NVIDIA CEO Jensen Huang: Forgetting Basic Math is No Big Deal
NVIDIA CEO Jensen Huang stated in an interview with The New York Times regarding a study of Chinese students that he does not consider it a problem that children are forgetting long division, multiplication tables, and square roots due to the use of artificial intelligence.
The study covered approximately 26,000 Chinese students. The host pointed out that while AI can accelerate homework completion and improve grades, knowledge retention is insufficient, leading to declines in test and exam performance. Huang acknowledged that poor retention is likely occurring among children who use AI extensively, but he clearly stated that he is not worried about it.
Huang cited examples of how difficult it is for children to perform long division now, that multiplication tables are starting to be forgotten, and square roots are even less remembered, indicating that basic math is being forgotten. When pressed on whether there are still essential skills that must be retained, he agreed that important skills exist but questioned whether basic math is among them, stating that people will discover new skills.
He then used himself as an example, admitting that many skills are not actually important. He confessed that he cannot remember his home address, panicked when asked for his postal code while refueling a few years ago, and does not remember his phone number, adding that his wife can attest to this, and he can continue living with these forgettings.
This statement does not entirely align with his previous public remarks. In 2025, he advised young people to continue learning math, reasoning, logic, and programming, emphasizing that even if they do not need to do it themselves, they should master the ability to break down unfamiliar problems using first principles; in 2024, he repeatedly stated that programming would no longer be a necessary condition for success. This interview directly categorizes basic calculations that can be outsourced to models as skills that do not need to be retained.
From a market mechanism perspective, this is an event-driven narrative release: those selling computation and models are reducing society's obsession with the idea that "the human brain must master basic calculations" to allow education, office work, and content production to more quickly delegate tasks to reasoning systems. The ones paying are cloud vendors, model companies, and school purchasers that require training and reasoning capacity; the ones under pressure are traditional educational support and assessment institutions that focus on repetitive training of manual calculations and exam skills. Funding continues to flow into chips, data centers, and applications that can do homework for children, rather than into human training that reintroduces multiplication tables into the classroom. Whoever frames "forgetting basic calculations" as an acceptable civilizational evolution is providing cultural permission for the next round of computing orders.
Source: Public Information
ABAB AI Insight
Huang himself entered chip design through an undergraduate degree in electrical engineering and a night master's from Stanford, having worked at AMD and LSI Logic early in his career before co-founding NVIDIA in 1993. He has publicly stated that math was his favorite subject and has emphasized grades. A person who considers linear algebra and architecture as the foundation of his career now categorizes long division and multiplication tables as skills that do not need to be retained, indicating that his judgment is not based on children's cognitive science but rather on whether NVIDIA's next order book should continue to expand.
The capital path is clear. NVIDIA's market value is built on one thing: society is increasingly delegating symbol operations that were once performed by the human brain to models on GPUs. Education is an untapped reasoning market. When children's homework is handed over to models, schools purchase AI tutoring, and parents pay for "faster homework completion," these will all turn into tokens and computational consumption. Declaring basic math as a forgettable skill equates to rewriting K12 from "training the human brain" to "calling models," shifting resources from teacher hours and educational materials to clusters, networks, and software subscriptions.
Historical comparisons are not new. After the widespread adoption of electronic calculators, schools debated whether to continue practicing manual calculations; after navigation became common, urban residents no longer remembered routes; after spell check became widespread, spelling accuracy declined but writing output increased. Each time follows the same structure: outsourcing automatable foundational processes and elevating scarce attention. The difference is that calculators do not require continuous purchases of power-level computing from the same company, whereas large models consume chips and cloud resources from NVIDIA's ecosystem each time they solve a problem. Just as OpenAI outsources writing and Google outsources retrieval, NVIDIA aims to outsource computation itself, positioning itself at the revenue end of outsourcing agreements.
This belongs to technological substitution, where the mechanism is the shift of pricing power for cognitive processes. When basic calculations transition from "a skill everyone must internalize" to "a service that can be called upon at any time," the scarcity of ability shifts from memory and proficiency to who masters the model, who pays for reasoning, and who defines new skills. Classrooms no longer pay for multiplication tables but begin to pay for interfaces and computational power. Forgetting is not a side effect; it is a necessary step in shifting demand from human brain training to machine reasoning.
ABAB News · Cognitive Laws
- Whoever sells computational power declares that the human brain no longer needs to calculate.
- The stronger the tools, the more basic skills are declared obsolete.
- The forgetting list is written for the public, while the order list is kept for oneself.