LONDON: Artificial intelligence could have a 10% chance of causing the extinction of humanity within the next decade, according to a safety researcher working for AI company Anthropic.
Anthropic researcher Jacob Kosen warned that companies are racing to develop self-improving superintelligent AI systems, raising concerns about the potential risks to humanity. Kosen recently announced that he was leaving the company.
In a post on X, formerly known as Twitter, Kosen said companies were competing to develop systems capable of improving themselves without human intervention and warned that people were effectively “playing with our lives.”
Self-improving AI systems capable of autonomously enhancing their own abilities do not currently exist, but several AI companies are working toward increasingly advanced systems.
Kosen urged people not to underestimate the potential power of the technology. He warned that future AI systems could become powerful enough to hack almost anything, rapidly transform entire industries and acquire significant control and resources.
He also said AI systems were being developed with the belief that they could have the ability to kill everyone by the end of the decade.
Evan Hubinger, head of Anthropic’s science team, responded to Kosen’s warning by saying he believed Kosen’s concerns were valid. Hubinger said he personally estimated there was a 10% chance that AI could cause human extinction within a decade.
Hubinger added that while he believed Anthropic was making the best possible efforts to address the risks, the company did not currently have a plan for dealing with the potential dangers associated with superintelligent AI.
He said the risks posed by existing AI models were relatively low compared with the potential dangers from future superintelligent systems, adding that progress toward such technology was occurring faster than expected.
Anthropic had not issued an official statement on the claims at the time of publication.