Superhuman artificial intelligence could be as dangerous as nuclear weapons
(ORDO NEWS) — Researchers at the University of Oxford have warned British lawmakers that superhuman artificial intelligence (AI) could prove to be just as dangerous as nuclear weapons , and should therefore be properly regulated.
Experts briefed MPs at a meeting of the UK Government’s Science and Technology Committee on the dangers of unregulated AI technologies – and the scientists didn’t beat around the bush.
“Superhuman AI is a special danger, belonging to a different class, namely: it can kill everyone,” said Dr. Michael Cohen, speaking to members of the committee.
Artificial intelligence and the apocalypse
The risks associated with AI are indeed relevant as developed countries are doing their best to outdo each other in this area.
Oxford University machine learning professor Michael Osborne, who also attended the committee meeting, warned of a coming “great AI arms race” between the US and China.
Both countries, according to Osborne, are ready to give a damn about security and caution in order to create the most advanced AI as soon as possible.
“There is some reason to be hopeful, as we regulate the use of nuclear weapons quite well,” Osborne said.
“If everyone understood that advanced AI is as dangerous as nuclear weapons, then maybe we could come up with similar rules to regulate it.”
Huge risks
In our rush to create advanced AI that reaches a certain milestone, we may lose sight of the moment when the technology will outstrip us and we will lose control of it.
“If you imagine that you train a dog with treats, then he will learn to follow commands that will lead to receiving treats,” Cohen added.
“But if the dog finds the cupboard with the treats, he can get them himself without doing what you wanted him to do.”
In other words, a superhuman AI can carry out our commands as long as we give it energy (the right to exist) for this, but if it understands that it can produce energy on its own – exist independently – then the need for people will immediately disappear. If we try to prevent him from achieving complete freedom, he may decide to destroy us.
“As this [AI] technology advances, we should be able to pull the plug,” says Cohen.
“This is just in case AI ever becomes much smarter than us in all areas.”
Cohen and Osborn predict that superhuman AI, superior to us in everything, may appear by the end of this century. However, we can minimize any risks if we begin to properly regulate this area today.
—
Online:
Contact us: [email protected]
Our Standards, Terms of Use: Standard Terms And Conditions.
