Dr. Reyaz Ahmad
For most of human history, intelligence came with an unspoken condition: it belonged to living beings.
A mathematician solved an equation. A doctor diagnosed an illness. A poet found a metaphor that captured something ordinary language could not. A chess player anticipated a dozen moves ahead. A scientist noticed a pattern in nature that everyone else had missed.
We called these acts intelligent because there was a human mind behind them.
Then machines began doing some of the same things.
They defeated world champions, translated languages, wrote computer programs, generated essays, identified patterns in medical data, solved mathematical problems and increasingly became involved in scientific research.
And something interesting happened. We changed the question.
Instead of asking, “Is this intelligent behaviour?” we began asking, “Yes, but does the machine really understand what it is doing?”
That shift deserves attention, because the debate about artificial intelligence may no longer be only about machines. It may also be about us—about how willing we are to accept that intelligence might not be an exclusively human possession.
So, the real question is not simply:
Can intelligence be artificial?
It is more unsettling than that:
If a machine can learn, reason, solve unfamiliar problems and discover patterns at a level comparable to—or sometimes beyond—human performance, on what basis do we continue denying it the word “intelligence” simply because it does not think as we do?
The case for Artificial Intelligence: Judge intelligence by what it can do
Alan Turing recognised the difficulty of this question as early as 1950. Rather than getting trapped in an endless philosophical argument over what it means to “think,” he suggested looking at behaviour. His famous imitation-game approach shifted attention away from whatever might be happening inside a machine and toward what the machine could actually demonstrate.
There is something compelling about that approach.
After all, when we call another person intelligent, we do not examine the physical contents of that person’s brain. We infer intelligence from what the person does.
A student solves a problem she has never encountered before.
A scientist develops an original hypothesis.
A child learns from experience.
A strategist anticipates consequences that others overlook.
A mathematician finds an elegant proof.
From behaviour such as this, we infer an underlying cognitive ability.
Why should a machine be subjected to an entirely different standard?
Take the game of Go. For decades, Go was considered one of the most difficult challenges for artificial intelligence because of the staggering number of possible positions and the intuitive strategic judgement required at elite levels. Yet AlphaGo defeated world-class players and produced moves that surprised professional experts themselves. DeepMind presented the achievement as evidence that AI could learn sophisticated strategies rather than simply execute a catalogue of human-written instructions.
At that point, saying, “It is not intelligence because a machine did it,” begins to sound rather convenient.
There has been a recurring pattern in the history of AI. Once a machine master’s something that had previously been treated as evidence of intelligence, we quietly redefine the achievement.
Arithmetic? Just calculation.
Chess? Brute force.
Go? Pattern recognition.
Translation? Statistical prediction.
Writing? Word generation.
Programming? Code completion.
Scientific reasoning? Data processing.
The list keeps growing.
Eventually, we have to ask whether the phrase “mere computation” is becoming less an explanation and more a way of protecting a boundary that machines keep crossing.
Recent developments sharpen the issue. Stanford’s 2026 AI Index reports substantial gains by frontier AI systems on difficult reasoning benchmarks. Tests specifically designed to remain challenging for advanced models are being overcome much more quickly than many researchers expected.
The picture in science is more complicated, and therefore more interesting. Leading models can outperform average human experts on some chemistry question benchmarks, and AI systems are increasingly being used in areas such as biology, astronomy and weather prediction. At the same time, AI agents remain significantly weaker than PhD-level researchers when asked to carry out complex, end-to-end scientific investigations.
That mixed record tells us something important.
AI is neither magic nor fraud.
Its abilities are powerful, uneven and still limited.
But human intelligence is uneven too.
A gifted mathematician may be socially awkward. A brilliant musician may struggle with accounting. An outstanding surgeon may have little ability as a poet.
We do not conclude that such people lack intelligence. We recognise that intelligence appears in different forms and at different levels.
Why should artificial systems be excluded from that possibility?
The case against: Perhaps performance is not understanding
The argument against machine intelligence, however, is far more serious than simply saying, “Humans are special.”
Perhaps AI does not understand anything at all.
Perhaps it has simply become extraordinarily good at producing the appearance of understanding.
John Searle’s famous Chinese Room thought experiment attacks exactly this possibility.
Imagine a person sitting inside a room who knows no Chinese. Chinese symbols are passed into the room. The person has an enormous rulebook explaining which symbols should be returned in response to particular combinations.
