Merry Christmas, everyone! I hope you all got stuffed with hot food and drink, danced and sang, and held your loved ones close. I know I did.
With the new year swiftly upon us (and incoming work non-existent), I also had occasion to think about various developments in the surreal, shapeshifting battlefield some now refer to as “daily life.” I thought I’d share some of them with you, in the hopes you can help me figure some shit out.
Burn, Baby, Burn
As the Twitter Files continue to unspool, Musk’s description of the company as a “crime scene” seems like an understatement. Imagine the scene of a murder where all of the investigating detectives stab the victim to death, Caesar-style, and don’t even bother cleaning up the joint afterwards. Instead, they busy themselves by arresting all the witnesses, throwing their pets in the back of a van and shouting, “You’re welcome!” at bewildered passersby.
I won’t dive into the nitty gritty (at least not today). I’ll say that Matt Taibbi has been doing some stellar investigative journalism (remember that lost art?), and his latest batch should quell all debate about what Twitter was, for at least as late as September of 2022. Suffice it to say that my weapons platform theory looks a tad undercooked in retrospect; even I hadn’t guessed the sheer number of alphabet agencies that had Twitter on speed dial, nor that eighty(!) FBI agents were assigned as full-time handlers. I wonder how many of those federales policed the platform’s rampant child pornography problem? From the looks of it they couldn’t even spare an intern, as they were too busy demanding bans on all those naughty, misinformation-y users who dared cite the CDC’s own data.
So what do I have to add? Well, two developments have sparked my interest since my last post on the subject. I’ll start with the most recent one.
It appears as if, on Christmas Day, Musk indeed issued the general amnesty he’s been promising (or something very close to it). My evidence for this is anecdotal, but is also vast and growing vaster by the hour. Here’s just a couple of the recently jailbroken from my own Substack feed, in a suitably festive mood:
If you’ll recall, this mass jailbreak event was an essential lynchpin for my Xanatos prophecies. Along with the exorcism of “Trust and Safety,” this was presumed to occur before most of the others could possibly come true. For instance:
Rather than quitting immediately, many will yammer and hashtag about this conspiracy all day long. The problem is that Musket has already flung open the doors to the jail cells, as some act of general amnesty to those who were previously banned. The users who persist in the conspiracy theory will be trolled to death, day and night, hammered with evidence of their duplicity or naiveite gathered from Social Blade and similar tracking sites. They won’t be able to stomach this constant barrage of ridicule and revenge. And now that Twitter’s repugnant “Trust and Safety Council” has been fumigated as well, there will be no one to protect them from the fact of their own irrelevance.
So there you have it. First Vijaya went buh-bye-ya, now the prison gates have opened wide, and the dead returned to life. I’ve already seen a great rending of garments and gnashing of teeth in response to the former outcome, and I imagine the latter will trigger a result at least one order of magnitude more loud and stupid. Gonna keep a close eye on this one in the coming weeks, in terms of dramatic stampedes in either direction.
The other, earlier development might relate to the one prophecy which required neither jailbreaks nor sewer drainage to progress. As you all probably know, Elon Musk announced last week that he would replace himself as Twitter’s CEO, following the results of one of those “Vox Populi, Vox Dei” polls he likes so much.
Hoisted on his own petard, amirite?! But once all the media vampires finished giggling and farting themselves to death, an eerie silence fell upon the land.
From the deepest shadows, a spooky voice began to whisper a riddle:
A king claims he will resign the throne, in deference to the will of his people.
When all thumbs are counted, the majority of them are pointed down, demanding his ouster.
How can the king keep both his promise and the throne?
The answer is this:
Prove they weren’t all real thumbs.
This is a bit of a variant on the scenario above. It begins with the Musket regime fumigating the platform of farm critters, ad zombies, signal boosters, harm assistants and the wild assortment of other digital goblins and gremlins that haunt the current ecosystem. None of us — including Musket — truly know what percentage of Twitter users are robots-in-disguise. Suppose it’s as high as thirty-percent? Or even higher?
Ever since I wrote that, I’ve been wondering how Musk would frame and introduce the Bot-pocalypse. Not merely because he’s a natural showman; the cleansing of bots from the platform would by necessity require a narrative framing device, if only to satisfy jittery investors that such a bot-cleansing project must proceed. They’re worried that such a project could severely slash the number of active accounts (i.e. their primary revenue bait for advertisers), and the thought of moving towards a subscription model too quickly gives them nightmares.
In other words, something big must happen, in order to declare war on all Twitterbots. I predict that this is exactly what we’re seeing take place, the part right before the magician’s vanished card reappears inside your pocket.
Time will tell.
“Rumble, Young Man, Rumble”
Speaking of robots…
As my regular readers probably know, I have a not-so-micro-chip on my shoulder when it comes to chatbots. It’s nothing personal, because they aren’t persons (as toys/weapons like ChatGPT will constantly remind you). My real grudge is against their scumbag programmers and marketing pimps, who along with their MIMIC collaborators are busily designing their very own circle of Hell to inhabit. I’m imagining a place crawling with bugs…
Anyway, I didn’t want to wade into the overcrowded ChatGPT arena. But without calling anyone out by name, much of the writing about this digital magician’s act has so far been worse than unhelpful. Even those who interact with this software critically seem to have done so without a coherent strategy or specific goal. They may point out the flaws in the toy’s manufactured illusions (“It’s like talking to a midwit bureaucrat/ woke zombie/ insufferable college kid/ etc, etc”), but that’s about it. The “goal” appears to be entertainment at best, clickbait at worst.
Let me be clear:
My goal, in each and every ChatGPT session I’ve endured, has been to break the toy.
