Anthropic CEO Dario Amodei argues that the next generation of powerful artificial intelligence could quickly reach capabilities equivalent to a “country of geniuses in a data center,” with computational capacity surpassing the collective knowledge of tens of millions of Nobel Prize winners. In a long essay titled “The Adolescence of Technology,” he predicts that by around 2027, interconnected AI clusters might run millions of AI instances operating at superhuman speed. This metaphor underscores the unprecedented scale of intelligence and computing power that could soon emerge from advanced AI systems.
Amodei highlights that such super-capable systems may pose serious global risks, including the possibility of hostile behavior, disruption of the global economy, and large-scale unemployment. He suggests that if an AI “country” were to act autonomously or fall into the wrong hands, it might outpace human capacity to control or counteract its actions, potentially destabilizing worldwide social and economic systems.
He connects these concerns to broader predictions about the near-term impact of AI on jobs, including controversial views that many entry-level white-collar positions could be displaced within the next few years. Amodei has repeated similar warnings at venues like the World Economic Forum in Davos, emphasizing that AI progress could accelerate rapidly, moving past familiar benchmarks before society fully grasps its implications.
Despite these stark warnings, Amodei rejects outright “doomer” perspectives and points to Anthropic’s own approach to safer AI development, including the use of what the company calls Constitutional AI — a training method intended to imbue AI systems with guiding values rather than relying solely on prohibited actions lists. He also advocates for greater transparency in how powerful AI models are built and trained, urging policymakers to adopt disclosure rules that could help manage the technology responsibly.
