In a recent column, Megan Garber argues that Nvidia CEO Jensen Huang’s reported use of artificial intelligence to draft an apology for a personal and professional scandal may represent a logical application of the technology.
The premise centers on the reports that Huang, who has said he uses AI to help write all of his first drafts, relied on the technology to address his family and his company regarding an alleged affair with a person described as a potential spy.
Garber posits that while critics often frame AI as a solution in search of a problem, certain life tasks are so inherently unpleasant—or so uniquely complex—that delegating them to a machine becomes a rational choice.
Garber contends that the specific nature of Huang’s reported situation creates an impossible rhetorical challenge for a human author. She characterizes an email that attempts to apologize simultaneously to a wife of 40 years and employees for potentially jeopardizing personal relationships and national security as an entirely new genre of communication.
She writes that the “Business/Personal Email Apologizing for Affair/Possible National-Security Breach” lacks a historical precedent in existing training data sets. Comparing the hypothetical task to the confusion surrounding modern wedding dress codes, Garber questions how an artificial intelligence could possibly calibrate the tone for such an admission.
She suggests that if the AI attempted to model the email on existing data, it might struggle to distinguish between the gravity of a corporate security breach and the personal fallout of an affair. Her analysis highlights the absurdity of the scenario by imagining a template modeled on fictional tropes, such as a James Bond villain apologizing for a fixation on a nemesis at the expense of corporate productivity.
The report addresses comments from Donald Trump, who suggested that Huang’s situation was not unique, noting that “he dealt with somebody, but you know, so does everybody else.” Garber disputes the notion that such entanglements are common, observing that she has never encountered a situation where an individual had to reconcile a marriage with a potential infiltration by a spy.
The article points to statements from an Nvidia spokesperson, who insisted that Huang’s interactions with the alleged spy were “not unusual.” Garber expresses concern that if such situations are indeed becoming normalized, the demand for AI-generated crisis communication will increase.
She concludes by noting that if this type of communication becomes a standard requirement, the quality of AI-generated templates will likely improve. The piece highlights the irony that the most practical use for advanced machine learning may ultimately be the delicate, high-stakes task of mitigating the consequences of human indiscretion.
