sherlock Holmes or Captain Proton on the holodeck of an interstellar vessel. Does that count as a bright future for the people who will die in the coming wars and the lean times to follow?

Who counts as "we?" From my perspective, "we" must include me and my children at the very least. Since Apjooz asked the question, he or she must also count as a member of the "we" who will enjoy continued technological and social progress. Given that definition of "we," here are a few developments that would suggest to me that "we" will enjoy a bright and prosperous future:

The carbon content of the atmosphere stabilizes and starts to move back toward pre-industrial levels.

The number of species lost to extinction each year slows dramatically.

The Gini Coefficient, a statistical measure of economic inequality, stops climbing and begins to fall in rich and poor countries alike.

The prison population in the United States decreases to pre-Drug War levels.

Politicians, tech executives, and thought leaders acknowledge publicly that the wealth gains created by information technology have mostly been captured by a tiny oligarchic elite. They would then need to propose some means by which to include all humans as stakeholders in the glorious high-tech future.

If any three of those five developments manifest in my lifetime, I will probably join the technoutopians and agree that the future of technological civilization is bright.