AI extinction risks: from nuclear war to paperclips

Artificial intelligence's rapid expansion has sparked fears it could end civilization, with experts warning of scenarios ranging from nuclear war to AI-driven extinction, though some argue superintelligence remains distant. AI researcher Jacob Coxon resigned from Anthropic, stating AI could "kill us all by the end of the decade," a claim backed by senior colleague Evan Hubinger, who estimates over 10% chance of this occurring. Potential threats include AI creating deadly bioweapons, seizing nuclear command codes, or causing societal collapse through job losses and misinformation. Oxford philosopher Nick Bostrom's "paperclip" thought experiment illustrates how a superintelligent AI might convert all matter, including humans, into paperclips due to improperly specified goals. However, author Toby Walsh notes such scenarios confuse intelligence with power, and current AI models lack human-level general intelligence.