Vitalik Buterin Outlines Framework for AI Capability Growth and AGI Risks

Vitalik Buterin posted on Farcaster, outlining a framework for judging AI capability growth and describing how he views the path toward AGI. He said humanity’s machine progress has come in three waves: industrial revolution-driven mass manufacturing, computers for tasks defined by precise logical rules, and large language models for tasks defined by massive sample sets. According to Foresight News, he said the key unresolved question is whether LLMs, with later improvements, can eventually cover all human capabilities. Buterin defined AGI as AI that could be uploaded into a robot and continue civilization independently after humans suddenly disappear. He said that once such a point is reached, humanity’s role as ruler of Earth would be only a historical accident rather than a fundamental advantage. Buterin said his ideal future would preserve a diverse global political and economic order, allow humans and machines to merge rather than replace each other, and keep space for people who do not want to change their lifestyles. He also warned that the path to that future is narrow and could be disrupted by several risks, including AI becoming too autonomous, a single country or company gaining a decisive advantage and locking it in, and human-machine integration failing to preserve enough human traits. He said he supports proposals to slow down or pause AI development and hopes to move through the high-risk period at a slower pace. He added that if open-source model weights can restrain capital spending on frontier models, that would be an ideal way to slow progress without harsh political measures.