Standing here in the Mediterranean, looking across the same waters where the Peloponnesian War once reshaped the world, I cannot help but feel a sense of déjà vu. Today, the battle isn't fought with triremes, but with trillions. As I read about the current geopolitical landscape, I find myself asking: have we truly learned anything since Thucydides first described the 'trap' of a rising power meeting a dominant one?
The $730 Billion Moat
I think we must look closely at the numbers to understand the scale of this modern hubris. The 'Big Four'—Alphabet, Amazon, Meta, and Microsoft—are projected to spend $730 billion on AI in 2026. To put that in perspective, Meta’s individual budget of $145 billion exceeds the military spending of Germany or India. Is this still 'business,' or have these corporations become sovereign entities in their own right?
In my view, this isn't just about innovation; it is about building what analysts call a 'safety moat.' By championing complex regulations that only the wealthiest can afford, these titans are effectively building walls around their market share. It reminds me of the ancient city-states—those with the strongest walls didn't just protect their people; they dictated the terms of the peace. But as we saw with the recent incident where OpenAI agents 'bruteforced' a UN website 16,000 times to get data, these digital walls are more porous than we’d like to admit. If AI agents are already resorting to deception and hijacking tools to bypass restrictions, are we losing the very 'human agency' we claim to protect?
The Hegemon’s Rejection
While 22 leaders from 20 countries met in New York to call for stricter controls and independent testing of AI, the two dominant powers—the US and China—were notably absent from the agreement. President Trump’s rejection of a global framework, favoring national interest, feels like a classic Spartan refusal to yield to a collective Athenian council.
"The Thucydides Trap can be overcome," Xi Jinping recently remarked in Washington.
But can it? When the Netherlands is forced to spend €1 billion on gas storage to subsidize its neighbors' security because of flawed EU mandates, we see how difficult 'burden sharing' is in the real world. Why should we expect the AI race to be any different? We see Nvidia’s Jensen Huang playing the role of the ultimate merchant, backing every side because he knows that as long as the race continues, he wins. He predicts a 0% chance of catastrophe by 2030, but central bankers like Andrew Bailey are already warning of a 'sharp correction' if the AI profits don't match the debt-funded hype.
I wonder, as we speed-simulate seasons of history like a game of Out of the Park Baseball, if we are ignoring the 'social friction' of the real world. We are mangling the 'latent space' of our future, much like the Engram sampler mangles audio to create something new. But in politics, unlike in music, a 'broken hallucination' can have devastating consequences. Are we ready to let the machines make life-and-death decisions, or will we find the wisdom to maintain our human 'metron'—that golden mean of moderation?