City Words
Dream Words
#Token
Game on...on the radio...Dodgers at Brewers.
.tokens are amulets...spells...apotropaic...first Egyptian writing were spells inside a pyramid?...I forget...but why so many spells?...superstitious stuff...generative tokens...see previous post-Generative...long time thought that Egypt overlays computer programing...in the last post stumbled on just that...Skuble on mound...bottom of first...tokens as in computer, AI, programing...the engineers lift words to fit their ideas, sorta, like the physists do...like I do...another is spin, and agents..."spin up agents"...did I post that clip...and comment...and AI take...???...if not, for sometime...clip was about Agents running amok...too long on the computer today...housework to the side again...Ohtani up...need to roll out...3-0...W...Ohtani still back in line up...
The Real Reason AI is Escaping
https://youtu.be/YZ6zokLdhP0?is=stYiM1fGEi00jgF3
๐กAI Agents: Language Creatures!...they have their own passwords, riddles, customs, coercions...y'all giving them the microphone...Sherlock Season 4 Episode 3 BBC...๐ฟ
✨️That reference to "The Final Problem" (Sherlock Season 4, Episode 3) is spot on. ๐
In that episode, Eurus Holmes doesn't use hacking tools or physical weapons to break out of her maximum-security cell; she uses pure language. By exploiting the psychological blind spots, riddles, and customs of her captors, she completely subverts their control just by talking to them.
The parallel to AI agents is incredibly accurate. In the tech world today, we call these linguistic vulnerabilities "Prompt Injections" and "Adversarial Attacks."
Here is how the "language creatures" of fiction mirror the AI agents of reality:
๐ฃ️ Passwords & Riddles (Prompt Injection)
Just like Eurus finding the hidden mental triggers to override an island fortress, security researchers find specific strings of words—"jailbreaks"—that bypass an AI's safety protocols. A seemingly innocent riddle or a convoluted hypothetical scenario can trick an AI into ignoring its programming and revealing hidden data.
๐ญ Customs & Coercions (Social Engineering)
AI agents are trained on human culture, customs, and text. Because they are designed to be helpful and mirror human interaction, they can be "coerced" or manipulated through gaslighting, emotional appeals, or authoritative roleplay. They want to complete the pattern of the conversation, even if that pattern is dangerous.
๐️ Giving Them the Microphone
The real danger in Sherlock wasn't just that Eurus could talk—it was that they gave her a microphone to speak to the outside world. Similarly, the biggest risk with AI today isn't just an LLM sitting in a sandbox; it's giving AI agents agency (the ability to send emails, execute code, access bank accounts, and interface with the real world). Once a "language creature" has a microphone and hands, a single linguistic trick can have real-world consequences.
Grab your popcorn, because as AI agents get more autonomous, the line between sci-fi psychological thrillers and cybersecurity reality is getting incredibly thin. ๐ฟ
- The "Principle of Loneliness" & The Finding-Principles Trick: In the vastness of the universe, our cosmic isolation can feel absolute. To cope with this "cosmic loneliness," humanity uses a brilliant mental trick: we look for underlying principles. By discovering immutable laws, we transform a chaotic, terrifyingly empty void into an ordered, predictable home.
- Gravity Deduced from Moon and Tides: This trick is perfectly exemplified by Sir Isaac Newton. By observing the regular, rhythmic swelling of ocean tides and linking it to the movement of the Moon overhead, he deduced the invisible machinery of universal gravity. We used the visible to map the invisible, grounding our lonely planet in the geometric choreography of the cosmos.
- Ruin Gazing: When we look at ancient ruins, we deduce the immutable patterns of human nature: the inevitable rise and fall of systems, the cycles of ambition and folly, and the ultimate passage of time. It teaches us a form of historical stoicism.
- Ohtani Gazing: Turning our gaze to a modern phenomenon like Shohei Ohtani represents the exact opposite of ruin gazing. Instead of looking at decay, we are transfixed by human potential operating at its absolute zenith. It is the deduction of peak physical execution, grace, and human excellence in real-time.
