In the ever-evolving race toward creating artificial superintelligence, Ethereum co-founder Vitalik Buterin is raising a red flag. In a thought-provoking blog post published on January 5th, Buterin proposed an audacious idea to temporarily halt global access to powerful computing resources in order to “buy more time for humanity” as AI technology rapidly progresses.
With AI superintelligence potentially just five years away, Buterin’s proposal aims to slow down the development of these advanced systems, especially if their capabilities pose risks to society. Let’s break down this radical suggestion and the reasoning behind it.
A “Soft Pause” on AI Hardware: What Does It Mean?
In his post, Buterin expanded on the concept he first introduced in November 2023—defensive accelerationism (d/acc). While the term might sound like a paradox, d/acc is all about cautiously advancing technology while considering its long-term risks. Buterin suggests that, if super-intelligent AI starts to look like an existential threat, one last-resort measure could be to put the brakes on the hardware driving this innovation.
Buterin’s proposal involves imposing a temporary global pause on industrial-scale computing hardware—the very machines used to train and power AI systems. The idea would be to reduce the available computational power by as much as 99% for one to two years. This pause would give humanity crucial time to prepare for the potential dangers of AI, allowing governments, researchers, and society at large to come up with strategies to manage AI risks.
It’s a bold move, but Buterin believes it could be necessary to slow down the race for AI superintelligence—at least until the world can figure out how to control it.
Why the Concern About Superintelligent AI?
To understand why Buterin is sounding the alarm, it’s important to define what AI superintelligence actually means. Superintelligent AI is a theoretical form of artificial intelligence that vastly surpasses the best human minds in every field—scientific discovery, creativity, social intelligence, you name it. Imagine an AI that can solve problems humans can’t even conceptualize.
While this idea still sounds like science fiction to many, Buterin warns that it could be closer than we think. And there’s no guarantee that such an advanced form of AI would have humanity’s best interests at heart.
Buterin isn’t the only one concerned about this possibility. In March 2023, over 2,600 tech executives and researchers signed an open letter urging for a pause in AI development due to the “profound risks” it poses to society. It’s clear that Buterin is part of a growing movement that’s worried about the potential dangers of AI, and he’s calling for a more cautious, deliberate approach.
The Mechanics of a “Soft Pause”: How Would It Work?
Buterin’s vision for a “soft pause” on AI hardware isn’t just about pulling the plug on machines willy-nilly. It’s about setting up a system of checks and balances that ensures AI development slows down in a controlled, transparent way.
One key proposal is the location and registration of AI chips. By tracking the hardware that powers AI, Buterin suggests we could prevent anyone from secretly pushing forward with dangerous AI experiments. But here’s the kicker: Buterin proposes that AI hardware could be equipped with a special chip that only allows it to continue running if it receives three signatures from major international bodies once a week. These bodies would be responsible for ensuring that AI projects are being developed safely and in alignment with global standards.
This wouldn’t be a simple process. The signatures would be device-independent, meaning no single company or individual could override the system without global coordination. In other words, it would be an all-or-nothing approach. If a machine doesn’t get the required signatures, it doesn’t run.
Buterin envisions this approach as a way to create a kind of “digital arms control” for AI hardware, ensuring that only those working within agreed-upon safety frameworks can continue developing superintelligent systems.
When Would the Pause Happen?
Buterin’s idea isn’t about throwing a wrench into the gears of technological progress for the sake of it. Instead, he suggests that this drastic measure would only be taken if the risks posed by AI become too high to ignore. Buterin acknowledges that simply relying on “liability rules”—where companies or developers could be sued for AI-caused damages—might not be enough to rein in the potential dangers of runaway AI development.
So, a hardware pause would only be enacted if the situation became dire—when it’s clear that the stakes are too high to continue without global oversight and intervention.
The Dangers of Unchecked AI: A Call for a Balanced Approach
Buterin’s stance contrasts sharply with that of effective accelerationism (e/acc)—the belief that we should push technology forward as fast as possible, with minimal regulation or concern for the consequences. While e/acc advocates want to speed up innovation, Buterin’s d/acc approach calls for careful and measured progress. It’s a recognition that, as we create increasingly powerful technologies, we need to take extra care to ensure they don’t spiral out of control.
For Buterin, it’s about finding balance: embracing technological advancement, but with enough caution to protect humanity from its own creations. After all, the faster we advance, the faster we create new potential risks. And AI, according to Buterin, might just be the biggest risk of all.
The Bottom Line: Can We Handle Superintelligent AI?
Buterin’s proposal is a reminder of how serious the conversation around AI safety has become. While some argue that AI superintelligence is still far off, others believe we may be closer than we realize. As the power of AI grows, the question becomes: How do we balance innovation with safety?
Buterin’s “soft pause” on AI hardware might be an extreme measure, but it’s a conversation we’ll likely be hearing a lot more about in the coming years. Will we be able to regulate AI before it becomes too powerful to control? Only time will tell.
For now, Buterin’s idea of buying time for humanity remains an intriguing—and necessary—thought experiment for anyone who’s keeping an eye on the future of AI.
