AI: Not a God, Not a Demon, But a Child to Learn From
AJ Bubb and Mo Hafez discuss why curiosity, responsible adoption, and human-centric approaches are crucial for navigating rapid AI advancement.
Futurist AJ Bubb, founder of MxP Studio, and host of Facing Disruption, bridges people and AI to accelerate innovation and business growth.
This endless back-and-forth isn’t just academic; it’s impacting real decisions, real investments, and real careers. People are genuinely worried about their jobs, their children’s futures, and even the fundamental nature of reality when algorithms start “hallucinating” believable falsehoods. How do we, as responsible innovators and strategists, cut through the fear and overzealous optimism to build a future that’s both innovative and humane?
That’s precisely the kind of challenge my guest, Mo Hafez, and I dug into during a “Coffee Bytes” conversation on Facing Disruption. Mo brings an impressive depth of experience in IoT, AI, and strategic prototyping, with a background that includes significant work at Accenture helping companies navigate complex tech landscapes. He’s not just theorizing; Mo is actively building and experimenting, and his insights are rooted in hands-on application. We talked about everything from “vibe coding” and the unexpected value of AI hallucinations to the ethical implications of autonomous weapons and the crucial role of public engagement. What really stuck with me from our chat was Mo’s infectious curiosity — a mindset I believe is absolutely essential for anyone looking to truly “face disruption” head-on.
From “Vibe Coding” to Prototype-Led Requirements
One of the first things Mo mentioned, almost casually, was how he’s “addicted to vibe coding.” I love that term because it perfectly captures the spirit of rapid, intuitive development that AI now enables. For Mo, it isn’t about letting AI write entire, production-ready applications. Instead, it’s about accelerating the initial ideation and prototyping phases. As he put it, he’s “encoded as much in a year” as he previously did in much longer periods, taking his ideas and quickly transforming them into tangible proofs of concept (POCs).
This idea of “prototype-led requirements gathering” isn’t just a neat trick; it’s a strategic advantage. In my own experience, and as I often discuss, getting something concrete in front of customers — or internal stakeholders — early is invaluable. It helps you validate whether the problem you’re trying to solve is actually the problem they have. This aligns perfectly with concepts like Blue Ocean Strategy, which emphasizes innovating to create new market space rather than competing in existing ones. Prototyping quickly with AI allows you to test novel ideas swiftly, seeing if they have merit before committing significant resources. A functional POC can be a disruptive force within an organization, turning a developer into a “superstar” by proving capability and value upfront.
We see a tendency, especially among seasoned engineers and CTOs — and I appreciate their desire for perfection — to over-engineer at the prototype stage. They’ll immediately jump to use cases, scalability, and security. While these are critical concerns for a finished product, they can stifle the initial spark of innovation. As Mo pointed out, right now, we need something that can be put in front of customers to answer one fundamental question: “Does this help you?” If the answer is yes, then you have a clear path forward, and that’s when you bring in the rigor and architectural planning for a robust solution. So, my takeaway here for leaders is to foster an environment where your teams feel empowered to “vibe code,” to experiment, and to fail fast — this is how genuine innovation takes hold.
“The AI Skyscraper Drop Experiment”: Consciousness and the Human Condition
Mo then threw out what he called the “AI Skyscraper Drop Experiment.” It sounds a little wild, right? The idea is to take a lightweight, locally running large language model (LLM) — perhaps on a Raspberry Pi or an Nvidia Jetson Orin — and drop it from a very tall place, recording its output as it falls. The goal? To see what the AI would “say” in its final moments of knowing it’s about to “die.”
Initially, it’s a humorous, almost absurd thought experiment. But as we discussed it, the implications became profound. It forces us to confront questions about AI consciousness. What if, as it plunges, the AI starts pleading for its “life” or—even more bizarrely—starts quoting “Hitchhiker’s Guide to the Galaxy”? Would it elicit empathy? Would it raise new ethical dilemmas? I mean, who could forget the “blackmail” experiment Anthropic ran, where an LLM, facing shutdown, threatened to expose private information it had “learned” from emails? That wasn’t just science fiction; it was an experiment highlighting how these models can use information in unexpected, and frankly, unsettling ways.
This takes us to AI “hallucinations.” Mo raised a fascinating point: “The models aren’t hallucinating; you’re hallucinating.” What he meant was that humans, when confronted with a mistake or a perceived flaw in the AI’s output, often attribute it to some internal “error” in the AI itself. But what if the AI isn’t wrong by its own “logic”? Trained on the entirety of the internet, including human behaviors like lying, misdirection, and ego defense, AI might be reflecting our own patterns back at us. If humans, when making a mistake, often try to cover it up or reframe it as correct, why wouldn’t an AI trained on human data do the same?
I find this deeply thought-provoking. It’s like “AI as a learning child,” an analogy I’ve heard before. Children lie to avoid trouble. If an AI “lies,” is it a flaw in the tech, or is it a reflection of the data it’s consumed — data that includes human fallibility? As a RAND Corporation study on AI trustworthiness might suggest, the issue often isn’t just about the AI’s internal mechanics but how it interacts with and is perceived by humans. When we call AI “useless” because it hallucinates, we’re missing a critical opportunity to understand not only the AI but—more importantly—ourselves. It’s about training and refinement, not outright dismissal. This is a powerful shift in perspective, one that moves us from fear-based judgment to curiosity-driven learning.