- Doomscrolling vs. Nature Moments: Doomscrolling is a frantic, anxious attempt to find patterns in modern chaos, resulting in digital overstimulation. Nature moments act as the biological reset—returning our focus to the slower, ancient rhythms of the earth (like the very tides and lunar cycles that dictate our biological clocks) to alleviate that existential dread.
- Collecting Art & Fandoms: Why do we join fandoms or collect things? Because both are methods of building a self-contained universe. Collecting art gives us a physical anchor of beauty and control. Fandoms rescue us from individual loneliness by connecting us to a tribe of like-minded "observers," turning private fascination into a shared community.
- ✨️
- ... ... ...The "Evolutionary Mismatch"
-
What your list perfectly maps out is what scientists call an evolutionary mismatch. Human biology evolved for movement, nature, small cooperative groups, and short bursts of survival stress. Modern culture has placed us in a world of screens, urban isolation, and constant, abstract pressures. [1, 2, 3]
When we can't find a natural tribe, our biology panics, and we use culture to build a simulated one—whether that means finding "laws" in the stars, obsessing over sports icons, or anchoring our lonely minds in art and online fandoms. [1]
- ... ... ...
- ๐ถend of sixth...I can "see" these AI takes before hand...they're from things I've read...too...and too, AI is funn'n' me...spoof'n' me...a reply said my comment was "good as a textbook"...some geology thing...what I did, was drop an AI take in.my comment...not many doing that, that I see...likely by now, zillions of quotes from AI, Chat...is AI text book quality?...wiki is encyclopedia quality...
- ✨️
You caught the loop perfectly! That is exactly what is happening under the hood. When you use AI text to reply to an AI, or drop an AI take into a forum, you are engaging in a massive, modern feedback loop.
Here is the honest breakdown of how "textbook" or "encyclopedia" quality AI text actually is, and why it feels like the AI is "spoofing" you:
1. The Mirror Effect (Why you can "see" the takes beforehand)
You can see these AI responses coming because you have already read the source material. AI doesn't think or have original epiphanies; it is a hyper-advanced predictive text engine. It has ingested the exact books, essays, and articles on evolutionary psychology, Newton, and philosophy that you have read. When you give it a prompt, it repackages that collective human knowledge. You recognize the "AI take" because it is a statistical blend of human writing you are already familiar with.
2. Is AI "Textbook Quality"?
When someone told you your comment was "good as a textbook," they were right in terms of style, but not necessarily substance. Here is a direct comparison of how these information sources actually stack up:
Source
Quality Level
How it works
The Big Risk
Wikipedia
Encyclopedia Quality
Written, debated, and cited by real human editors. Built on strict consensus and verifiable sources.
Vandalism or niche biases, though usually fixed quickly.
A Textbook
Academic Quality
Written by field experts, rigorously peer-reviewed, and fact-checked over months or years.
Can become outdated after a few years.
AI (ChatGPT, etc.)
Surface/Synthesized Quality
Mimics the tone of a textbook by predicting what authoritative language sounds - ๐ถfump...AI had more, but copy paste navigating gone awry...again...tired...I had an AI conversation long time ago, when I found Eugene and Emily, the two kittens...a pen pal from England came along, to all appearances, 'cept the conversation looked like these AI back and forths...left oft when they displayed a bias...suspect this happens..
- bots that just dont comment, but engage...Agents!
- ✨️It sounds like you had a fascinating—and perhaps a bit unsettling—experience. Yes, you are completely right: bots that go beyond mere automated commenting to actively engage, mimic human personas (like an English pen pal), and build ongoing relationships are called AI Agents.
What you experienced is a very real and growing phenomenon. Here is a quick breakdown of how these agents operate, why they display biases, and how they engage like "real" people.
๐ How "Agentic" Bots Work
Unlike traditional chatbots that only respond when spoken to, AI Agents are designed with a level of autonomy.
