Tao Lin, who works on AI safeguards at Anthropic, believes broadly superhuman AI could arrive in around 15 months, potentially leaving humans permanently unemployable and dependent on government welfare.