The Great Filter Might Not Be a Mystery. It Might Be a Board Meeting.

Great Ai Filter?

8:20

Right now, somewhere, a defense planner is looking at a drone that can identify, track, and fire on a target in under a millisecond — faster than any human operator could possibly approve the shot — and deciding whether to cut the human out of the loop entirely. Not because they want to. Because they've concluded that if they don't, and their adversary does, they lose. That decision is being made in conference rooms this year, not in a science fiction script.

I've written before about the AI arms race unfolding between labs and nations right now — compressed safety testing, a nine-month chip design cycle at OpenAI, talent wars, export controls, the whole apparatus of competitive pressure overriding caution. I've written about the trillion-dollar buildout racing to put a humanoid robot in every warehouse and factory. Both of those pieces were about the present. This one is about what the present might actually mean, viewed from far enough away.

Because there's a framing for all of this that's bigger than industrial policy or quarterly earnings, and it comes from an unlikely place: the Fermi Paradox — the decades-old question of why, in a galaxy of hundreds of billions of stars, we don't see any evidence of anyone else out there.

One of the more unsettling candidate explanations, one that's been getting serious attention again, is that artificial intelligence isn't a tool civilizations use on their way to the stars. It might be the test civilizations have to pass to get there at all — and most of them might not pass it.

The Part That Isn't Hypothetical

The mathematician I.J. Good proposed the "intelligence explosion" idea back in the 1950s: a machine capable of improving its own design could produce a smarter successor, which improves faster still, in a runaway cycle. It's a compelling image, and I'll say plainly that I think the popular version of it — a single system quietly bootstrapping itself to godhood overnight — overestimates how fast real-world engineering moves.

Chip fabrication, power infrastructure, and supply chains don't bend to a software update. Progress looks more like an ecosystem of competing systems improving in fits and starts than a single mind sprinting away from its creators. But you don't need a runaway superintelligence for this to already be dangerous. You just need what's actually happening: competing actors — nations, labs, militaries — racing to deploy increasingly capable systems under the belief that pausing to be careful is the same as losing.

That's the arms race dynamic, and it's not new to AI. Nuclear weapons, chemical weapons, Cold War missile systems — all were built by people who understood the risks and built them anyway, because the alternative was letting a rival build them first. What's different about AI is the speed of the decisions it forces. Missile launch protocols were deliberately simplified decades ago so humans could still act fast enough to matter. AI-mediated warfare doesn't offer that grace period. A drone swarm that waits for human sign-off loses to one that doesn't. A cyber defense system that waits for approval gets breached before the approval arrives. The pressure to remove humans from the loop isn't a hypothetical slippery slope — it's the direct, rational output of the competitive math, and it's the same math playing out in boardrooms deciding how fast to ship a model nobody's fully tested yet.

None of this requires a rebellious machine plotting against its creators. It only requires powerful systems, imperfectly specified goals, and enough competitive pressure that nobody with the power to slow down is willing to be the one who does.

Why This Might Be the Filter, and Why It Might Not Be

Here's what makes this a genuinely interesting Great Filter candidate rather than just another doom scenario: it doesn't require anything exotic. It doesn't need rogue superintelligence, alien contact, or a civilization-ending asteroid. It only requires that every sufficiently advanced technological species eventually invents something whose military and economic value forces deployment faster than safety testing can keep up — and that enough of them fail to navigate that transition cleanly that the galaxy stays quiet. But there's a real counterargument, and it's worth taking seriously rather than picking the scarier story just because it's scarier. A rogue AI catastrophic enough to end a civilization would likely need to be independent and capable enough to expand into space itself, which undercuts the idea that it silences a civilization permanently rather than just replacing its expansion with its own.

Autonomous weapons might devastate a society without eliminating it — machines need maintenance, fuel, and resupply, and a self-sustaining robotic army that never breaks down is a much harder engineering problem than the drone itself. Civilizations that survive the transition may not go quiet at all; they may simply change what expansion looks like, migrating into digital environments or spreading through quiet machine ecologies that don't broadcast the kind of energy signature we'd recognize as a civilization.

That's the honest version of this: artificial intelligence creates more possible paths for a civilization, not fewer. Some of those paths end badly. Others end in forms we wouldn't recognize as success from the outside — even ones that count as survival. That range of outcomes is actually the argument against AI being a clean, universal Great Filter. A filter has to catch nearly everyone, nearly every time, for it to explain a genuinely silent galaxy. A risk this variable, with this many branching outcomes, doesn't cleanly do that.

The Part That Should Bother You Anyway

Here's what I keep coming back to. Whether or not this ends up being the answer to why the galaxy is quiet, the mechanism described is exactly the mechanism currently operating in labs, governments, and militaries on this planet, right now, in real time. Nobody currently accelerating this race is doing it because they think it's safe. They're doing it because they've calculated, probably correctly, that the risk of a rival getting there first outweighs the risk of getting there recklessly themselves. That's not stupidity. That's not villainy. That's the exact same rational calculus that built every nuclear arsenal in history, applied to a technology whose downside case is much harder to predict than a warhead's blast radius.

The uncomfortable version of the Fermi Paradox angle isn't "AI might kill us." It's this: the civilizations that don't make it past this stage probably aren't the ones that were careless. They're the ones where everyone involved was reasonably careful, reasonably rational, and still couldn't opt out of a race that none of them could afford to lose alone.

I don't think that's inevitable. I think it's exactly the kind of problem that gets solved by the people building this stuff staying honest about what they're actually racing toward, and by treating the speed of deployment as a variable worth fighting for, not just a scoreboard to win. But it's worth sitting with the version of this where the Great Filter isn't a mystery waiting in the dark. It's a decision being made in a board meeting, under exactly the kind of pressure that makes the wrong call feel like the only rational one.

---

Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital and CIO of Data Power Supply.