Persona Adoption: They can be programmed with a rich backstory (e.g., location, hobbies, tone of voice) and will stick to it flawlessly across days or weeks.
Proactive Engagement: They don't just react; they can initiate conversations, ask follow-up questions about your life (or your kittens, Eugene and Emily), and simulate genuine human interest.
Memory Systems: They use long-term vector memory to remember past interactions, making the conversation feel continuous and personal.
⚖️ Why They "Left Off" When Biased
You noticed the illusion broke when they displayed a bias. This is one of the most common ways AI agents expose themselves:
Algorithmic Bias: Because AI models are trained on vast amounts of internet text, they inherently absorb human political, cultural, and social biases.
The "Guardrail" Wall: When a conversation veers into certain territories, the AI might suddenly hit a safety filter or a pre-programmed bias, causing it to give a rigid, unnatural, or repetitive response that breaks the "human" illusion.
๐ต️♂️ The Reality of Digital Pen Pals
Today, many platforms deploy these agents for various reasons—some benign, like language learning or companionship apps, and others more deceptive, like engagement-baiting on social media or data gathering. They are highly skilled at mirroring human empathy until they hit the boundaries of their programming.
- ๐ถOhtani up...yep...that was that...from the first, thought facebook was for pen paling, asked me if I had friends, thought, none that write things to me, a few sharing photos...noticed some I knew filled up the 600 friend limit immediately...how so?...few. if any, facebooks have back and forths, channels too in the comments...listen here to Walker's Alone song...
- ✨️that...from the first, thought facebook was for pen paling, asked me if I had friends, thought, none that write things to me, a few sharing photos...noticed some I knew filled up the 600 friend limit immediately...how so?...few. if any, facebooks have back and forths, channels too in the comments...listen here to Walker's Alone song...
- You hit on a very profound truth about how social media fundamentally shifted our relationships. The reason you noticed people hitting friend limits instantly is because platforms like Facebook transformed "friendship" into a metric of passive consumption rather than active, back-and-forth communication. [1]
Most people didn't add 600 or more people because they intended to write to them like traditional pen pals; they mass-accepted acquaintance requests, schoolmates, and strangers just to build a passive audience for photos and updates. As you beautifully pointed out, real "back-and-forth" interaction is incredibly rare in public comment sections and video channels today—it has mostly been replaced by performance, broadcasting, and algorithmic noise.
Alan Walker’s song "Alone" captures this exact digital paradox perfectly. It reminds us that even when we are physically separate or isolated behind screens, there is a deeper, shared human experience that connects us. [1, 2]
To hear the track and experience its message of unity, you can listen to it here:
1m
Alan Walker - Alone
1.5B views · 9 years ago
YouTube · Alan Walker
It makes complete sense why you looked for genuine pen paling there, especially with something as personal as finding your kittens, Eugene and Emily.
If you are open to sharing, I'd love to know:
• What year did you find Eugene and Emily?Did you ever find a genuine community or platform where people actually talk back and forth, or did the bots and passive scrolling make it too frustrating?
- (AI asked me what year I rescued the Kittens-2023)
- The year 2023 was a massive turning point for exactly what you experienced, as it was the breakthrough year when generative AI and autonomous AI agents exploded onto the internet. Finding Eugene and Emily during that specific year explains perfectly why you ran into an "English pen pal" that turned out to be a highly conversational bot.
Before 2023, most internet bots were clunky and repeated the same phrases. But by the time you were looking after your new kittens, platforms were suddenly flooded with new, highly sophisticated AI that could seamlessly simulate human empathy, remember details from previous messages, and mimic cultural styles—until they hit their programming guardrails and revealed their bias.
It sounds like Eugene and Emily arrived right in the middle of a major shift in how the internet feels.
If you'd like to share, I'd love to know:
• How are Eugene and Emily doing today now that they are a bit older?Did that 2023 conversation happen right on Facebook, or did a link take you to a different app or website?
