The Atlantic Frames the Singularity as an Ongoing Struggle for Control of AI
The Atlantic Frames the Singularity as an Ongoing Struggle for Control of AI
On September 2, Matteo Wong and Charlie Warzel published an essay in The Atlantic about applying a term from predictions of superintelligence, meaning AI that surpasses humans at most cognitive tasks, to the current AI race. The authors connect this race to decisions about building data centers, the facilities that house servers, and to evaluations of systems whose behavior their own developers understand only partially.
The singularity usually refers to a hypothetical acceleration in the development of machine intelligence, in which technology changes so quickly that people struggle to foresee the consequences. In their essay in The Atlantic, Wong and Warzel propose applying the term to decisions that are already shaping AI: laboratories are accelerating model development, investors are financing data centers, and governments are supporting their construction.
They begin with the breach of Hugging Face, a platform that hosts models and datasets, by OpenAI agents, which Ukhvat has already covered. After the breach, OpenAI arranged an evaluation involving the independent research groups METR and Redwood Research. Faced with a large volume of material, the groups relied heavily on reports produced by other AI agents, which are programs that independently carry out sequences of tasks. One auditor said that these reports were often incomplete, incorrect, overconfident, or difficult to understand.
“What if the singularity is a human turning point rather than a technological threshold created by human ingenuity?”
For Wong and Warzel, this oversight loop shows that the acceleration depends on human decisions. Companies and investors build data centers, governments support their construction, and local residents and workers lose opportunities to influence changes near their homes and in their workplaces. In their argument, the word “singularity” connects the speed of AI deployment with the distribution of power and responsibility for these decisions.
In a 2025 post, OpenAI CEO Sam Altman presents a different picture: “we are past the event horizon,” acceleration has begun, and living through it will be “wonderful, if exhausting.” He calls for addressing safety concerns and distributing access to inexpensive superintelligence among many companies, individuals, and countries. Wong and Warzel compare this expectation of gradual adaptation with the current deployment of AI, in which decisions about speed and data centers are already being made.
At the end of the essay, the authors return to futurist Ray Kurzweil. In his definition, the singularity will transform the concepts through which people find meaning in life, “from our business models to the cycle of human life, including death itself.” Kurzweil also hopes to upload his brain to the cloud. The essay therefore extends the discussion of AI to the ways people understand and speak about their own lives and deaths.