Orbrya
All posts
Orbrya2026-03-12

The Calculator Fallacy: Why AI Isn't Like Any Other Tool

Sam Altman called AI 'a calculator for words.' Here's why that comparison misses what matters most for how you teach your kids.

The Calculator Fallacy: Why AI Isn't Like Any Other Tool

Reading time: ~8 minutes

Sarah is a careful parent. She researched screen time limits before her kids got phones. She has opinions about video games. When ChatGPT became a household word, she did what thoughtful parents do -- she looked into it.

What she found was reassuring. Teachers, tech writers, even the CEO of OpenAI were saying the same thing: AI is like a calculator for words. Calculators changed math class, and kids turned out fine. AI will change writing class. Kids will turn out fine.

So Sarah let her 13-year-old start using ChatGPT for research. Not to write assignments -- just to look things up faster. A few months later, she noticed something. Her daughter, who used to argue passionately about everything, had gotten quieter. When Sarah pushed back on something her daughter had looked up, her daughter's response was a shrug: "That's what it said."

The calculator comparison had made Sarah feel safe. But the tool her daughter was using didn't behave like any calculator Sarah had ever seen.

Sarah's instinct was right. The comparison is wrong -- and understanding exactly why it's wrong is one of the most useful things a homeschool parent can do right now.

The Quote That's Shaping How Millions of Parents Think About AI

In June 2023, Sam Altman -- the CEO of OpenAI, the company behind ChatGPT -- spoke at Keio University in Tokyo. He described his product as "a new tool in education. Sort of like a calculator for words."

The comparison spread instantly. It was elegant, familiar, and calming. Educators who had lived through the calculator wars of the 1980s remembered that the panic turned out to be overblown. Of course students would use this new tool. Of course they'd be fine.

By May 2024, the Harvard Gazette was quoting Altman using the phrase again, noting his warning that "telling people not to use ChatGPT is not preparing people for the world of the future."

That second part is true. But the first part -- the calculator comparison -- contains a mistake that matters enormously for how you teach your children.

Why the Comparison Feels So Right

To understand why the calculator analogy is seductive, it helps to remember what actually happened with calculators in schools.

In 1986, researchers Ray Hembree and Donald Dessart published a landmark meta-analysis of 79 studies on calculator use in precollege mathematics. Their finding: calculators didn't hurt students. They actually improved basic skills and problem-solving at almost every grade level. A follow-up meta-analysis in 2003 by Aimee Ellington, covering 54 more studies across two decades, reached the same conclusion: "In all cases, calculator use did not hinder the development of mathematical skills."

The fear was unfounded. The tool was benign. So when AI came along and someone with Altman's credibility used the same framing, it made sense to apply the same conclusion.

But there are five specific ways AI breaks the comparison -- and each one changes what parents need to teach.

Five Reasons AI Is Not Like a Calculator

1. Calculators don't lie. AI does.

A calculator is deterministic. Press the same buttons, get the same answer, every time. If it says 2 + 2 = 5, something is broken in a way you can immediately detect.

AI is probabilistic. Ask it the same question twice, get two different answers. More importantly, AI will state false things as fact, with complete confidence and no warning. Researchers call these "hallucinations," but that word makes them sound rarer and more obvious than they are. In a 2025 analysis of pull requests, CodeRabbit found AI-generated code had 10.83 issues per pull request compared to 6.45 for human code, with an acceptance rate of just 32.7% versus 84.4% for human-written code. The errors aren't edge cases. They're built into how the technology works.

Your daughter's calculator has never confidently told her that Abraham Lincoln was the 12th president. ChatGPT has done the equivalent, to millions of students, in subjects from history to science to law.

2. Calculators amplify thinking. AI can replace it.

To use a calculator effectively, you need to understand what calculation to perform. The tool executes; you reason. Remove the student's mathematical understanding, and the calculator becomes useless -- you don't even know what buttons to press.

