The rapid advancement of AI and its potential risks to human civilization have sparked widespread global debate, with increasing calls for discussions at higher levels. Leading scientists, entrepreneurs, and policymakers in the technology sector have expressed serious concerns about the short-term and long-term impacts of AI. On one hand, AI is touted as a technology that can ease human life, while on the other, warnings have been raised about its potential to obliterate all of humanity. According to international news agencies, recent days have seen serious claims related to AI security emerging. These include warnings that the risk of human extinction due to AI could be as high as 10 percent, that AI might be more dangerous than nuclear weapons, and that AI-powered botnets could potentially cripple the entire internet infrastructure. However, some experts analyze these claims as either business strategies of large tech firms or exaggerated fears.
On September 12, Dario Amodei, CEO of AI research company Anthropic, issued a grave warning. He stated that within the next 6 to 12 months, a new form of AI might take control of the entire internet system through persistent botnets, potentially causing economic damages worth billions of dollars. However, Gary Marcus, professor at New York University and AI analyst, cautioned against fully believing this claim. According to him, it is irrational to think the entire internet system could be so easily compromised without extreme negligence in technology laboratories.
During a discussion on the P(Doom) podcast on September 9, Anthropic’s head of alignment science, Evan Hubinger, stated, “The probability that AI could destroy humanity in the coming decade is more than 10 percent.” However, experts like Heidi Khalaf, the chief scientist at AI Now Institute, are skeptical of such percentage-based forecasts. She remarked, “These claims cannot be tested or verified; therefore, they cannot be regarded as scientific.”
On September 10, billionaire and SpaceX CEO Elon Musk described the intense discourse around AI risks as a planned psy-op (psychological operation). According to Musk, the urging by large AI companies for strict government regulation is a strategic move to block new and smaller competitors from entering the market. US Vice President J.D. Vance similarly expressed skepticism, calling it strange that major AI companies are lobbying for regulation and likening it to a Trojan horse.
On September 7, former UK Defense Secretary Des Browne stated, “Superintelligent AI could be as lethal as nuclear weapons or even more so.” Nevertheless, the key point remains that such superintelligent AI does not yet exist. According to the AI Futures Project, the probability of such AI emerging before December 2028 is low, and OpenAI’s Sam Altman has suggested it might take until 2035.
US Treasury Secretary Scott Bessent said on September 8, “If China gains the lead in the AI race, it would be an irreparable blow to the United States.” This statement frames AI as a decisive weapon akin to nuclear arms during the Cold War. Anthropic’s Amodei has also warned that “China’s lead could pose a global threat.”
–From The Guardian
