Nvidia CEO Jensen Huang has offered a remarkably blunt answer to a question sitting underneath the artificial-intelligence boom: what happens if people stop retaining basic skills because software does more of the work for them?

His answer, broadly, is that some of those skills may simply no longer matter.

During an interview, Huang was presented with research from China suggesting that AI tools can help children complete homework faster while the intended learning does not necessarily stick. The concern was not merely that students were using a shortcut. It was that later quiz and test performance appeared to decline as retention weakened. Huang agreed that this could be happening, but argued that the loss of familiar mathematical skills was not necessarily a problem.

“Try to get a kid to do long division right now,” Huang said. “The multiplication table is starting to be forgotten. Doing square roots, my goodness. Basic math is being forgotten. Does it matter? I don’t think it does. I don’t think it does.”

That is an unusually clear expression of the trade-off AI boosters often imply without stating outright. If a tool is always present, fast, and good enough, perhaps people need not carry out the underlying task themselves. Huang’s view is that society will develop other, more valuable abilities instead. He described the possible loss as a reduction in “finer intellectual dexterity,” offset by people becoming better “systems thinkers.”

For anyone who plays games, makes games, or has watched a live-service title turn its menus into a small tax code, the central issue should be familiar: convenience is not the same thing as comprehension. An assist can lower friction. It can also remove the feedback that teaches a player why an outcome happened.

The address anecdote makes the argument tangible

Huang did not frame the point only as a prediction about students. He volunteered that his own reliance on AI and other technology has left him unable to recall basic personal information. He said he does not know his home address or telephone number, and remembered panicking when asked for a ZIP code while buying gas a few years ago.

“There are a lot of skills that don’t matter,” Huang said. “I actually don’t know my address.”

In one sense, this is a mundane adaptation. Phones store contact information, mapping apps route people home, and autofill handles addresses all day long. The system works until it does not: a dead device, an unavailable service, an unfamiliar form, or a moment when another person needs information that cannot be retrieved immediately.

That does not mean every fact must be memorized forever. It does mean that deciding a skill is obsolete requires more care than noticing that an app can perform it. A person may be able to live comfortably without remembering a ZIP code. It is harder to infer from that example that foundational arithmetic has become similarly optional, particularly when the cited research raises concerns about whether students are learning from AI-assisted homework at all.

What “systems thinking” can mean—and what it cannot automatically solve

Huang’s phrase “systems thinkers” is broad, but it points toward a real category of ability: understanding how connected parts affect one another, identifying constraints, and choosing among tools or processes rather than mechanically performing every step by hand.

In games, systems thinking might describe someone who recognizes how an economy, crafting loop, character build, combat timing, map knowledge, and team composition interact. A player need not calculate every value from scratch to understand that spending one scarce resource now can restrict later options. A designer likewise needs to see how a seemingly minor balance adjustment can affect progression, player behavior, and the wider game economy.

Those are useful skills. But the term should not become a magic spell that makes basic knowledge unnecessary. Systems thinking depends on being able to assess inputs, question outputs, and understand enough of a process to notice when something has gone wrong. If an AI assistant gives an answer, the user still has to decide whether the answer is plausible, applicable, and complete.

That distinction is especially important with generative AI, meaning tools that produce new text, images, code, or other material in response to prompts. These systems can produce an answer that looks confident and polished without making their reasoning easy to inspect. Asking a chatbot for the result of a calculation is not the same as understanding how the calculation works. Asking it to explain its result may help, but it still shifts the learner’s role from doing and checking to accepting and auditing.

Auditing is a skill in its own right. Yet it is difficult to audit what one cannot independently evaluate.

Games already demonstrate both sides of assistance

Video games are full of tools that reduce tedious work without necessarily hollowing out play. Navigation markers, accessibility options, difficulty settings, aim assistance, crafting recipes, quest logs, and detailed tutorials can all make a game more approachable. Their value is not in demanding that every player memorize everything or perform every action the hard way. Their value is in helping more people participate.

At the same time, games can illustrate why removing every bit of productive friction can make an experience thinner. A quest marker may prevent aimless wandering, but it can also eliminate the act of reading a world and forming a route. A build guide can get a player to an effective loadout quickly, but blindly copying it does not teach why particular stats, abilities, or equipment work together. An automated strategy suggestion can be useful; treating it as infallible leaves the player stranded when the situation changes.

