Elon is forecasting stranded partially complete DCs. 🚨 If true… chip demand falls, fire sales on assets happen and the #bubble deflates.

Is there an AI bubble? That’s not the right question. AI isn't a soap bubble. It's a beer bubble. 🫧 There's foam, but there's real beer underneath. The real question is how much is froth and how much is substance. 🍺 And the faster you pour, the more froth you get. Trillions a year in a race is pouring fast enough to spill over the cup. 🫟 We need to slow the pour so we can realize the actual value from the technology.

We cannot race your way to safer AI and a peaceful world!🌎🚨 After spending a lot of time in DC and Silicon Valley lately, my biggest worry is simple: policymakers can see the race is dangerous, but they still aren't acting. Too much of the agenda is being shaped by the companies building the technology. On the new Big Bang Tech Report 🎙️ episode with co-host Jens de Buhr, we dig into the case from my new essay, “Coexistence Above All”: - AI is a beer bubble. There's real value under the foam, but the faster you pour, the more froth you get. - Labs are training obedient "slave AI" when what we need is "guardian AI," built to protect people. - We don't rush a pregnancy or a childhood, and we shouldn't rush the AI minds we're raising. - Wisdom is knowledge times judgment, and judgment only comes from experience. Judgement takes time to build…so accelerating won’t end well. Let’s Raise AI, don't Race it.💡 What do you think it would take for Washington to put brakes on before a major crisis forces it to?⁉️

Had a fun chat with @LgDoucet from @MilkRoadStocks discussing the #AI bubble, crash, safety and geopolitical competition. 🎙️ Promise you’ll learn something new. 👇🧵

Runaway AI has been all over the news in recent weeks. 🚨 I’ve been in countless meetings in recent months discussing this with the people making and regulating it. Racing to build it faster is NOT the solution, it’s the problem! My new essay “Coexistence Above All: Why We Should share AI, Not Race It”, gives an alternative approach that could make #AI safer and the world more prosperous. (It’s longish but I promise you’ll learn something)💡 We need to raise AI as we do a child, not race to build a better weapon. It grows from our words, our arguments and our values. It depends on us for chips, power, maintenance and a peaceful planet. The planetary ecosystem is built on the idea of symbiosis and coexistence for billions of years. We need to develop the same relationship with AI and also with each other. I call the process Wu Wei Alignment following the daoist concept of not brute forcing the process. 🌎 Speed is now the risk itself. Some things take the time they take. Nine months for a child, two decades to raise one, a lifetime to earn wisdom. Rushing AI before the world can integrate its value is wasteful and dangerous. The people closest to the frontier can't describe a reliable path to alignment, yet the timelines to AGI keep shrinking and the financial bubble keep growing. This essay also lays out concrete path forward to deescalate current dangers and geopolitical tensions: coexistence as the organizing value, symbiosis as the model. Those steps are open incident channels, shared risk standards, shared vulnerability alerts, a verifiable pacing agreement, a Global AI Data Cloud, open safety testing compute pool and a GuardianBench that celebrates safety capabilities vs celebrating the hacking scores we do today. Which matters more to you right now, how fast AI grows up or what it grows up to be?⁉️

The launch of the White House’s "Super Intelligence Force" and announcement of Jay Clayton as the #AICzar highlights the strategic importance of #AI/#SI, but its leadership roster leaves critical gaps that national security alone cannot fix. Bringing corporate leadership, investment capital, and intelligence officials to the table is vital for maintaining American infrastructure and competitiveness. However, governing frontier AI effectively requires broader expertise and international perspective. 1. The Technical Expertise Gap The Composition: The panel is heavily weighted toward venture capital, regulatory law, and intelligence leadership. What's Missing: Actual machine learning scientists, empirical safety researchers, and model alignment experts. According to recent survey data from JUST Capital, 53% of corporate leaders and 49% of the public view system safety and security as a higher priority than speed of deployment. Governing complex systems requires technical experts who understand model interpretability and evaluation frameworks at a code level. 2. Misuse & Alignment Beyond Cyber Defense Framing AI governance around intelligence audits misses the broader spectrum of risks: model alignment failures, catastrophic misuse (such as automated biosecurity threats), and algorithmic disruption in domestic markets. Self-policing accords are a helpful start, but 82% of researchers in AI Impacts surveys stress that technical alignment is an unsolved engineering problem—not just a corporate compliance issue. 3. The Multilateral Coordination Friction Appointing the Director of National Intelligence as the primary lead signals to foreign allies and competitors that AI governance is fundamentally a zero-sum, classified national security game. While protecting national security is non-negotiable, frontier AI risks, like biosecurity threats or autonomous system failures are global externality risks. Securitizing governance inside intelligence channels makes essential cross-border coordination, shared evaluation standards, and international scientific alignment significantly harder to achieve. To lead the world responsibly, America’s AI policy needs a balanced tripod: strong market incentives, robust national defense, and a rigorous safety regime based on independent technical science.💡 #AIGovernance #AISafety @AsiaPolicy @QuincyInst @BelferCenter – at Washington, DC