‘I don't save for retirement now’: Former Google AI expert's alarming warning about artificial intelligence

A former Google and Microsoft engineer has stopped contributing to his retirement savings, saying he believes artificial intelligence (AI) could wipe out humanity within the next decade.

Nate Soares, a machine learning researcher and president of the Machine Intelligence Research Institute (MIRI), estimates that the probability of AI-driven human extinction over the next 10 years is higher than 50%.

Speaking on a recent podcast hosted by Nicholas Thompson, CEO of The Atlantic, Soares said he was increasingly concerned that humanity might not have much time left to address the risks posed by advanced AI systems.

“I guess right now it looks like we probably don’t have a decade on the clock,” Soares said, adding that there was “more than 50%” chance that humanity did not have another decade.

His concerns have also influenced his personal financial decisions. Soares said he stopped contributing to his retirement savings around a decade ago because he questioned whether he would live long enough to need the money.

“I do not save for retirement now,” he said. “I have a 401K from when I worked at Google, but I haven’t contributed to it for like a decade.”

Who is Nate Soares?

Soares currently serves as president of the Machine Intelligence Research Institute, a Berkeley-based nonprofit that focuses on the risks associated with advanced artificial intelligence.

The organisation is associated with the AI safety movement and concerns often described as “AI doomerism” — the belief that the development of superintelligent AI could pose an existential threat to humanity.

Soares has argued that AI systems designed to match or surpass human intelligence could eventually become difficult for people to control, potentially putting human survival at risk.

Why does he believe AI could wipe out humanity?

During the podcast conversation, Thompson questioned Soares' assessment, pointing to the limitations and shortcomings that remain in even the most advanced AI systems available today.