An AI Rant

Carl Handellen argues that warnings of AI-driven human extinction are alarmist and unsupported. He accepts that AI is already changing daily life and has transformed software engineering, but rejects predictions that advanced systems will soon eradicate humanity. He cites a 2023 statement comparing AI extinction risk with pandemics and nuclear war, Eliezer Yudkowsky’s claim that an uncontrolled powerful AI could kill all biological life, and Anthropic figures’ estimates: Evan Hubinger’s belief in a greater-than-10% chance of catastrophe within a decade and CEO Dario Amodei’s 25% chance that outcomes go “really, really badly.” Handellen says these claims come without a public, detailed account of how such a catastrophe would unfold.

He describes the feared scenario as depending on several uncertain steps: AI must become vastly more intelligent than humans, continue improving without hitting limits, regard people as an obstacle, and find a way to destroy humanity despite lacking direct access to the physical world. By contrast, he presents economic harm from slowing or halting AI development as a more immediate certainty, arguing that restrictions could trigger recession and widespread hardship; he points to a sharp Nasdaq decline following the prospect of regulation. He suspects, while acknowledging he is speculating, that some researchers’ beliefs reflect an existing “doomer” worldview and that AI-company executives may use existential warnings to justify regulation that entrenches their market power and limits open-source competitors.