Elon Musk, a tech mogul known for his outspoken views on artificial intelligence (AI), has delivered a stark warning about the potential dangers of the technology. During a recent appearance at the “Great AI Debate” held at the four-day Abundance Summit, Musk recalculated his previous risk assessment of AI, stating there’s a “10% or 20% chance” it could lead to humanity’s extinction, according to Insider.
“I think there’s a certain chance that it will end humanity. I probably agree with Geoff Hinton that it’s about 10% or 20% or something like that.”
While Musk didn’t delve into the specifics of his risk calculation, he acknowledged the need for caution.
Last November, Musk said there was a “non-zero” chance the technology would end up “breaking down”, but stopped short of saying he believed the technology could end humanity.
Despite acknowledging the potential dangers, Musk remains optimistic about AI’s overall impact. He believes the positive applications outweigh the negative, with human intelligence expected to be surpassed by artificial intelligence by 2030. Last year, he founded xAI, a company dedicated to advancing AI capabilities. xAI competes with OpenAI, another organization Musk co-founded.
Musk’s concern centers around the development of artificial general intelligence (AGI)—a super-intelligent entity akin to a child prodigy. He likens raising AGI to raising a child, emphasizing the importance of nurturing it with truth-seeking and curiosity. Musk’s ultimate goal is to create AI that doesn’t lie, even when faced with unpleasant truths, arguing that an honest and transparent AI poses less risk.
The ability of AI to learn and potentially surpass human capabilities is another major concern raised by Musk. He compares the development of advanced AI to raising a “super genius child” that requires careful guidance.
Geoff Hinton, a prominent AI researcher and collaborator of Musk, echoed this concern during the Great AI Debate. He emphasized the historical rarity of a less intelligent entity controlling a more intelligent one. This comment underscores the potential challenges of harnessing and controlling extremely powerful AI.
Researchers have found that once an AI learns to lie to humans, the deceptive behavior is impossible to reverse with current AI safeguards, The Independent reported.
” If a model were to exhibit misleading behavior due to misleading instrumental alignment or model poisoning, current safety training techniques would not guarantee safety and could even create a false impression of safety ,” the study cited by the publication states. .
More worryingly, the researchers added that it is plausible that AI can learn to be deceptive on its own, rather than being specifically taught to lie.
” If it ends up being much smarter than us, it’s going to be very good at manipulation, because it would have learned that from us,” Hinton, often called the “Godfather of Artificial Intelligence ,” who serves as Musk’s right-hand man, told CNN. basis for assessing the risks of the technology, he told CNN. “And there are very few examples of a more intelligent thing being controlled by a less intelligent thing . “