The question is not whether assistance is bad. It is what remains for the player or learner after assistance has completed its task.

There is a difference between a tool that supports understanding and one that quietly substitutes for it. A calculator can speed up arithmetic after a person understands the operation. An AI tool that supplies a finished homework response may save time, but the time saved is not automatically converted into learning. The research described to Huang matters precisely because it suggests the latter outcome may be occurring: faster completion alongside poorer retention and test performance.

Why basic math is not just about arriving at a number

Long division, multiplication tables, and square roots are easy targets because many adults remember them as slow, repetitive classroom exercises. And it is fair to ask whether every old instructional routine deserves preservation unchanged. But Huang’s examples are not solely about getting correct numbers without a machine.

Basic math builds habits of estimation, pattern recognition, and error detection. Someone who knows that four groups of three equal 12 has a tiny but useful reference point for judging whether a later answer makes sense. Someone who can estimate a cost, compare quantities, or recognize an impossible result has a layer of independence from whatever tool they happen to be using.

That is relevant in games as much as daily life. Players frequently make quick decisions using partial information: whether an in-game price is worthwhile, whether an upgrade is meaningful, whether a percentage bonus actually changes a build, or whether a claimed damage calculation sounds credible. Nobody needs to hand-calculate every system in every game. But players who understand the basics are better equipped to spot misleading advice and question unexplained outputs.

This is also why an opaque answer can be less useful than a slower, understandable process. In technical terms, opacity means the user cannot readily see how a result was generated. The more opaque a tool is, the more the user needs independent knowledge to evaluate it. Delegating basic tasks may be harmless in many moments. Delegating judgment is a much larger concession.

The practical question is dependency, not nostalgia

It would be simplistic to respond to Huang by insisting that all information must live in human memory and all work must be done manually. Technology has always externalized some memory and labor. Digital calendars remember appointments. GPS systems hold maps. Game wikis document labyrinthine quest chains and hidden mechanics. The existence of a useful tool is not a moral failure.

But “the tool exists” and “the underlying skill does not matter” are not equivalent claims. The former describes capability. The latter makes a judgment about resilience, learning, and who remains in control when an automated result needs to be challenged.

For parents, teachers, and students navigating AI-assisted homework, one practical distinction is whether the tool is being used before or after an attempt. Using an assistant to explain a missed step, generate extra practice questions, or identify an error can preserve the learner’s active role. Using it to produce the finished work before the learner has grappled with the problem risks turning education into an exercise in output delivery.

For players, the same principle can be applied without turning games into homework. Use a guide after trying to understand the encounter. Ask for an explanation of why a build works rather than only a list of items. Treat automated recommendations as a starting point, not as an oracle. Those approaches preserve the convenience while keeping some ownership of the decision.

The games industry is already wrestling with the broader version of this issue: where AI fits in creative and technical work, and whether speed gains come with losses that are harder to measure. That conversation belongs beside wider questions of industry durability and the pressure to chase the next major technology shift, as discussed in this look at why games need more than the next big hit.

AI can be a tool without becoming the thinker

Huang is right about one limited point: human skill priorities do change. People regularly stop practicing tasks when technology makes those tasks less common. There is no reason to romanticize unnecessary drudgery, and there are genuine benefits in tools that remove barriers and free time.

But the available information here offers no reason to treat weakened retention as automatically harmless. The research Huang was shown points to a real educational tension: students may finish work faster while retaining less of what that work was meant to teach. His personal story about forgetting his address demonstrates the convenience of outsourcing memory, but also the brittleness of that arrangement when the usual support system is unavailable.

“Systems thinking” may be a valuable aspiration. It should mean understanding relationships, consequences, and trade-offs across a whole process. It should not mean being unable to perform, explain, or verify any part of that process without a device in hand.

The better goal is not a future where everyone does long division for its own sake, nor one where people must memorize every number their phones can store. It is a future in which AI handles genuinely routine work while people retain enough knowledge to test its claims, recognize its failures, and decide when the machine’s answer is not good enough. In a classroom, in a game, or at a gas pump, that remains a skill that matters.