You vs ChatGPT: Who Would Actually Win?
| Fremont, California
Your opponent has read millions of books, can solve calculus problems in seconds, writes essays almost instantly, and never gets tired.
Sounds impossible to beat, right?
Thankfully, your next opponent is not an Olympic gold medalist or a Nobel Prize winner.
It’s a chatbot.
Over the past few years, ChatGPT has emerged as the unofficial study partner for millions of students. It’s explaining chemistry homework at midnight, debugging code five minutes before it’s due, translating Spanish assignments, and somehow making calculus seem a little less terrifying. Whether teachers love it or hate it, one thing is certain: AI has changed the way students learn.
However, after seeing ChatGPT answer everything from SAT test questions to programming assignments, I couldn’t help but wonder:
If you and ChatGPT were to take the exact same high school exam, who would really end up getting a higher score?
At first, the answer seems obvious.
After all, ChatGPT can talk about black holes, solve algebra, analyze literature, write essays, and even generate computer code in a matter of seconds. Humans simply can’t process information that quickly.
So… game over?
Not quite.
The biggest misconception about ChatGPT is that it “knows” things.
It doesn’t.
Unlike people, ChatGPT doesn’t remember information the way humans do. Instead, it predicts what words are most likely to come next based on patterns it learned from enormous amounts of text. Imagine reading billions of books, articles, conversations, and websites, not to memorize every page, but to become unbelievably good at recognizing patterns in language. That’s much closer to what ChatGPT is doing.
Because of that, it can sound incredibly intelligent. Most of the time, it gives accurate and well-organized answers that are easy to understand. However, sometimes confidence can be misleading.
Researchers use the term “AI hallucination” to describe moments when AI confidently provides information that simply isn't true. In other words, ChatGPT may invent a scientific paper, misquote someone famous, or explain a concept incorrectly while sounding completely certain. The answer is polished, but polish doesn’t guarantee accuracy.
Imagine confidently writing one of those answers on your biology test.
Suddenly, relying on AI seems risky.
Of course, humans make mistakes too. We forget formulas, confuse historical dates, and occasionally convince ourselves that option C “just feels right”. The difference is that when humans lack information, they often acknowledge their uncertainty. ChatGPT doesn’t experience doubt. It just keeps producing the most statistically likely response whether it is right or wrong.
In fact, some studies have suggested ChatGPT can perform surprisingly well on standardized exams, including portions of the SAT, law school, and even medical licensing exams. That is impressive, but passing a test isn’t the same as understanding the material.
So what happens if we compare subject by subject?
In the case of a simple memorization test, ChatGPT will probably dominate. Definitions, vocabulary, historical facts, and scientific concepts are all areas where AI performs exceptionally well, meaning it can explain most of them clearly and efficiently.
Math, however, is more complicated.
While ChatGPT has proven to be quite proficient in solving equations and explaining its work, it isn’t flawless. Long calculations, unusual word problems, or questions requiring several layers of reasoning can lead to mistakes. A tiny error made halfway through a solution can quietly produce the wrong final answer.
The same pattern applies to coding. AI systems are completely capable of writing very effective code, explaining algorithms, and even fixing bugs. That’s one reason AI has become so popular among computer science students. But anyone who has copied AI -generated code into their editor knows it isn’t magic. Sometimes the program works perfectly. Other times it produces errors, ignores special cases, or creates an entirely different solution from the one you asked.
Then there are essays.
Give ChatGPT a prompt asking it to analyze symbolism in The Great Gatsby, and it will likely produce a well organized response with strong grammar and good logic.
But if you ask it to describe the moment when you finally understood a difficult concept after weeks of frustration, the thrill of winning your first robotics championship, or the feeling of presenting a research project you poured months into, it can’t.
Not because it isn’t advanced enough.
Because none of those things have ever happened to it.
ChatGPT has never been nervous before an exam. It has never celebrated a hard earned success or learned from failure. It can imitate human experiences with great detail, but it cannot genuinely draw from them because it has never experienced them.
And that is something an algorithm can’t replicate.
That is where I think people underestimate human intelligence.
Being smart isn’t just about having the right answer to a question, but also questioning the information, recognizing when something doesn’t feel right, making connections between unrelated ideas, and learning from personal experiences. Intelligence includes curiosity, creativity, judgment, empathy, and common sense, all of which are qualities that cannot be included in a dataset.
For decades, success in school often meant remembering as much information as possible. Today, nearly every fact imaginable is only a few clicks away, and AI can explain just about anything within seconds. That doesn’t make learning less important, but it changes what learning should look like.
The students who will succeed in the future most likely won’t be those who refuse to use AI nor the ones who depend on it for everything. They’ll be the ones who know how to ask better questions, verify information, think critically about the answers they receive, and understand when technology should be a tool instead of a shortcut.
Ironically, ChatGPT might score higher than many students on today’s high school exams. But instead of asking whether AI is becoming more intelligent, maybe we should ask a different question: What exactly are our exams measuring?
So maybe the exam was never really the point.
ChatGPT can answer almost any question you throw at it. But it never leaves the conversation more curious than when it started.
The real test isn’t whether you can beat a chatbot. It’s whether you keep asking questions long after it has finished answering them.
Works Cited
Giannos, Panagiotis, and Orestis Delardas. “Performance of ChatGPT on UK Standardized Admission Tests: Insights From the BMAT, TMUA, LNAT, and TSA Examinations.” National Library of Medicine, 26 April 2023, https://pmc.ncbi.nlm.nih.gov/articles/PMC10173042/.
Gilson, Aidan. “How Does ChatGPT Perform on the United States Medical Licensing Examination (USMLE)? The Implications of Large Language Models for Medical Education and Knowledge Assessment.” National Library of Medicine, 8 February 2023, https://pubmed.ncbi.nlm.nih.gov/36753318/.
Liu, Mingxin. Performance of ChatGPT Across Different Versions in Medical Licensing Examinations Worldwide: Systematic Review and Meta-Analysis, JMIR Publications, 25 July 2024, https://www.jmir.org/2024/1/e60807.
Revankar, Saisuman, and Aruna Madrekar. “ChatGPT Assignment Statistics By Usage, Adoption, And Student Behavior (2026).” ElectrioQ, 14 May 2026, https://electroiq.com/how-to-guide/chatgpt-assignment-statistics/.
Westfall, Chris. “Educators Battle Plagiarism As 89% Of Students Admit To Using OpenAI’s ChatGPT For Homework.” Forbes, 28 January 2023, https://www.forbes.com/sites/chriswestfall/2023/01/28/educators-battle-plagiarism-as-89-of-students-admit-to-using-open-ais-chatgpt-for-homework/.
“Why language models hallucinate.” OpenAI, https://openai.com/index/why-language-models-hallucinate.
Comments