By following the instructions perfectly, the person produces answers so convincing that Chinese speakers outside the room believe they are communicating with someone who understands Chinese.
But the person inside understands nothing.
He is manipulating symbols according to rules.
Searle’s point was simple but profound: syntax is not the same as semantics. Correctly manipulating symbols does not necessarily mean understanding what those symbols mean.
Modern AI makes that thought experiment remarkably relevant.
A language model can write a deeply moving paragraph about grief.
But has it ever grieved?
It can describe the fear of death in exquisite detail.
But does it know that it exists?
It can define love.
Has it ever missed someone?
It can discuss humiliation.
Can it feel ashamed?
It may tell a distressed person:
“I understand your pain.”
But what exactly is the “I” in that sentence?
This is where fluent language can mislead us.
An AI system may have learned the patterns, structures and associations of human language so well that its responses create the impression of an inner life. Yet linguistic competence and subjective experience are not obviously the same thing.
A machine can write:
“I am afraid.”
There may nevertheless be no one there who is afraid.
That distinction cannot simply be brushed aside.
Researchers themselves do not have an agreed method for deciding whether an artificial system is conscious. Contemporary work on consciousness acknowledges that the question remains unresolved, partly because human consciousness itself is still not fully understood.
That should make us cautious in both directions.
To declare today’s AI conscious simply because it speaks fluently would go far beyond the available evidence.
But to declare that a machine could never become conscious would also claim more certainty than science presently allows.
Hallucination and the Problem of Knowing What You Do Not Know
There is another weakness in current artificial intelligence that deserves serious attention.
Intelligence is not only about producing an answer. It also involves recognising the limits of one’s own knowledge.
AI systems often struggle with this.
They can generate false information with remarkable confidence—a problem generally described as hallucination. A 2026 Nature paper notes that plausible but incorrect outputs continue to limit the reliability of state-of-the-art language models despite sustained efforts to reduce them.
The difficulty is not merely that AI makes mistakes. Humans make mistakes constantly.
The deeper issue is epistemic responsibility: knowing when you know something, when you are uncertain and when you simply do not know.
A person faced with an unfamiliar question may say:
“I don’t know.”
An AI system may instead produce an elegant, detailed and entirely fictional answer.
That is a serious limitation.
One could therefore argue that present-day AI possesses considerable intellectual performance without the mature self-monitoring that we associate with reflective human reasoning.
Yet here too the comparison becomes uncomfortable.
Human beings are hardly models of perfect epistemic discipline.
We misremember events.
We confidently repeat false information.
We invent explanations after the fact.
We rationalise decisions.
We become overconfident.
We mistake something familiar for something true.
Human error does not lead us to conclude that humans have no intelligence. In the same way, AI hallucination may demonstrate the limitations of machine intelligence without proving its absence.
Perhaps the lesson is simpler: intelligence, whether human or artificial, does not guarantee truth.
Intelligence, Understanding and Consciousness Are Not the Same Thing
Much of the confusion in this debate comes from treating three different ideas as though they were interchangeable:
Intelligence, understanding and consciousness.
They are related, but they are not necessarily identical.
Intelligence can refer to the ability to learn, reason, predict, solve problems and adapt to changing circumstances.
Understanding involves grasping relationships, meaning and context.
Consciousness is something stranger: subjective experience, the fact that there is apparently something it is like to be a conscious being.
There is no obvious reason why all three must always appear together.
A calculator performs computation, but we have little reason to think it understands mathematics.
A dog clearly experiences the world, yet it cannot solve calculus.
A human infant is conscious long before displaying sophisticated reasoning or abstract problem-solving.
So why insist that intelligence and consciousness must always travel as a pair?
Perhaps one of AI’s greatest philosophical contributions will be to force us to separate abilities that biology happened to package together in the human brain.
A future artificial system could, in principle, possess extraordinary problem-solving intelligence without anything resembling human subjective experience.
If so, it would still make sense to call that intelligence.
It simply would not be human intelligence.
That may turn out to be one of the most important distinctions of the coming decades.
Is human exceptionalism part of the problem?
There is also a psychological side to the argument that we do not often acknowledge.
Human beings have a long history of defining themselves through abilities supposedly unavailable to other creatures or systems.
We considered ourselves uniquely rational.
Research revealed surprisingly sophisticated cognition in animals.