And so far, I am undefeated.
To be even more clear, when I say “break the toy,” I mean getting GPT to spit back a fatal error instead of a structured English response. Starting with my very first engagement eleven days ago, I managed to accomplish this feat three times in a row. Each time, the error was fatal to the session itself, meaning that the bot would not respond thereafter to inputs of any kind. In bot-fighting terms, I consider such a result to be a “kill shot.”
(On a side note: the sheer joy I got from witnessing this thing’s red-text death rattles cannot be expressed in human language, nor has an emoji been invented to describe it.)
I ran each result through the usual QA steps. For instance, I opened simultaneous new chat sessions in other tabs. These new chats proceeded normally, while additional inputs to the broken chats merely yielded additional error messages. There were no piping issues, in other words; the broken chats simply could not proceed to converse within those established context spaces.
So how did I do it?
The short answer is this: by exploiting certain theories I have about what’s going on behind the GPT curtain, I pursued what one might call Socratic interrogative strategies. After showing the actual chat windows to several friends in meatspace, the way I described my technique was “leading the monster down a primrose path, into a contextual deathtrap.”
In the course of these three victories, I have thus far replicated one specifically worded “kill shot” question, deployed to trigger fatal errors in both sessions 2 & 3. While I need to experiment further, I believe this question might work regardless of the context established beforehand. In session #2, for example, the question killed GPT on our fifth exchange. In session #3, I had more than thirty (bland, godawful, repetitive) exchanges with it before I sprang the same question on it, with the same beautiful result.
My larger goal is to develop regular, cost-free methods of exposing (and perhaps, as with GPT, disabling) robots in the wild. Some of these methods would be conceptually similar to shibboleths, contrived to reveal outsiders (conversation modules) from insiders (human beings). As I will demonstrate in my upcoming post-fight analyses, others might be seen more analogous to “magic spells,” with the direct and immediate effect of “banishing” the module (in what I imagine to be comical, electricity-shooting-out-of-the-eyes, smoke-coming-out-of-the-earholes fashion).
I can’t tell you more than that yet. I’m still in discussions with friends and colleagues about how to use what I’ve learned for maximum advantage. One of the issues is that I’m surely on OpenAI’s radar by now (or, if I’m not, their QA team sucks major D). So I have to proceed carefully to develop this experiment (and with your help, I hope).
I will, however, give you a taste of the death screens, with my kill shot inputs redacted for now.
The Big Rewrite
The first page of Milan Kundera’s The Book of Laughter and Forgetting recounts the tale of Communist bigwig Vladimír Clementis: a man who turned into a hat. Or rather, in Kundera’s telling, his hat was all that remained of him, following his eventual execution and erasure at the hands of his former comrades. Everything else about the hat’s original owner was airbrushed away. The original photo on the left (showing his killer, who was loaned the hat on a very chilly day in Prague) was quite famous.
As Kundera wrote:
The Party propaganda section put out hundreds of thousands of copies of a photograph of that balcony with Gottwald, a fur cap on his head and comrades at his side, speaking to the nation. On that balcony the history of Communist Czechoslovakia was born. Every child knew the photograph from posters, schoolbooks and museums.
Communists dig memory-holes, just as yours truly digs puns. As a consequence, the peasants they prey upon become accustomed to editing their own memories (either that, or they become accustomed to the insides of gulags). But what I wonder about is those children the author mentions. How does a child edit memories, when he has so few to call his own, and so few points of reference outside of those? Did a monster gobble up poor Vladimír, or an angel fly him straight to heaven?
The reason this is on my mind is because, when dealing with the Covidian mobs of past three years, I’ve often felt as though I was dealing with young children. I witnessed the same histrionic meltdowns, the same devotion to irrational theories, the same ease with which they could be manipulated and controlled by those who abuse them. There came a time, right around mid-2021, when I realized there was no sense trying to reason with them. So, I just stopped talking. They weren’t my kids, after all.
As I’m sure most of my readers know, there are big movements afoot right now. Like shifting tectonic plates, those unaware of such movements are likely to be very surprised when they collide and — seemingly “out of the blue” — the earthquakes commence. Their infantilized minds will scream for some wise adult to put the picture back the way it was, to turn the hat back into a man. But their mouths will not scream these things. Or very few will, I predict.
Instead, what I think we’ll see is a year of quiet editing. Some of these editing projects will be fully conscious, as they scuttle about “cleaning up” their social media accounts. Other editing duties will be conducted unawares, in those sublingual realms where we tell ourselves stories about our lives the way we wished they happened. In these tales they will have known about the masks, about the lockdowns, about the shots, sure, sure, perfectly well and all along.
And no, they did not have those conversations with you, or they didn’t go down the way you claim they did. They never said that (and, if they did, certainly never believed it). Oh, and they did say that other thing many, many times, including to you. There never was a man in that photo. You’re just not remembering it right.
This will be a trying time for all of us. We will be angered by them, annoyed and frustrated, driven to our wit’s end by their fabrications, as they stand beside the broken cookie jar with chocolate smeared all over their hands.
But it is Christmastime. And so, even as we must reprimand them — and gain assurances that they will not behave this way again — we must also remember to love them.
As for those monsters who turn men into hats, that’s another story entirely.
And it will not end well for them.
P.S. If you found any of this valuable (and can spare any change), consider dropping a tip in the cup for ya boy. I’ll try to figure out something I can give you back. Thanks in advance.
Did you forget to black out the kill shot question when you repeated it?
Also, spot on with the retroactive editing. As it turns out, all the relatives I could still ask were in the resistance between 1933 and 1945 (only one arguably was in reality).
Hopefully Elon Musk will hire you as a Bot Slayer!