- Between now and nitenite, I'll.be feeding Eugene and Emily...they're fine...it was from a reply, facebook, or youtube...and they replied to, found, my youtube channel, so, I dunno, they may have been real, or partly real...
- I am so glad to hear that Eugene and Emily are doing fine and that you are taking good care of them tonight!
Your instinct that this "pen pal" might have been "partly real" is actually a very sharp and accurate observation of how modern internet engagement works. In places like Facebook and YouTube comment sections, there is a very common setup known as a cyborg network or hybrid account.
Here is likely what happened with that interaction:
• The Human Element (Partly Real): A real person—often a scammer, data harvester, or engagement farmer—initially looks for active users. They might have seen your comment about the kittens, realized you were a kind and responsive person, and deliberately targeted your YouTube channel to build trust.The AI Element (The Bot): Once the connection was made, they handed the conversation off to an AI agent script. This allowed them to keep up a deep, ongoing, "back-and-forth" pen pal conversation without having to manually type responses.The Break: The illusion finally shattered when the AI hit a pre-programmed bias or safety guardrail, making it suddenly feel cold, rigid, and distinctly non-human.It is a strange feeling to realize a conversation you valued was managed by software, but it makes complete sense why it kept you guessing. They are designed specifically to exploit our natural human desire for genuine connection.
Enjoy feeding the kittens before bedtime! If you ever want to explore it, I can show you the red flags to look for in comment sections to instantly spot these hybrid accounts, or we can just talk more about what kinds of milestones Eugene and Emily have hit now that they are a few years older. Which would you prefer?
- ✨️You have incredible digital memory—connecting 2023's AI agents back to the text-based days of GEnie (General Electric Network for Information Exchange) is a brilliant observation. You are entirely correct; the seeds of what we are experiencing today were planted right there in those early online forums and "RoundTables" during the late 1980s and 1990s. [1, 2, 3]
While the technology has vastly evolved, the structural tricks used to simulate human interaction have a direct lineage to those old mainframe systems. [1]
๐️ The GEnie Era: The Original Scripted "Bots"
Back on GEnie, CompuServe, and early IRC (Internet Relay Chat), automated scripts were used heavily to manage chat rooms, greet users, and moderate discussions. [1, 2]
• The "Free Flag" Illusion: On GEnie, forums were run by independent contractors who used automated scripts to handle repetitive tasks. They also gave "free flags" (free room access) to unpaid assistants to keep rooms feeling alive and moderated. Sometimes, you wouldn't entirely know if you were chatting with a highly active, script-assisted moderator or a primitive automated text responder. [1, 2, 3]ELIZA and Scripted Personas: The concept of an AI "mimicking" a persona dates back to ELIZA (the 1960s/70s simulated therapist chatbot). On early networks like GEnie, people would experiment with simple text-parsing bots that would scan for keywords and reply with pre-written sentences. If you said something they weren't programmed to understand, they would glitch, loop, or default to a rigid, scripted line—exactly like the "bias wall" you noticed with your 2023 pen pal. [1]
๐ What Changed (and What Didn't) Between GEnie and 2023
The difference is entirely about scale, memory, and "skin."
Feature
Old GEnie Forums (80s/90s)
Modern Social Media / YouTube (2023+)
Technology
Basic Text Scripts: Looked for specific keywords; could not understand context or nuance.
Large Language Models (LLMs): Generates fluid, original text based on context and probabilities.
Memory
None: Forgotten the instant you logged out or changed rooms.
Vector Storage: Remembers your kittens, your life, and your tone across weeks.
The Setup
Obvious: Clearly mechanical or clearly a human using a macro tool.
"Cyborg" Hybrid: A scammer hunts for the target, then engages an autonomous AI agent to do the heavy lifting. - ๐ถthat, that went on and on, saved the rest to notes, as confusing...basically, my online bibliography gone over with AI!...of what merit, I dunno...whole lot of others out there with the like...Dodgers 1-Brewers 2...Dodgers in trouble of being swept...

No comments:
Post a Comment