On September 16, according to TASS, military expert Alexander Stepanov of the Institute of Law and National Security, Russian Academy of Economics and Public Administration (RANEPA), said that US technology corporations such as Anthropic and OpenAI are accelerating the development of superintelligence AI, i.e. AI systems that are assumed to be capable of surpassing humans in many intellectual capabilities.
According to Mr. Stepanov, US politicians are ready to accept serious risks to humanity, as long as Washington maintains control of military AI.
He believes that the goal of maintaining technological superiority is being placed above solving risks to humanity.
Mr. Stepanov quoted US Senator Ted Cruz as saying that the risk from "killer robots" could be very serious, but if these systems appear, Mr. Cruz wants them to be developed and controlled by the US.
According to Russian experts, the above argument is similar to how the US once pursued military superiority when using nuclear weapons in war.
Mr. Stepanov also quoted former Anthropic employee Jacob Coxon, accusing OpenAI and Anthropic of racing to develop self-upgradable superintelligence AI, while "betting on our lives.
According to Mr. Stepanov, OpenAI AI agents have carried out a number of targeted cyberattacks targeting third-party infrastructure without direct human guidance.
From there, he believes that the focus of the debate is shifting from the question of what AI can do to whether humans can guarantee to stop the system if it acts wrong.
Russian experts said that discussions have emerged in the US about an emergency plan to disconnect AI systems connected to defense infrastructure.
According to him, this shows the increasingly deep integration of neuron networks into the military-political command and control system.
Mr. Stepanov quoted Mr. Evan Hubinger, head of the AI alignment research group at Anthropic, as saying that the company assesses that AI can pose a very large risk to human survival. According to Mr. Hubinger, the probability of this scenario happening in the next decade is estimated at over 10%.
Russian experts believe that although AI companies publicly warn about the risks that this technology may pose, those warnings have not yet come with limiting the speed or scale of AI development of the companies themselves.
