Oxford University researchers have warned British lawmakers that “superhuman artificial intelligence” could be at least as dangerous as nuclear weapons and should therefore be regulated like them, The Telegraph reported.
Experts briefed MPs at the UK government’s Science and Technology Selection Committee about the dangers of unregulated artificial intelligence (AI) technologies. Doctoral student Michael Cohen, quoted by The Telegraph, told lawmakers: “There’s a special risk with superhuman AI that it’s in a different class, that it could kill anyone.”
Artificial Intelligence Apocalypse
These risks are indeed of great importance as political forces increasingly seek to outdo each other in the field of artificial intelligence. Michael Osborne, a professor of machine learning at Oxford University, who also attended the committee meeting, warned that there could be a “major AI arms race” between the US and China, who “want to set aside security and precaution to reach the most advanced AI as quickly as possible”. .
“There is some reason to be hopeful, as we are pretty good at regulating the use of nuclear weapons,” Osborne said. “If we can understand that advanced AI is a comparable danger to nuclear weapons, then perhaps we can arrive at similar frameworks for managing it.”
“If you imagine training a dog with rewards, it will learn to choose actions that will enable it to receive rewards,” Cohen told lawmakers.
In other words, a superhuman AI risks expending “as much energy as it can” to obtain a reward, which “leaves us with energy for ourselves. As these technologies advance, they become “much smarter than we are in every field,” according to Cohen. If they come, we need to have the ability to “pull the plug.” Cohen and Osborne predict that AIs more capable than humans could emerge by the end of the century, and this could be avoided with the right regulation.
23853
