Ethereum co-founder Vitalik Buterin has sparked renewed debate within the technology community after suggesting that a universal kill switch should be implemented for all artificial intelligence applications. His proposal arrives amid ongoing disagreements about the anticipated pace and implications of AI development, particularly concerning its impact on society and the workforce.
Growing divisions among AI experts
Buterin’s recent statements on X outlined a divide between those who expect the emergence of “superintelligence” by 2040—often referred to as the “AI 2040” scenario—and skeptics who question the urgency of such concerns. Supporters of the AI 2040 outlook argue that exceptional AI capabilities could develop within the next two decades unless decisive action is taken. Meanwhile, critics contend that this view underestimates humanity’s ability to coordinate and manage risks, while also warning that certain regulatory proposals could restrict freedom without addressing the real dangers.
Acknowledging both perspectives, Buterin admitted uncertainty about which scenario is likely. He wrote that if current-day AI remains similar to existing technologies, he would agree with the skeptics, but that faster-than-expected advances would bring him closer to the AI 2040 camp.
Buterin indicated, “If I was confident that superintelligence is coming in 2030 by default, I would be closer to the AI 2040 camp … But I am open to slowing or pausing AI development if the risks become significant.”
His remarks attracted responses from notable figures, including Meta’s chief AI scientist Yann LeCun, author Daniel Jeffries, and policy analyst Adam Thierer.
The debate over AI safety and governance
LeCun argued that AI safety represents an engineering problem that can be addressed through iterative design, much like how aviation safety progressed over decades. He expressed skepticism toward fears of near-term superintelligence and pointed out that current large language models function as limited “autocomplete machines.”
AI commentator Harry Hawk sided with LeCun, suggesting that, just as aircraft have become safer, future AI systems will be designed for reliability. He pushed back against the belief that advanced AI will eliminate all human jobs.
Buterin countered by highlighting the possibility of Artificial Superintelligence (ASI)—an AI so powerful it could accomplish virtually any task without human intervention. The discussion also engaged AI researcher Yann LeCun, policy analyst Adam Thierer, and others who criticized or supported various ideas on AI regulation and openness.
Part of the debate was prompted by Daniel Kokotajlo’s AI Futures Project, which released the “AI 2040: Plan A” report. The report advocates for collaboration between the United States and China to slow development toward superintelligence, including provisions for open sharing of research and “mutually assured compute destruction.”
Richard Ngo, an AI researcher, claimed the report exaggerates the speed with which powerful AI systems could emerge and does not address the political challenges such technology may create within countries.
Mini dictionary: Artificial Superintelligence (ASI), a hypothetical form of AI that surpasses human intelligence across all domains and could autonomously perform any conceivable task.
| Viewpoint | Key Argument | Main Proponent |
|---|---|---|
| AI 2040 Scenario | Superintelligent AI may emerge by 2040. Strong regulation needed. | Daniel Kokotajlo (AI Futures Project), Supported by Buterin (cautiously) |
| Skeptics/Critics | Humanity can control AI risks; focus on freedom and coordination. | LeCun, Thierer, Jeffries |
Open source as a safeguard?
Open source models have become central to the debate. LeCun argued that the true danger lies in the concentration of AI development within a handful of powerful companies, and stated that open source foundation models are crucial for “AI sovereignty.”
LeCun noted, “The only solution to AI sovereignty is open source foundation models,” emphasizing the importance of broad accessibility in AI technology.
Author Daniel Jeffries agreed, warning that efforts by what he called “safetyists and hawks” to restrict open source could stifle innovation. Adam Thierer, a senior fellow at the R Street Institute and key contributor to the US House AI Task Force, labeled the current US legislative process on AI review as disorganized and secretive. He cautioned that increased regulation could harm open source AI, instead supporting “permissionless innovation.”
Thierer recommends using existing laws against harmful applications, creating test zones, mandating openness for certain models, and increasing funding for research, instead of imposing heavy new restrictions.
Mini dictionary: The R Street Institute is a Washington, DC-based think tank that focuses on free markets and limited, effective government, often contributing research and analysis on technology policy issues including AI governance.
