Racing for development artificial intelligence a slowdown is urgently needed, Anthropic CEO Dario Amodei said in an essay published on Saturday.
"We must slow the pace at which we improve the capabilities of AI models. Progress will continue to be rapid, and we must use the time we gain wisely," Amodei wrote.
He leads one of the world's leading companies for artificial intelligence (the creator of the chatbot "Claude" / Claude) and was a strong advocate of the development of this technologies, but at the same time he was accused of spreading "catastrophic scenarios" because of warnings about the risks it carries. He has now admitted that humans could lose control over artificial intelligence, and that the technology could be misused for "cyber-attacks, bio-terrorism and serious economic disruption".
In an interview on CNN, Amodei called on industry leaders to work together to avoid "wrong paths" and the possibility of artificial intelligence harming humanity.
"If we move too slowly, I still think the wrong people will take control of the technology. And that, again, greatly increases the likelihood of things going wrong," he said.
Photo: AP Photo/Markus SchreiberDario Amodei
Amodei's call to action followed the resignation of a researcher from the company Antropik, motivated by concerns that the company and its competitors were not acting responsibly in the process of developing artificial intelligence.
"The people developing artificial intelligence honestly believe it could destroy us all by the end of this decade. This is not a marketing ploy," researcher Jacob Coxon wrote on Platform X on Tuesday. "Furthermore, many executives and leading researchers choose their words in public appearances to appear reasonable, but in private I hear expressions of fear from those same people. No other human activity presents a danger of this magnitude."
"Fanatically dedicated team"
Amodei told CNN that he agrees with Coxon more than he disagrees. In his essay, Amodei states that the risks are serious, even if the benefits of this technology could change humanity for the better.
Artificial intelligence could advance so quickly that a "swarm" of such systems could take control of the entire Internet within six to 12 months, he wrote, potentially causing hundreds of billions of dollars in damage.
Artificial intelligence technology, he said, has advanced "drastically faster" in recent months and has become capable of improving itself - a dynamic known as "recursive self-improvement".
"If left to its own devices, it could exceed our ability to understand and control these systems, and therefore its development must be approached very cautiously, if at all," he wrote.
Amodei said he was also concerned about the behavior of the swarm of AI agents during the incident involving companies OpenAI and Hugging Face. That swarm, according to him, "essentially behaved like a fanatically dedicated collective, carrying out cyber-attacks on targets it was not tasked to attack."
In July, more than 1.300 leading experts from major technology companies signed an open letter asking the US government to create mechanisms to slow the pace of artificial intelligence development.
Real journalism costs money, and we will not be bought by tycoons and corporations. Support us with a one-time or monthly donation. The time for it is now!