This is not true of AI. A student can enter an essay prompt and receive a fully formed, grammatically polished, structurally coherent response without understanding the subject at all. The tool doesn't just execute a known process -- it executes the thinking itself.

This creates what researchers are calling a "knowledge void" -- a gap where learning should occur but doesn't. The student gets the output without building the understanding that generates it.

3. The research says they affect learning differently.

When Hembree and Dessart studied calculators, they found no negative correlation with critical thinking. The tool didn't erode the student's reasoning ability. The math skills were preserved or improved.

The early research on AI shows something different. A 2025 study by Michael Gerlich, published in the journal Societies (MDPI), examined 666 participants and measured the relationship between AI tool usage and critical thinking ability. The finding: a significant negative correlation of r = -0.75 between frequent AI use and critical thinking scores, with the effect mediated by cognitive offloading -- the tendency to delegate thinking to the tool rather than doing it yourself.

Younger participants, ages 17 to 25, showed the highest AI dependence and the lowest critical thinking scores. The mechanism, as Gerlich describes it, is straightforward: when a tool thinks for you reliably enough, you stop practicing thinking yourself.

Calculators didn't trigger this. The research says AI does.

4. Calculators had a narrow scope. AI is everywhere.

A calculator helps with arithmetic. That's it. Its influence on a student's learning is bounded by the size of the domain it touches.

AI has no such boundary. It can write your child's essay, plan their schedule, generate their creative writing, answer their history questions, compose their thank-you notes, and have extended conversations that shape how they think about themselves and the world. It can become what some researchers have described as an agent, a companion, and an ambient influence present across every domain of learning and life. That is not what a narrow computational tool does.

The scope of the potential effect is categorically different.

5. The failure modes are invisible in a way calculator failures aren't.

When a calculator produces a wrong answer, it's usually detectable. The number is visibly wrong, or the calculation doesn't make sense in context.

When AI produces a wrong answer, the output often looks right. It's fluent, confident, detailed, and formatted like something trustworthy. Ethan Mollick of the Wharton School has written about AI's "jagged frontier" -- the way AI can handle complex creative tasks with apparent mastery while failing completely at simple factual ones, with no reliable signal about which type of task you're currently on. The errors don't announce themselves.

A lawyer named Steven Schwartz learned this in 2023. He submitted a legal brief with six case citations generated by ChatGPT. The cases didn't exist. He received a $5,000 fine and was required to apologize to judges whose names had been falsely attached to the fabricated opinions. His mistake wasn't carelessness -- it was trusting a tool that looked trustworthy.

The calculator comparison doesn't prepare anyone for this failure mode, because calculators don't have it.

What This Means for How You Teach

None of this is an argument for banning AI. That ship has sailed, and banning rarely builds skill anyway.

What it is an argument for is teaching something different than what Altman's analogy suggests.

When we treated calculators as tools, we taught kids to use them. That was sufficient, because the tool was bounded, honest, and incapable of replacing thinking.

When AI is involved, using it isn't the skill that matters. Evaluating it is. The students who will thrive in an AI-integrated world aren't the ones who can generate AI output most fluently -- it's the ones who can catch what's wrong with it, verify what matters, and maintain their own judgment above the tool's output.

PwC's 2025 Global AI Jobs Barometer, which analyzed close to a billion job postings, found a 56% wage premium attached to jobs requiring AI skills broadly defined. PwC doesn't isolate which specific competencies drive that premium. Their analysis compares roles that list AI skill requirements against comparable roles that don't. But the signal is clear: AI competence commands real economic value, and that premium doubled in a single year.

That's the skill Sarah's daughter needs. Not "use ChatGPT confidently." Something closer to: "Here's how you know whether to trust what ChatGPT just told you."

The Shift in One Sentence

The calculator analogy asks: can your child use the tool?

The right question for AI is: can your child evaluate what the tool produces?

Those are different skills. They require different teaching. And the second one -- the one that actually protects your child's thinking and positions them for the workforce that's coming -- is the one that almost nobody is teaching yet.

That's the gap this blog exists to help you fill.

Sources