I cover artificial intelligence for a living. I read the terrifying essays about superhuman AI. I know the arguments from people who believe machines will one day replace us all. So when I sat down to watch The AI Doc: Or How I Became an Apocaloptimist on Prime Video, I expected to nod along with familiar warnings. Instead, the documentary left me with something far more unsettling than fear of robot overlords. It left me with uncertainty. And one statistic in particular made my stomach drop: an ai safety imbalance that is almost impossible to ignore.
The Workforce Gap That Should Keep Us Up at Night
According to the documentary, more than 20,000 people around the globe are currently working on artificial general intelligence — the kind of AI that can match or exceed human cognitive ability across any task. Meanwhile, fewer than 200 people are focused specifically on making sure that AI is safe. That ratio — 100 to 1 — is the kind of number you hear and immediately forget because it feels too absurd to be true. But it is true.
Think about it. We are pouring resources into building something powerful while barely staffing the department that checks the brakes. If you were constructing a skyscraper and hired 20,000 architects but only 200 safety inspectors, would you feel confident walking into that lobby? That is exactly where we stand with AI today.
The documentary does not frame this as a conspiracy. The people building AGI are not villains. Many genuinely believe their work will cure diseases, solve climate problems, and transform education. But the ai safety imbalance means those optimistic builders vastly outnumber the people asking hard questions about what happens when the system goes wrong.
Reason 1: The Scarcity of Safety Researchers
Let me put a finer point on it. Two hundred people is a small team in a medium-sized company. OpenAI alone employs thousands. DeepMind has hundreds of researchers. Meta, Google, Microsoft — each pours enormous budgets into advancing AI capabilities. But the entire field of AI safety, as a dedicated discipline, operates on what amounts to a shoestring staff.
The documentary highlights interviews with safety researchers who describe working in a kind of isolation. They publish papers. They attend tiny conferences. Meanwhile, the capability side of the field throws lavish parties. The result is that safety questions — like how to align an AI with human values, how to prevent unintended behaviors, how to test for catastrophic risks — advance at a snail’s pace compared to the raw power of the models themselves.
This is not a judgment call. It is a numerical reality. And it is the first reason the documentary scared me. Because if you believe even a small chance exists that advanced AI could be dangerous, then a workforce of 200 people feels like we are flying blind.
The Political Reversal That Removed Guardrails
Just days before I watched the documentary, President Trump scrapped the executive order that had required additional AI safeguards. That order was not perfect. It was a first step. But removing it sent a clear signal: the regulatory brakes are off.
The documentary explores how AI regulation has always lagged behind development. But the political climate makes it worse. One administration puts up a fence. The next tears it down. Meanwhile, the models keep getting bigger, faster, and more integrated into our daily lives.
Reason 2: Regulation Moves at Government Speed, AI Moves at Silicon Valley Speed
The documentary draws a direct parallel to the early days of social media. Platforms like Facebook and YouTube exploded globally before anyone understood what misinformation, algorithmic radicalization, or mental health erosion looked like at scale. By the time governments started reacting, these platforms were already woven into the fabric of everyday life. You cannot unring that bell.
AI is moving much faster. Social media took about a decade to reach mass adoption. ChatGPT reached 100 million users in two months. The pace is not just fast — it is accelerating. Every few weeks, a new model demonstrates capabilities that would have seemed science fiction a year earlier.
And yet, regulation proceeds at the speed of committee meetings, public comment periods, and legislative cycles. By the time a law passes, the technology it aims to govern already looks antique. This timing mismatch is the second reason the documentary shook me. We are effectively building the plane while flying it, and the manual is being written by people who are not even on the plane.
The Uncertainty That Is Worse Than Doom
I could handle a documentary that predicted a clear dystopia. That is easy to process — you know what to fear, and you can argue against it. What unsettled me was the documentary’s refusal to give a tidy answer. The filmmakers call it “apocaloptimism.” The word captures the tension perfectly: we can simultaneously believe AI might save us and might destroy us, and we cannot resolve that contradiction yet.
Reason 3: No One Knows Where the Ceiling Is
The documentary interviews developers, ethicists, and everyday users. No one agrees on the timeline. Some say AGI is five years away. Others say thirty. A few say it will never happen. But the uncertainty itself is the problem. Because if you do not know how close you are to the edge, you cannot know how much caution is appropriate.
You may also enjoy reading: 5 Features Claude Pro Needs to Be Close to Perfect.
Imagine driving a car at night on a winding road. You cannot see the curves ahead. You do not know if there is a cliff. But you keep accelerating because the engine is powerful and you are curious how fast you can go. That is where we are with AI. The documentary does not need to show a terminator-style apocalypse to be scary. The lack of a speed limit is enough.
The filmmakers, Daniel Roher and Charlie Tyrell, are both expecting children. Watching them grapple with the question of what kind of world their kids will inherit made the uncertainty feel personal. They are not detached observers. They are parents-to-be asking: will AI make life better for my child, or will it create a world my child cannot navigate safely?
Reason 4: We Are Already Deeply Embedded
The documentary forced me to look at my own habits. I use ChatGPT for research. I use AI for brainstorming, organizing my calendar, and even drafting emails. Millions of people do the same. The integration feels natural, even trivial. But the documentary points out that we are handing over decision-making and cognitive work to systems we do not fully understand.
When you rely on a large language model to summarize a medical study or suggest a financial move, you are trusting that the model’s training data, architecture, and guardrails are sufficient. But those models are black boxes. Even their creators cannot always explain why a particular output emerges. So we are outsourcing judgments to inscrutable systems, and we barely notice.
The documentary shows how this applies to education (students using AI to write essays), to art (AI-generated content flooding creative fields), and to relationships (AI companions designed to form emotional bonds). Each use case seems small. But together, they represent a fundamental shift in how humans interact with the world. And we are not pausing to examine that shift. We are just rolling with it.
The Incentive Problem No One Wants to Solve
Why are companies racing so fast? Because the incentives are enormous. Whoever builds the most capable AI first could dominate search, software, healthcare, media, and entire industries. The documentary does not demonize companies. It simply observes that market forces reward speed, not caution.
Reason 5: The Race Has No Finish Line
Consider the competitive dynamic. If one company pauses to focus on safety, another company will speed ahead and capture the market. The first company loses. So no one pauses. This is the classic prisoners’ dilemma applied to technology. Everyone would be safer if everyone slowed down. But no individual player can afford to be the one who hits the brakes.
The documentary explores this through interviews with people inside the industry. They talk about the pressure to release products, the fear of being left behind, and the belief that if they do not build the next big model, someone else will. The result is a collective action problem that no single company or government can solve alone.
That is the fifth reason the documentary scared me. Not because AI is malevolent. But because the system of incentives pushing us forward is stronger than the system of caution trying to hold us back. And that ai safety imbalance — between builders and safety researchers, between speed and regulation, between optimism and uncertainty — is not going to correct itself.
