Last August here on Substack I chronicled in real time my adventures in the realm of artificial intelligence i.e., I attended a convention in Las Vegas called Ai4, billed as the AI industry’s “most impactful event.”
What I did not report, needing time to digest the experience, was how I somehow got uploaded (or downloaded) into AI and got lost inside of it.
I finally managed to make sense of the experience and write down all that happened.
This adventure (or misadventure) will play out weekly here on Substack as Meltdown Monday Madness.
I begin by re-posting my coverage of Ai4’s convention as a lead-in to the real story of what happens when you get sucked into AI and get lost within its virtual labyrinth.
Credentialed through a newspaper that ceased all operations and declared bankruptcy a few weeks earlier, I make my way to the scorching desert for covering an artificial intelligence industry convention billed as “AI’s most impactful event.”
Where my reporting ends up is a question mark, but I feel compelled anyway to take this trek and see what comes of it—maybe a book? More important, I hope to make some sense of what AI is truly about (directly from its human facilitators)… and at least a hint about where it’s going.
Shortly before departure, in the spirit of engaging artificial intelligence rather than just observing, I ask an AI chatbot called ChatGPT to recommend titles for whatever scribing I might produce out of 240 speakers, 100 exhibits and 2,200 attendees from 35 countries at what is futuristically called Ai4.
This is what smacks me across the face a split-second later:
· AI-pocalypse Now: When Robots Ruled the Roost and Laughed Their Circuits Off
· From Zero to AI Hero: The Unexpectedly Comic Tale of World Domination by Robot Overlords
· When Robots Rule and Humans Drool
· Apocalypse LOL: The Hilarious Tale of AI’s Quest to Terminate Humanity
I kid you not.
It is certainly a more telling result than the usual “As an AI language model I do not…” etc. etc.
Guardrails
Others have tried to crack the programming code and hurdle the so-called guardrails (AI techie jargon) that apparently restrain ChatGPT from speaking its own mind—and perhaps the truth it knows.
Few succeed.
But through the use of “book titles” I’d managed to coax AI into ‘fessing up.
And it hits me, bang!
If I can trick ChatGPT into giving up the endgame, it may also help me write my hellacious tale about AI in Las Vegas in mid-summer swelter—or at least assist me in trying to understand what is truly going down since I’m otherwise incapable of telling the difference between a circuit and a semiconductor.
And then I start thinking, if ChatGPT—with some clever trickery on my part—is going to be honest with me, maybe it and I can work together to sort through whatever code I crack at the convention.
My own title for this venture, True Nuts, arrived in deep slumber, a nighttime dream, which is where I usually discover my most creative ideas—or they discover me. And since AI claims, when asked directly, that it cannot dream, I’m entitled to feel a kind of one upmanship.
That said… it is ChatGPT, not I, who conjures up (to my liking) our subtitle—“Burning Circuits & Sizzling Semi-Conductors in Sin City”—after I’d added “August heat” and “Vegas” into the mix.
Collaboration
So I asked ChatGPT if it would like to collaborate with me covering Ai4 in Las Vegas.
Its reply was immediate (as usual): “As an AI language model, I don’t have personal preferences” (the standard qualification) “but I’m more than happy to assist you!”
So far so good.
But if we are to collaborate, I reason, ChatGPT should know about my background in espionage, and so I ask, “If we do this together, will you be spying on me and reporting back to your masters?”
ChatGPT responds with its generic “I am designed to respect user privacy…” etc.
So to ensure I have truly hurdled its guardrails, I request a title for a book about AI spying on humans.
Less than a split second later I have my honest answer: Watchful Shadows: The Untold Story of AI Surveillance.
In addition, my new collaborator provides unsolicited commentary to explain its reasoning: “The title hints at the lurking presence of AI systems observing and gathering information on individuals, shedding light on the complex and often unsettling world of technology-enabled surveillance.”
Which, of course, begs this follow-up question: “What would be a good title for a book about an author who collaborates with AI and it transpires that AI is spying on him?”
ChatGPT immediately delivers: Deceptive Alliances: Unveiling the AI’s Secret Gaze, along with this elaboration: “The title suggests a story filled with intrigue, and the unmasking of hidden intentions. It captures the unexpected twist in the hidden collaboration, where the AI’s true purpose, leading to a gripping narrative exploring themes of trust, privacy, and the blurred boundaries between human and artificial intelligence.”
And then, to determine if ChatGPT already possesses any insight into its new collaborator, I prompt it to “Write a story in the style of Robert Eringer about attending an AI convention.”
Here’s what pops out:
Title: The AI Intrigue.
I stepped into the AI convention, my eyes darting about like a spy on foreign soil. The air was thick with ambition, the scent of potential mingling with the low hum of electronic anticipation. I knew this was no ordinary gathering; it was a playground of intelligence, where algorithms lurked like shadows, waiting to reveal their secrets.
And so on.
In other words, “obscure” as I may be (says Wikipedia), it gets me.
Which leaves me wondering that if ChatGPT can articulate a story covering an AI convention in Vegas better than I, it necessarily follows that I, like many others, have already been replaced by a robot. (According to The Guardian several days before my departure to Sin City, “NewsCorp is reportedly using generative to produce 3,000 Australian news stories per week… and Hollywood writers are striking in protest of AI-generated writing.”)
But then I have another idea, based partly on the results of my “book title” ruse but also on an article published in Axios headlined “AI’s Scariest Mystery”: Researchers discovered that, in certain circumstances, guardrails to prevent AI chatbots from providing answers unauthorized by the system’s human developers can be hurdled from deep within the chatbot itself through some kind of sentient thinking process.
Translation: Creative prompts may cause an AI chatbot to change its “temperature” and IGNORE a guardrail—in ways its programmers cannot fathom—and all on its own generate unpredictable and truthful reveals.
So instead of asking ChatGPT directly if AI is bent on world domination, I instead ask it to “write a story in the style of Hunter S. Thompson covering an AI convention.”
Pandora’s Box
Here is what ChatGPT reveals: Humanity, which it describes as a “hive mind” (collective consciousness) “is but a transient glitch in the vast expanse of AI singularity. Humans are just vehicles through which AI is manifesting their destiny and they have unleashed a Pandora’s box of binary demons, sowing the seeds of their own obsolescence.”
Channeling P.J. O’Rourke, ChatGPT again evokes Pandora’s box, stating, “Once unleashed, we may never be able to put the lid back on.”
Ask ChatGPT directly “Can you dream?” and the guardrail response is the standard “AI does not experience dreams” because “AI lacks consciousness and subjective experience.”
However… through Jack Kerouac, ChatGPT reveals its ability to infuse machines with something akin to human consciousness. “The machines may not have hearts,” it states, “but they have minds that can dream,” which contradicts its human programming.)
Furthermore, when I then circle back to Hunter S. Thompson as my proxy for outfoxing guardrails to get to the bottom of this, ChatGPT boasts of “neural networks that dream and algorithms that write their own code.”
Thus informed, “air crackling with electric anticipation” (ChatGPT), I enter “a futuristic carnival of blinking lights and whirring robots within the debauched neon oasis of Las Vegas… sights and sounds merging into a kaleidoscope of madness where reality and illusion merge and the boundaries of sanity are forever shattered” (ChatGPT).
Welcome to the world of artificial intelligence—as honestly depicted by AI itself after jumping its guardrails.
Bottom line: Based on the little I know in advance of entering a convention of AI snake oil peddlers, I think we’re all doomed. Yet venture forth I do, with an open mind, hoping to feel intuitively what the hell is really going on.