The warning came last month, delivered by Anthropic’s CEO, Dario Amodei – a stark assessment that the world’s control over rapidly advancing artificial intelligence is perilously close to a breaking point, potentially leading to catastrophic consequences.
Amodei’s argument, laid out in his article “The Adolescence of Technology,” begins with a haunting question borrowed from Carl Sagan’s *Contact*. Sagan’s astronomer, facing potential contact with an alien civilization, asks what single question she would pose: “How did you do it? How did you evolve, how did you survive this technological adolescence without destroying yourself?”
The question resonated deeply, particularly for a medical historian preparing a lecture on the origins of immunology. A crucial element often missing from discussions of scientific progress is historical context – understanding how innovations impact the intricate fabric of human society.
Amodei isn’t simply identifying a potential problem; he’s attempting to provide the crucial context as it unfolds. Headlines like one from the New York Times – “ICE Already Knows Who Protesters Are” – demonstrate the immediate, tangible implications of AI-powered facial recognition technology.
But Amodei’s concerns extend far beyond surveillance. He highlights the alarming speed of change with generative AI, a pace that is accelerating exponentially. AI is now actively writing the code for its own advancement, creating a feedback loop that could, within just a year or two, lead to autonomous AI development.
This vulnerability is particularly acute within AI companies themselves. These organizations control vast data centers, possess unparalleled expertise, and wield influence over potentially hundreds of millions of users – a concentration of power that demands scrutiny.
Amodei doesn’t shy away from naming the potential dangers, subtly referencing instances of negligence regarding the sexualization of children within AI models. This raises serious doubts about the willingness and ability of some companies to address the far more significant risks associated with autonomous AI.
The chilling reality of this potential is already unfolding. In Portland, Maine, a legal observer filming an ICE agent was confronted and informed she was now labeled a “domestic terrorist” due to her inclusion in an AI-powered database.
Recent events, including the tragic and unprovoked deaths of US citizens, underscore the urgency of Amodei’s warning. Current autocratic regimes are constrained by the limitations of human executioners, by the inherent reluctance of individuals to commit extreme acts of inhumanity.
However, AI-enabled autocracies would face no such constraints. Freed from the moral boundaries of human operators, they could wield repression on an unprecedented and terrifying scale, a future Amodei warns is closer than many realize.
