The most unsettling thing about AI is not how fast it is progressing: it's that its creators don't know how it acquires emergent capabilities

Dario Amodei, CEO of Anthropic, stated in an April 2025 essay that neither his team nor he fully understands how their AI models operate, particularly regarding emergent capabilities. The admission highlights the uncertainty and potential safety risks surrounding interpretability in advanced AI systems.

Summary written by Kernelia from the original article by Xataka. The story and its rights belong to its author.