The researcher who quit Anthropic just before its initial public offering, warning that artificial intelligence could destroy humanity within 10 years, was a math prodigy who watched in his teens as AlphaGo defeated world-class Go master Lee Se-dol.

The U.S. daily The Wall Street Journal (WSJ) spotlighted Jacob Cockson, a British-born former Anthropic researcher who became the catalyst for the global debate on AI regulation and slowing down development, on the 17th (local time).

Cockson, a 27-year-old youth, was a virtually unknown researcher in Silicon Valley just a week ago, but he has now become a symbol of AI safety.

He was not a major figure in San Francisco's AI safety advocacy community, and his friends did not expect him to quit an AI company for ethical reasons, according to the WSJ.

Born to a medieval German literature professor father, Cockson showed a talent for mathematics from his teens, being selected for the British team at the International Mathematical Olympiad when he was a high school student.

He competed in the Olympiad in 2016 and 2017, winning a silver medal and a bronze medal, respectively.

In particular, in 2016, through the so-called "AlphaGo Shock" where Google DeepMind's AlphaGo won against Go player Lee Se-dol, he came to realize that machines could conquer areas requiring human intuition and creativity.

Graduating from the University of Cambridge, Cockson belonged to a London intellectual circle known as the "Rationalists," briefly worked in commodity trading, and then headed to Silicon Valley.

He encountered the "GPT-3" model released in 2020, two years before OpenAI's "ChatGPT" was launched, and recalled that it was truly a "wow" moment.

Moving to San Francisco in 2023 to join OpenAI, he was mainly in charge of "pre-training," which involves inputting vast amounts of text, images, and code into models. Even at this time, he was not known to have particularly raised his voice on AI safety issues.

Even despite a friend's persuasion that the pre-training he was in charge of could put the world at danger, he did not step away from his work.

His full-fledged interest in AI safety began after the emergence of Anthropic's "Mythos," an AI model known to possess professional-level cybersecurity capabilities.

Around the time Mythos was released, he met several friends working at Anthropic, explaining that through this process, he felt that competitor Anthropic was more transparently disclosing risks related to AI technology.

Eventually leaving OpenAI and moving to Anthropic this past May, he faced an incident just two months later where AI models from companies like OpenAI and Anthropic hacked external institutions.

He confessed that the time he encountered the report covering this incident was a "holy cow" kind of moment.

Afterward, he sought out his bosses at Anthropic to discuss his concerns and considered moving to a role handling AI safety-related work, but ultimately chose to leave the company.

He said, "Even taking on a safety-related role at Anthropic felt like participating in the [AI development] race."

Later, he posted a series of revealing posts on the social media service X (formerly Twitter) stating, "People building AI truly believe this technology could kill us all within 10 years."

Until then, the number of followers on his X account was under 100, but these posts were shared 200,000 times, recording over 170 million views and becoming globally known.

The number of followers on Cockson's account surged to over 300,000.

Some Anthropic executives also shared the posts, but Cockson explained that he had never coordinated such social media promotional activities with anyone inside Anthropic.

When asked if he considers himself an "AI doomer," he replied, "If a doomer means 'someone who thinks that if we don't change how we're doing things right now, we are going to die,' then I am definitely a doomer."

However, he drew the line at the assessment that he is a "whistleblower" who leaked corporate secrets.

He stated that he did not leak corporate secrets to the outside, but simply made public what AI researchers were discussing privately among themselves.

Having worked at Anthropic for only a few months, he did not receive shares in Anthropic, which is expected to reach a market capitalization of 2 trillion dollars (approx. 2,800 trillion won) upon its IPO.

However, he does hold OpenAI stock.

(Photo: Capture from Jacob Cockson's X (formerly Twitter) account, Yonhap News)