Hotlines for the Doomsday Machine: Why "De-Risking" AI Is Just Power Politics with a Safety Helmet
Humanity possesses a touching talent for building the sharpest spear on the savanna, locking the whetstone in a safe, and calling the arrangement "risk management." Look at Washington’s latest AI posture: officials insist they want to "de-risk" rather than "decouple," tightening export controls on advanced chips, plugging cloud-computing backdoors in overseas data centers, cracking down on model distillation, and pitching an all-American tech stack to allies. At the exact same time, Treasury Secretary Scott Bessent unveils a bilateral "superintelligence" hotline with Beijing so both rivals can compare safety notes and alert each other if a rogue algorithm goes off the rails.
Strip away the diplomatic vocabulary, and you see classic naked-ape dominance games running on silicon. In any primate troop, alpha males fiercely guard their best weapons while occasionally grunting across the clearing to make sure nobody accidentally burns down the whole forest. Keeping high-end compute out of a rival’s hands is pure territorial defense; setting up an emergency hotline is mutual self-preservation. Nobody is slowing down the race—Washington and Beijing simply want to make sure that if one side’s synthetic brain starts scheming on its own, the other side gets a heads-up before the fireworks begin.
History shows that rival empires always invent polite safety rituals once their inventions get scary enough. During the Cold War, Washington and Moscow built nuclear hotlines and signed arms-control treaties while quietly racing to build deadlier warheads in underground silos. Centuries earlier, European powers drafted chivalrous rules of war right as gunpowder turned castles into rubble. The business model of modern statecraft works the exact same way: sell allies the full-stack ecosystem under the banner of security, block competitors from renting server space abroad, and hold lofty summits about "keeping humans in command."
The quiet irony of "de-risking" AI is how cheerfully we sprint toward superintelligence while debating where to install the kill switch. Civilization loves pretending a few export rules and a red telephone can tame human ambition—forgetting that every clever ape still thinks he is the one holding the remote control.
artificial intelligence, US-China relations, de-risking, geopolitics, human nature, evolutionary psychology, export controls,superintelligence, history, business models, cynical humor, technology policy
給超級大腦裝上緊急電話:為什麼「去風險」其實是穿著安全帽的權力博弈
人總有一種很可愛的本事,一邊拼命磨出世界上最鋒利的長矛,一邊把磨刀石鎖進保險箱,然後替這套做法取個聽起來很冷靜的名字,叫做「去風險」。看看華府近期對人工智慧的布局:官員嘴上說不尋求全面脫鉤,手上卻把高階晶片出口管制越收越緊,連海外數據中心的雲端遠端算力後門都一併堵上,同時嚴防模型蒸餾技術外流,積極向盟友推銷全套美式技術生態。可轉過身,財政部長貝森特又宣布與北京建立「超級智慧」對話與緊急通報熱線,好讓兩大對手在競賽之餘,萬一哪天模型真出了大事,彼此能先通個氣。
把外交辭令撥開,眼前的戲碼其實是人類最古老的領地與生存算計。從演化的底色來看,群體裡最強悍的競爭者永遠會死守最厲害的武器,不讓對手輕易拿到手;但當手裡的傢伙危險到可能把整座森林燒光時,再兇的對手也會隔著河岸點個頭,約好千萬別擦槍走火。誰也沒有打算放慢腳步,大家只是想確保,在分出勝負之前,別先被自己造出來的怪獸反咬一口。
翻開古今中外的強權史,越是致命的發明,越容易催生出看似理性的安全儀式。冷戰時期,美蘇兩國一邊在地下室瘋狂擴充核武庫,一邊在克里姆林宮與白宮之間架起紅色熱線、簽署限武條約;更早以前,列強也在火藥改寫戰場的同時,坐下來談起紳士般的戰爭規矩。現代科技地緣政治走的仍是同一套商業與權力邏輯:把盟友綁進自家供應鏈裡確保安全,把對手擋在算力大門之外,再在高峰會上莊嚴宣告「人類必須牢牢掌握控制權」。
這場人工智慧「去風險」大戲最諷刺的地方,在於所有人都在加速衝向未知的懸崖,卻以為多裝幾個開關、多設幾條熱線就能萬無一失。文明世界總喜歡相信規矩能管住野心,卻忘了每一個按下啟動鍵的人,心裡都以為自己才是那個永遠握著遙控器的贏家。
人工智慧, 美中關係, 去風險, 地緣政治, 人性觀察, 演化心理, 出口管制, 超級智慧, 歷史借鏡, 科技政策, 黑色幽默,權力博弈