What is AI 2040 about?

[UltimaOnline]

Ask AI : What is AI 2040 about?​

"AI 2040: Plan A" is a highly detailed, optimistic scenario and policy roadmap published by Daniel Kokotajlo’s nonprofit, the AI Futures Project. It serves as a direct sequel to their widely discussed "AI 2027" report, which previously predicted an aggressive, terrifying intelligence explosion and potential human extinction. Released on platforms like the AI Alignment Forum and LessWrong, "Plan A" outlines how humanity can strategically steer away from doomsday toward a safe, cooperative future. [1, 2, 3, 4]
The core narrative and policy mechanisms of AI 2040 revolve around several main concepts:

Key Concepts of the Plan

• Delaying Superintelligence: The plan proposes an intentional, internationally coordinated slowdown. Instead of letting an artificial superintelligence emerge unchecked by 2030, the United States and China broker a deal to delay its creation until 2040. [1, 2, 3, 4]
• The "Sweet Spot" Timeline: This delay keeps humanity in a manageable range of AI capability. It gives researchers enough time to use highly advanced (but not all-powerful) AI systems to solve the core mathematical and technical problems of AI alignment before it is too late. [1, 2]
• Mutually Assured Compute Destruction (MACD): To prevent cheating, the US and China build their frontier AI data center clusters in neutral territories (e.g., the US builds in Mongolia, China builds in Canada). If the diplomatic treaty is violated, the host nations can instantly pull the plug, or the owning nation must self-destruct the hardware. [1]
• Radical Transparency and Open Source: Rather than locking frontier AI technology away inside corporate labs, the scenario mandates making AI research public. It heavily promotes information diffusion so that dozens of global companies can catch up safely. [1, 2, 3]
• Explosive, Managed Prosperity: Despite a "slowdown" in raw model scaling, the timeline projects massive economic transformation. It envisions massive GDP growth by 2032, a Universal Basic Income (or citizen dividend) rolling out by 2033, and major diseases being cured by 2036, culminating in a stable "handover" of certain governance powers to aligned AI by 2040. [1, 2]

Reception and Debates
The scenario has sparked intense debate among industry insiders, researchers, and tech philosophers: [1, 2, 3]

• The Critique of Plausibility: Prominent figures and commentators on platforms like Hacker News and Reddit have called the plan a "fairy tale," pointing out that it requires an unprecedented level of global political coordination that ignores real-world geopolitical greed and friction.
• Vitalik Buterin's Take: Ethereum co-founder Vitalik Buterin weighed in on X (formerly Twitter), noting that he appreciates the plan's open-source focus and its move away from letting a few powerful AI labs dominate the world. He pointed out that whether one supports this plan depends heavily on how close they think superhuman AGI actually is.
• Handover Concerns: AI safety researchers, such as Richard Ngo, have published criticisms alongside the text, questioning whether the sudden transition to "passing the torch" to AI in 2040 is too abrupt and democratic enough for normal citizens to consent to. [1, 2, 3, 4, 5, 6]

 

[UltimaOnline]

 

[UltimaOnline]

AI Debate Driven By Competing Views

Buterin’s piece framed the larger discussion between “AI 2040” proponents and opponents as a conflict of “totally incompatible worldviews” about how quickly and how far AI would go, and noted that his own perspective rests on assumptions he can’t definitively prove.

The Ethereum co-founder said he remained open-minded to AI slowdowns or pauses given his uncertainty about which scenario was correct, while pushing back on arguments he said concentrate power in the hands of a small number of major AI labs or governments.


Ethereum's ‘d/acc’ Framework Remains Central

He returned to his original stance of connecting to Ethereum's 'd/acc' architecture, investing in cryptography, formal verification, secure hardware and biosecurity, arguing that these investments are worthwhile regardless of what AI scenario unfolds.

Buterin said his own proposal to retool the platform "is also naive" but that he currently sees "zero plans" for managing an AI transition that are not naive in some way.


AI 2040 Proposal Sparks Industry Debate

The comments come days after "AI 2040: Plan A," a report published by former OpenAI (OPEAZZX) researcher Daniel Kokotajlo's AI Futures Project, proposed that the U.S. and China cooperate to delay the arrival of superintelligence to 2040 deliberately. The plan calls for full research transparency between the two countries and a verification framework it terms "mutually assured compute destruction," modeled loosely on nuclear deterrence logic.

The report has drawn criticism from AI researchers, including Richard Ngo, who argued that it overestimates how imminent superintelligence really is and underestimates the domestic political disruption AI will cause well before any US-China "race" plays out.