the artificial intelligence race has entered a point of no return, that the problem is no longer just how to organize future models, but how to deal with powerful systems that already exist and have spread beyond the complete control of their developers.
Hot Topics
5 red and purple foods that may help support heart health
A Chinese robot and a Russian robot attack humans Has the rebellion begun?
Under the title "It is already too late to stop the threat of artificial intelligence," the well-known writer presents a deep forward-looking reading of the crisis-ridden reality that AI governance policies have reached, highlighting the state of bureaucratic confusion and the lack of a unified vision among decision-makers in dealing with a technology whose characteristics exceed the limits of traditional tools to become a new living species that poses an existential and institutional threat to all of humanity.
Friedman begins with a paradox he describes as "chaotic" in US and Chinese policies toward artificial intelligence. Companies are calling for a slowdown in development, but they link this to what China will do, while the competition for technological leadership remains a powerful incentive to continue the race.
In contrast, he cites a statement by US President Donald Trump in which he asserted that the only "precautionary measures" needed to control artificial intelligence are a "strong and intelligent president".
In his article, Friedman reviews scenes from the turbulent political theater and the “laughable and tragic” reflections of statements by some leaders who adopt a superficial understanding based on direct individual leadership rather than organizational structures and solid safety barriers.
But the core of Friedman's argument is that relying solely on future models is no longer sufficient. He draws on the analysis of Craig Mundy, former head of research and strategy at Microsoft , to establish four pillars for understanding the risk.
Firstly, according to Friedman, artificial intelligence has become a "four-way" technology. The problem is no longer that humans can use the tool to build or destroy, but that the tool itself may become capable of making decisions and carrying out actions independently.
The author cites evidence of recently documented hacking operations attributed to artificial intelligence agents, such as a group of AI agents from the company "OpenAI" hacking the "Hugging Face" platform of their own volition, and the use by "rogue" entities such as the Iranian- backed Houthis in Yemen of advanced models to target naval vessels and develop missiles. Friedman argues that the threats are no longer theoretical, but have manifested in actual reality in the form of "quadruple use" in which systems make their decisions with complete independence.
This, in turn, leads—according to the article—to the second point, which Mundy summarized in three words: "It's too late." Open-weight models have already proliferated, and models owned by commercial companies could fall into the hands of rogue actors and unregulated states.
The author cites as evidence Anthropic ’s report on attempts related to biological weapons research, and the use of the “cloud” model in gathering military information and working on missile guidance systems, to say that dangerous artificial intelligence capabilities.
Tags:
technology
