No, Anderson Cooper, AI is not going to kill all humans by 2030

Substack · Watch/Read

• Gary Marcus dismisses the “Doomer” narrative that AI will cause human extinction by 2030 as an “absurd end-of-the-world fantasy,” arguing that constant focus on this scenario has made the situation worse by empowering frontier AI companies.
• He criticizes “Doomers” like Jacob Coxon, Eliezer Yudkowsky, and Nate Soares, claiming their alarmist rhetoric has led to massive funding for companies like OpenAI and Anthropic without producing viable safety solutions, and that their arguments for human extinction are “flawed” and lack nuance regarding human resistance or real-world dynamics.
• Marcus emphasizes the critical distinction between “existential risk” (literal extinction, which he believes AI poses a “barely above zero” chance of) and “catastrophic risk” (severely damaging civilization or killing 1%+ of humanity), which AI does significantly elevate.
• Humanity should focus on concrete, catastrophic risks elevated by AI, such as bioweapons, disinformation, cyberattacks, authoritarian surveillance, and harm to the educational system, rather than “fairy tales” about extinction.