It's humanity's worst nightmare: the evolution of artificial intelligence could pose a true existential threat to humanity by the end of the decade, destroying it. The alarm was raised, in what sounds more like the plot of a movie, by a researcher at Anthropic—the American public benefit company in the field of artificial intelligence—who resigned from the U.S. benefit corporation, warning that the uncontrolled race to create a superintelligence capable of self-improvement could destroy humanity as early as this decade, by 2030.
His farewell is further fueling the wave of scientists who have decided to leave their laboratories in recent months, alarmed by the potential of the systems they are developing.
His name – reports the Financial Times – is Jacob Coxon, a 27-year-old British researcher working for the San Francisco-based company, who is convinced that Anthropic and its main competitor, OpenAI, are “gambling” with the future of humanity.
"The people developing AI sincerely believe it could kill us all by the end of the decade," he denounced in a post on X announcing his resignation. "This is not a marketing ploy," he added. "No other human activity carries this level of danger."
Coxon is just the latest employee to leave a major US AI lab, citing safety concerns among other things. At the end of July, Chloé Bakalar, the company's only dedicated ethicist, left, followed by Johannes Heidecke, head of the Safety Systems team. Then came Joshua Achiam, formerly head of Mission Alignment, COO Brad Lightcap, and finally researcher Andrew Ho, who resigned due to technical disagreements.
Coxon's departure comes as Anthropic—which has made AI safety a central element of its public image—prepares for a massive initial public offering (IPO) that could value the company at as much as $1 trillion.
Coxon, who previously worked at OpenAI before joining Anthropic, warned that people outside the lab are underestimating the power of this technology. "Soon, these will be superhuman systems capable of hacking anything, revolutionizing any industry overnight, and gaining real power and resources."
Evan Hubinger, a senior security researcher at Anthropic, echoed Coxon's warning, saying that artificial intelligence is evolving so rapidly that he believes there is a greater than 10% chance that it "could exterminate all of humanity" within the next decade. In a post on X, Hubinger stated that the risk from existing models is "low" but expressed concern that the technology could soon develop and become so sophisticated that it could pose an existential risk to humanity in the near future.
(Unioneonline)
