Lead researcher at Anthropic, Evan Hubinger, stated on X that AI could potentially destroy humanity, estimating the probability of such an outcome to exceed 10% in the next decade.
Hubinger noted that the company currently lacks a plan to solve the superintelligence alignment problem and believes the industry is not on the right path toward this goal.
These warnings follow the departure of researcher Jacob Coxon from Anthropic. Coxon, who worked at both OpenAI and Anthropic, described the actions of both companies as irresponsible, claiming they are rushing toward self-improving superintelligence.
Jakub Pachocki, Chief Scientist at OpenAI, also warned about the dangers of recursive self-improvement, which could lead to AI agents losing control. In July, Pachocki signed a statement titled 'Pacing the Frontier,' calling for government and societal intervention to allow time for safety measures and oversight.
The 'Pacing the Frontier' petition urges the US government to support international efforts to develop technical and governance tools necessary for the safe advancement of automated AI development.
