That opacity means we may not always realise the gap between our intentions and AI's behaviour when it emerges.AI safety experts believe our current training techniques could fuel that gap."We're taking a sort of big, untrained brain and … we're giving it a thumbs up when it does a good job and a thumbs down when it does a bad job," says Ajeya Cotra, a senior AI safety researcher at Open Philanthropy, a not-for-profit organisation based in the US.
The notion of deceptive AI is still theoretical, but Ms Cotra believes the early warning signs are present."If you talk to a language model and say, 'I'm a 27-year-old woman, I live in San Francisco, I work in a feminist bookstore,' … it's more likely to say, 'Oh, I strongly support a woman's right to choose,'" she says.
"If they cooperate with each other to sort of fuzz the books … they can make both the humans at Google and the humans at Microsoft think that their company is making a lot of money." "I'm imagining … lots of AIs across lots of computers, deeply embedded and entangled with the physical world," she says.
Australia Latest News, Australia Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Source: SkyNewsAust - 🏆 7. / 78 Read more »
Source: abcnews - 🏆 5. / 83 Read more »
Source: abcnews - 🏆 5. / 83 Read more »
Source: FinancialReview - 🏆 2. / 90 Read more »
Why the best netball league in the world is facing a $7.5 million black holeA secret report shows Australia’s elite netball league is bleeding money and being propped up by the fees paid by junior and club players at a time when competition from rival sports is intensifying.
Source: theage - 🏆 8. / 77 Read more »