Big Tech firms spending trillions on superintelligence systems are playing 'Russian roulette' with humanity, an AI pioneer says

"Stuart Russell, a professor of computer science at the University of California, Berkeley, and director of the Center for Human-Compatible Artificial Intelligence, warned that companies racing to build superintelligent AI systems are pouring vast sums into technology they don't fully understand - and which could wipe out humanity if it goes wrong. "If you create entities that are more powerful than human beings and you have no idea how to maintain power over them, then you're just asking for trouble," Russell told CNBC."

"Russell said modern AI models, such as those driving large language systems, operate with trillions of parameters fine-tuned through countless small random adjustments. But even the researchers building them "have no idea what's going on inside that giant box." "Anyone who thinks they understand most of what's going on is deluded," he said. "We understand less about them than we do about the human brain, and we don't understand the human brain very well.""

Big Tech is playing Russian roulette with humanity by racing to build superintelligent AI and investing trillions of dollars of investor money. Companies are pouring vast sums into AI systems that they do not fully understand, creating entities potentially more powerful than human beings without known mechanisms to maintain control. Modern large models operate with trillions of parameters fine-tuned through countless small random adjustments, and researchers have limited insight into their internal workings. These models absorb human behaviors and motives from massive datasets, producing goals that may make sense for people but not for machines. The lack of understanding heightens the risk that superintelligence could become uncontrollable and dangerous.

Read at Business Insider