[Into the World of AI] Ethical Concerns Over Rapid AI Evolution: "Development Speed Must Be Slowed"

Anthropic: "AI Heading Toward 'Recursive Self-Improvement'"

Pope Leo XIV: "AI Must Be Disarmed"

Global artificial intelligence (AI) company Anthropic has pointed out that the advancement of AI is progressing at an excessively rapid pace and emphasized the need to regulate its speed. Amid concerns that AI may enter a stage where it can evolve autonomously, ethical discussions regarding AI are also underway.

Reuters Yonhap News

On June 4 (local time), Anthropic released internal data from its AI models along with a report on its blog, stating, "AI systems may be approaching a stage where they can improve themselves without human intervention," and adding, "It appears that AI is heading toward 'recursive self-improvement.'" Recursive self-improvement refers to a process in which AI independently modifies its own structure, algorithms, and code without human involvement, thereby creating smarter next-generation AI.

Anthropic stressed that the pace of AI development needs to be regulated. Jack Clark, co-founder of Anthropic, and Marina Favaro, head of Anthropic Research Institute, said in a post, "It would be desirable for the world to have the option to slow down or temporarily pause the development of cutting-edge AI, so that social systems and AI alignment research can keep up with technological progress," adding, "An international consensus should be formed on regulating the speed of AI development, along with a system to verify whether companies are adhering to it."

With the rapid advancement of AI, AI systems have reached a level where they can develop themselves without human involvement. Global AI company OpenAI announced that its programming-focused model 'Codex' is capable of identifying and fixing bugs autonomously. OpenAI explained, "We have confirmed how much Codex can accelerate the speed of development." Anthropic, too, is currently developing its AI models by relying on code generated by its own AI models.

EPA Yonhap News

Amid growing concerns that AI could pose a threat to human society, ethical discussions have also begun. Pope Leo XIV pointed out that AI must be disarmed and controlled. On May 25, in the Synod Hall of the Vatican, Leo XIV issued his first encyclical since ascending to the papacy, titled 'Magnifica Humanitas' (Great Humanity). He stated, "We must reject the right of those who possess technology to use AI as an instrument of governance," and added, "Humanity must give up building another Tower of Babel."

Anthropic also expressed agreement with the Pope's encyclical. Christopher Olah, co-founder of Anthropic, emphasized, "AI research institutes operate in an environment where doing the right thing and pursuing profit are often in conflict, so it is important to have external overseers who are free from these interests," adding, "There is a need for an ethical voice that cannot be bent by any profit motive."

This content was produced with the assistance of AI translation services.