Artificial intelligence could become more intelligent than the combined capabilities of all humans within the next five years, Elon Musk has predicted. He warned that the technology’s rapid advancement may be impossible to stop.
The world’s richest man made the projection during an interview with The Economist published on Thursday, saying AI would eventually outperform humans in almost every field, with the exception of the human experience itself.
He said the development of advanced AI would most likely usher in an era of unprecedented abundance rather than destruction, allowing people to access solutions to almost any material need. Musk told The Economist that he believed AI could exceed the sum of human intelligence in around five years, and that there would likely be nothing AI couldn’t do better than humans, apart perhaps from being human. He added that the most likely outcome was an age of amazing abundance in which anyone could have anything they could think of, acknowledging that the idea might sound far-fetched but suggesting it would be worth revisiting by 2036.
The Tesla and SpaceX chief executive said the momentum behind AI development had grown beyond the control of any individual or organisation, including himself.
Musk, who has previously raised concerns about the dangers of artificial intelligence, said the risks remained but that he had accepted the technology’s continued advancement. He explained that although he still believed there was non-zero risk associated with AI and robots, he had reached a philosophical conclusion to focus on the positive, saying he could see no way to genuinely halt the momentum behind AI and robotics. He said his current philosophy was to enjoy the ride, describing the progress in AI and robotics as seemingly inexorable, to the point that even if he wanted to stop it, he would be unable to.
Responding to whether he still believed AI had a 10 to 20 per cent chance of causing human extinction, Musk said the possibility remained but his approach to the threat had changed.
He also called for leading AI companies to hold regular discussions on safety and security, suggesting weekly or bi-weekly meetings where competitors could review new models before their release. He argued that competing companies could help keep each other honest, and that rival firms were better placed than governments to determine whether an AI system posed a security risk. He urged the industry to implement such measures quickly, noting that the pace of innovation meant delays could be costly, pointing out that AI breakthrough announcements were emerging constantly, sometimes multiple times a day.
Musk’s prediction, however, has been questioned by some AI experts, who argue that existing systems still do not fully understand or replicate the complexity of human intelligence.