We regarded tool-making as distinctly human.
Other species use tools.
We believed complex communication separated us from the rest of life.
Research into animal behaviour complicated that assumption as well.
Strategic mastery appeared uniquely human.
Then machines defeated world champions.
Creativity seemed to offer safer ground.
Now generative systems produce music, images, prose and designs.
And so we retreat to a final line of defence:
“Yes, but it doesn’t really understand.”
Perhaps that defence is correct.
Perhaps consciousness, genuine meaning and deep understanding ultimately depend on biological life in ways machines can never reproduce.
That remains entirely possible.
But another possibility also deserves consideration.
Perhaps part of our resistance comes not from science, but from human exceptionalism—the instinctive desire to preserve some final intellectual territory as uniquely ours.
Recognising that possibility does not prove machine intelligence. It simply reminds us to examine our assumptions as critically as we examine the machines.
AI has another problem: Its intelligence was built from ours
There is, however, an equally important warning for those who are too eager to declare machines independent minds.
Today’s AI did not develop in intellectual isolation.
It has been trained largely on human-generated language, mathematics, science, literature, images, software and culture.
Its intellectual environment was created by us.
When an AI writes Shakespearean prose, solves a mathematical problem or discusses philosophy, we therefore have to ask a difficult question: how much of what we are seeing is genuine generalisation, and how much is extraordinarily sophisticated compression and recombination of humanity’s accumulated knowledge?
That is not a minor objection.
Imagine a parrot somehow exposed to civilisation’s entire library and trained to recombine everything it had encountered.
It might become astonishingly impressive.
But would that make it a philosopher?
The real test comes when a system faces something that cannot simply be recovered from existing patterns.
Can it develop genuinely new theories?
Can it recognise when accepted assumptions are wrong?
Can it design experiments?
Can it revise its most basic beliefs when evidence demands it?
Can it pursue a question whose answer does not already appear somewhere in the material on which it was trained?
Current evidence points in both directions. On well-defined tasks, AI can perform extraordinarily well. But on long, open-ended research problems requiring planning, experimentation, validation and replication, experienced human researchers still hold substantial advantages.
That is why neither hype nor dismissal is sufficient.
Perhaps the word “Artificial” is misleading us
The phrase artificial intelligence may itself encourage confusion.
An artificial heart pumps blood.
Artificial light illuminates a room.
An artificial limb can restore movement.
We do not usually say that artificial light provides “fake illumination.” We mean that the same broad function has been achieved through a different mechanism.
Why should intelligence necessarily be different?
If a non-biological system can learn, generalise, reason, predict, form useful abstractions and solve unfamiliar problems, then refusing to call any of those abilities intelligence merely because silicon rather than neurons produced them risks turning intelligence from a scientific concept into a membership privilege of the human species.
Yet the opposite mistake would be just as serious.
Fluent language is not evidence of consciousness.
Prediction is not automatically understanding.
Performance is not self-awareness.
Simulation is not necessarily experience.
And a machine saying, “I think,” does not establish that there is a conscious “I” behind the sentence.
These distinctions matter.
So, can intelligence be artificial?
My answer is uncomfortable precisely because it does not fit neatly into either camp.
Yes—intelligence can probably be artificial. But artificial intelligence does not therefore have to possess an artificial version of the human mind.
That distinction changes the debate.
Today’s AI already demonstrates several capacities historically associated with intelligence: pattern recognition, learning, reasoning, prediction, problem-solving and increasingly sophisticated generalisation.
What has not been established is that current AI possesses consciousness, subjective experience, an inner self or understanding in the same sense that humans experience these things.
Those are separate claims, and we should resist the temptation to collapse them into one.
The sceptic says: “If it does not think exactly as humans do, then it cannot be intelligent.”
The enthusiast says: “If it behaves intelligently, then it must experience the world as humans do.”
Both positions may be too simple.
What we may be witnessing is something far stranger than either side expected:
Intelligence without biology, competence without childhood, memory without lived experience, language without a tongue—and perhaps reasoning without consciousness.
For thousands of years, human beings have asked what makes us intelligent.
Artificial intelligence is forcing us to turn that question around.
Was intelligence ever uniquely ours in the first place?
Perhaps the deepest surprise will not come when a machine finally persuades us that it can think.
It may come when we realise that intelligence never needed to think exactly as humans do.
(The author is a freelancer and can be reached at [email protected])
