AI isn’t smarter than people: A simple test will show everything

Artificial intelligence is smarter than most people. This thought comes to mind of almost everyone who regularly uses modern language models. And you know what? This thought is based on our perception error.

Now I’ll show you why we all fell into the same comparison trap. And how one simple experiment completely changes this perception.

Imagine typical day of person who discovered AI. In morning you ask Claude about complex technical problem. Get detailed, structured answer. During day you ask ChatGPT to help with data analysis. It provides insights you wouldn’t have thought of. In evening Gemini writes letter for you so that you yourself are amazed at elegance of formulations.

And now remember your real day of communicating with people. Colleague can’t formulate thought in correspondence. Neighbor tells same story for hundredth time. Cashier in store can’t solve simplest arithmetic task without calculator. And there it is – this thought. AI is smarter than people around me. And not just smarter – noticeably smarter.

But let’s stop and think. What exactly are we comparing?

Here’s the trick. When you open AI application, you use models that are in top three, maximum top five in world. Claude, ChatGPT, Gemini – this is pinnacle of modern technology. These aren’t just good models. This is the best that humanity has created at moment in this field.

And now question. Do you live surrounded by top five smartest people on planet? No. Even top hundred smartest people of your country? Probably not. Top hundred smartest people of your city? Still no for most of us.

Understand where I’m going? We compare absolute AI elite with ordinary people from our environment. This is like if Lionel Messi went to play against yard team, won with score of thirty to zero, and we would conclude that Argentine footballers are better than all people in football at all. Absurd, right?

Want to check this theory? Conduct simple experiment. Go to any language model ranking. There are plenty on internet. And try to use models that aren’t in top ten, not in top twenty. But somewhere around hundredth place.

Know what you’ll discover? There will be quite decent models from known companies. They can answer questions, write texts, solve tasks.

But you’ll feel difference. They make more mistakes. Get confused with complex questions. Give less structured answers. Sometimes tell outright nonsense. And now most interesting. Compare your feelings from working with these models with communicating with real people. Suddenly picture becomes quite different. Average AI no longer seems so smart compared to average person.

And now imagine you took model from top five hundred. Or some completely weak model that isn’t even included in rankings. Believe me, such AI loses in intelligence to practically any person with basic education.

Let’s play honest game. If we take top three AI and compare them with top three smartest people of your city – who wins? Almost certainly people. Because among these people there may be outstanding scientists, inventors, doctors with thirty years of experience.

If we take random AI model from hundred existing ones and random person from street – result will be unpredictable. Sometimes model wins, sometimes person.

But if you take average model – say, five hundredth in ranking – and average person with average education, this will be honest comparison. And result may surprise you. In many tasks person will be smarter.

But why did we all fall into this trap? Answer is simple – we have choice in technologies, but no choice in people.

When you choose AI, you consciously go and download best application. Read reviews, watch comparisons, choose top model. This is rational decision – why use worse when best is available for free or for same money? But with people it doesn’t work that way. You don’t choose your colleagues by intelligence level. You don’t select neighbors by IQ test results.

Your environment forms randomly – work, residential area, hobbies, random acquaintances.

And this asymmetry creates distorted perception. We compare world’s best AI models with ordinary people. And conclude that models are smarter. But this is statistical survivorship bias in reverse – success bias.

There’s another important point. Top AI models are indeed outstanding – but in very narrow areas. They’re brilliant in text processing. Great at writing code. Excellent at analyzing and structuring information.

But ask AI to fry eggs. Or calm crying child. Or come up with way out of complex social situation where there’s no right answer, only least bad options.

Human, even average human, possesses universal intelligence. We can switch between abstract thinking and physical tasks. Between logic and emotions. Between learned rules and creative improvisation. AI is still narrowly specialized, even most advanced.

Let’s talk about what even most perfect language models don’t have. Life experience. Real empathy. Intuition based on millions of non-verbal signals.

Here’s example. Your friend writes you: “Everything’s fine”. Any person from message, sending time, situation context and previous communication experience can understand. Fine or catastrophically not fine. AI sees only three words.

Or another example. Doctor with twenty years of experience enters ward, throws one glance at patient. And says: “Urgently take tests, something’s wrong”. He can’t explain why. Thousands of patients before created pattern in his brain that works on feeling level. No AI can do that.

Or mother who by one sound of infant’s crying distinguishes – child is hungry, or stomach hurts, or just bored. Or chef who by eye determines meat doneness. Or mechanic who by engine sound knows what exactly broke.

When we say “smarter”, we often mean narrow area of verbal intelligence. But there are dozens more types of intelligence where people are beyond competition.

And yet something revolutionary happened. Before, access to expert knowledge was limited. To get consultation from high-level specialist, you had to pay money. Make appointment, wait. And now knowledge level comparable to good specialist lies in pocket of each of us and is instantly available.

This is culture shock. We’re not used to intelligence of such level being so accessible. Before, “smart helper” meant either expensive consultant or rare luck with environment. Now you just open application.

Comparison bar shot up. We started comparing ordinary people not with each other. But with machine trained on all human knowledge. This is like if possibility appeared to call professor of any university in world at any moment. And we would start complaining that our colleagues aren’t as erudite as professor.

People didn’t become dumber. We just now constantly hold in hands tool that in certain tasks works at expert level.

And this changes our perception of norm.

So, what did we find out? AI isn’t smarter than most people. We just compare AI elite with ordinary people. And this is unfair comparison.

If you communicated every day with best minds of humanity. Like Nobel Prize winners, top scientists, brilliant writers. You would think just same that these people are smarter than AI. Because that’s how it is.

Value of modern language models isn’t that they’re smarter than us. But that they democratized access to expertise. They gave every person opportunity to get help at level of good specialist at any moment.

This is tool. Very powerful, very convenient tool. But this isn’t new life form surpassing us. This is best we created to complement our own intelligence.

Use AI as amplifier of your thinking. But don’t use it as measure of human intelligence. Because you’re comparing incomparable. Selected machine with ordinary people.

And now simple test for you. Try to work for week not with top model, but with some from bottom of ranking. Or with last year’s version of same ChatGPT. And write in comments whether your perception changed after this experiment. I bet yes.

And remember – we live in amazing time. When best achievements of human thought are accessible to everyone. But this doesn’t devalue ordinary people. This elevates all of us, giving tools that previous generations could only dream of.