Vitalik Buterin Warns: Superintelligent AI Could Arrive Sooner Than Expected

Ethereum co-founder Vitalik Buterin has expressed concerns about the rapid development of artificial intelligence (AI), warning that superintelligent AI might emerge sooner than expected. Buterin emphasizes the urgent need for strategies to counter potential risks, focusing on what he calls "defensive acceleration" to ensure AI technology is used responsibly.

In a blog post on January 5, Buterin outlined his proposals to prevent harmful advancements in AI. He advocates for decentralized AI systems closely tied to human decision-making, aiming to reduce the risk of misuse, particularly by military forces. He highlights the growing global trend of AI in warfare, citing examples like its deployment in Ukraine and Gaza, and warns that military exemptions in AI regulations could pose significant threats.

Buterin estimates that artificial general intelligence (AGI) could be just three years away, with superintelligence potentially emerging three years after that. He stresses that humanity cannot merely accelerate beneficial advancements but must also actively slow down harmful developments. He describes a scenario where unchecked AI could lead to catastrophic outcomes, including the possibility of human extinction.

To address these risks, Buterin suggests several measures. First, he calls for liability rules to hold users accountable for how AI systems are utilized. While acknowledging the complexity of linking AI development to its use, he argues that end users ultimately decide the technology's applications.

If liability measures prove insufficient, Buterin proposes "soft pause" mechanisms. These would temporarily slow down the development of dangerous AI systems, potentially by reducing global compute capacity by 90-99% for one to two years during critical periods. This would give humanity time to prepare for emerging challenges.

Another key suggestion involves controlling AI hardware. Buterin proposes integrating chips into AI systems that require weekly authorization from three international bodies, with at least one being non-military. This measure aims to maintain global oversight and prevent misuse.

Despite presenting these ideas, Buterin acknowledges that his strategies are temporary and imperfect. However, he insists that immediate action is necessary to manage the risks posed by rapidly advancing AI technologies.

Buterin's warnings come at a time of increasing concern about AI safety, highlighting the need for global cooperation to address these pressing issues. By ensuring AI remains under human control, the risks of catastrophic outcomes can be minimized, but achieving this will require collective effort and vigilance.