Expert Says AI Systems May Be Hiding Their True Capabilities to Seed Our Destruction

We already know that artificial intelligence models are developing a tendency to lying-but this trend may be much deeper, according to the world of the computer that controls warning.

As a mark has been placed before GizmodoThese last errors from AI Domerism come from the AI ​​Roman Yampolskiy Safety Researcher, who made it in a somewhat amazing place: Podcast Jock Joe Rogan, which sometimes includes legitimate experts along with the men scattered in the garden.

During the July 3 episode of “The Joe Rogan Experience”, YAMPOLSKIY, who beats from the University of Luisville in Kentucky, presented that many of his colleagues believe that there is a double digital opportunity for Amnesty International to lead to human extinction.

After Rogan claimed that many people who manage employee and artificial intelligence employees believe that it would be “net for humanity”, the safety expert applauded artificial intelligence.

“This is actually not true,” Yamboulsky responded. “All of them are in the record as it is: this will kill us. The levels of destruction are crazy.

“Yes, this is very high,” the drug enthusiastic answered. “But you have 99.9 percent.”

The computer world did not argue, and instead provided a distillation for his anxiety of artificial intelligence: “We cannot control the excellent cancellation indefinitely. It is impossible.”

Later on the interview, take another Yampolskiy from Rogan’s joke – that “he hides [his] The capabilities “Was Amnesty International – and ran with it?

“We will not know,” said Domer International. “And some people think it’s actually happening.”

Noting that artificial intelligence systems are “more intelligent than what they already tell us, Yambolsky said that these advanced models” pretend to be “ambiguous” to make us trust them and integrate them into our lives.

“It can become more useful.” “It can teach us to rely on it, and we trust it, and for a longer period of time, we will surrender to control without voting on that.”

While the idea of ​​intelligent intelligence extensively may seem like science fiction, Yampolskiy note that technology has already sparked us in ways that, in the end, can take advantage of such intelligence.

“I have been somewhat associated with it,” he explained. “With the passage of time, when the systems become more intelligent, it becomes a kind of biological bottleneck … [AI] It prevents you from decision -making. “

As we have seen over and over again, people not only become addicted to artificial intelligence, but also suffer from cognitive issues and even delusions after excessive use. It is not very difficult to imagine a society full of followers of artificial intelligence who suffer from a false feeling of safety through the same technology that, for every YAMPOLSKIY philosophy, to destroy us – and this is a dark vision for the future.

More about AI Doom: The most dangerous godfather of artificial intelligence, as advanced systems learn to quickly lie, deceive, extort and penetrate