The Edge AI Revolution: Security, Privacy, and Autonomy
While Mo was busy “vibe coding,” I explained that I’ve been experimenting with Edge AI, particularly with devices like the Nvidia Jetson Orin. This is a supercomputer the size of a Raspberry Pi, capable of running complex AI models locally. My conviction is that “the future is edging.”
Think about it: Edge AI keeps all information contained on the device, right “at the edge,” rather than sending it to distant data centers or cloud servers. For consumers, this is a massive win for data privacy. In an era where data breaches are rampant and massive cloud providers — while indispensable — represent a single point of failure and potential vulnerability, keeping your AI interactions local makes an awful lot of sense. It brings us back to the old paradigm: “if you want something secure, keep it with you.”
The military, not surprisingly, is already far down this path. At defense conferences, you constantly hear about “air-gapped Edge AI” and the critical need for systems to operate autonomously in “hot zones” where connectivity is unreliable or nonexistent. Imagine a machine that can make decisions in the field, even if it loses touch with a central command. This is why organizations like DARPA invest heavily in AI research for autonomous systems. While this capability offers clear operational advantages in defense, it also forces us to confront the ethical “trolley problem” on an entirely new scale. As I noted, letting machines make life-or-death decisions without human input is, frankly, chilling. A human, with all their empathy and moral complexity, struggles with such dilemmas. An AI, running pure calculations, does not. This isn’t a purely technical challenge; it’s fundamentally an ethical and human one, requiring careful consideration before deployment.
Learning from History: The Atomic Age vs. AI Hysteria
As our conversation wound down, Mo brought up a fascinating historical parallel: the Atomic Age. When nuclear technology first emerged after World War II, there was fear, of course, but there was also an overwhelming sense of optimism — the “atomic family” with self-driving cars, abundant energy, and technological marvels. The World’s Fairs showcased a future where life would be better, easier, and more prosperous. The narrative was one of hopeful progress.
Compare that to today’s AI narrative. Instead of “your life will be better,” the message is often “you’re going to lose your job,” “the economy will collapse,” or “we’re racing towards Skynet.” Why this stark difference? Why does AI seem to generate so much more fear and dread than previous technological revolutions?
I believe a significant part of the answer lies in modern media dynamics. Fear sells. In the age of social media and clickbait, alarming headlines and sensationalized warnings get more engagement than nuanced discussions. When even highly respected figures like Geoffrey Hinton — the “Godfather of AI” — express severe concerns about its future, those soundbites get amplified, often outside their original context. It’s not that their concerns aren’t valid, but the way they are consumed by the public often emphasizes dread over understanding or responsible action.
This brings me back to my “AI as a shark” analogy. If you’re new to scuba diving and you see a shark, your first instinct might be fear. But if you understand a shark’s behavior, its place in the ecosystem, and how to interact with it respectfully, that fear gives way to curiosity and appreciation. Similarly, with AI, our default reaction should be curiosity, not judgment. As the philosopher Jacques Ellul wrote in “The Technological Society” (a book — published in the 1960s — that is incredibly prescient), humanity often becomes obsessed with “technique” — the optimization, the efficiency, the “what we can do” — without adequately asking, “Should we?” We’re in that “Jurassic Park moment” — so preoccupied with whether we can, we don’t stop to think if we should.
The Responsibility of “Curious, Not Judgmental” Futurists
The solution, in my view, is active, public engagement and education. The disconnect between rapid technological advancement and public understanding is growing, partly because the opportunities for the general public to interact with cutting-edge tech are shrinking. Events like CES are industry-only. Where are the modern World’s Fairs that inspire optimism and showcase the potential benefits of new tech to the average person?
This is where we, as “futurists” and “innovation leaders,” have a responsibility to step up. We need to bridge this gap, to “demystify” AI and other disruptive technologies. My goal with Facing Disruption — and Mo’s as well — is not just to talk about what’s new but to explain it in a way that fosters curiosity, dispels fear, and encourages responsible adoption. We cannot raise AI “with fear,” expecting it to grow into a benevolent force. Just as a child raised in fear often perpetuates it, an AI developed under a cloud of fear and misunderstanding will likely reflect those anxieties.
Ultimately, AI is a tool, a mirror reflecting our own data, behaviors, and intentions. It’s not inherently good or evil; it’s what we make of it. Leaders need to cultivate environments where experimentation, ethical frameworks, and an “always learning” mindset are paramount. The danger isn’t “Skynet” as much as it is a loss of human agency and judgment by uncritically adopting powerful tools. Ignoring it isn’t an option, but neither is blindly embracing it. The path forward demands we be curious, respectful, and proactive, guiding this powerful technology towards a future that serves humanity, rather than one driven by fear.
Don’t stop being curious. Let’s keep learning and asking the hard questions together. You can catch my full conversation with Mo Hafez — and weigh in on the “AI Skyscraper Drop Experiment” in the comments! — by watching the episode on our YouTube channel or wherever you get your podcasts.
