Human minds’ distinctiveness under scrutiny as AI advances

Rowan Elmsford3 May 2026

A human hand and a robotic hand reaching towards each other against a digital background

Artificial intelligence can now compose polished prose, win medals in mathematics and defeat the world’s best players at the most complex games — achievements that challenge the long-held belief that human intelligence is uniquely special. Tech executives promise that superhuman AI is just around the corner. But according to Tom Griffiths, professor of information technology at Princeton University, the question is not whether machines will overtake us, but whether we are measuring intelligence on the wrong scale.

AI’s recent achievements

In the past few years AI systems have matched or surpassed human performance in domains once considered exclusively human. Google DeepMind’s AlphaGo defeated the best human Go players after training on many lifetimes of games. Large language models such as ChatGPT produce fluent prose and pass exams. More recently, AI has demonstrated the ability to solve mathematical problems well enough to win competition medals. These successes have prompted debate about whether human minds are becoming obsolete — or merely also-rans.

Griffiths, who is also author of The Laws of Thought (William Collins), argues that the framing of AI “catching up” to humans is based on a flawed analogy. “My parents used to mark the heights of my younger brother and me on the doorframe of our laundry,” he writes. “Each year he would get a little closer to me, until one year the unthinkable happened and he outgrew me. The current moment feels a bit like that, as we look at these new younger siblings with concern that they might overtake us.”

But intelligence, he insists, is not like height. “There is only one way to be tall, but there are lots of ways to be smart.” Other animals demonstrate this diversity: birds navigate vast distances, ants cooperate in complex societies, and spiders hunt with precision. Each species has been shaped by its environment to be smart in a different way.

Human limitations as strengths

Human intelligence, Griffiths argues, is equally shaped by biology — and it is human limitations, not abilities, that make minds special. “We only live for a few decades and have to learn everything we are going to learn and do everything we are going to do in that short time,” he says. “All that learning and doing will be carried out at the direction of a kilogram or so of neurons trapped inside our bony skulls. We can only share our thoughts with others by making noises with our mouths or tapping and wiggling our fingers.”

AI systems face none of these constraints. They can process more data than a human might see in a lifetime, expand their capacity by adding more computers, and share information instantaneously. Yet Griffiths contends that it is precisely our short lives, squishy brains and mouth noises that make us special. “Human intelligence is a response to our limitations,” he writes.

To make the most of limited experience, humans have developed an extraordinary ability to learn from small amounts of data. No AI system, he notes, can produce sentences with the creativity of a human five-year-old when exposed to the same vocabulary and interactions. Our limited brains force us to become skilled at recognising patterns and using attention wisely. And because we rely on slow, noisy communication — mouth noises — we have invented tools such as language, writing, teaching and science to pool knowledge across people and generations. That requires us to think about what is going on inside other people’s heads and to cooperate toward shared goals.

Different paths

Because humans and machines face fundamentally different constraints, Griffiths expects them to find different solutions to the same problems. He offers two concrete examples. The first is a simple counting task: how many letters are in a sequence of as? For a human, it is straightforward to count them. For an AI system, however, it is trickier. AI models break words into parts called “tokens” and tend to favour sequences that appear more often in their training data. OpenAI’s GPT-4 — hailed by some as showing “sparks of artificial general intelligence” — was more likely to correctly answer when given 30 letters rather than 29, simply because the number 30 is written down more often.

The second example involves a pharmacist who needs a drug concentration of 785 parts per million (ppm). Two test tubes are available: one at 685 ppm and another at 791 ppm. A human would pick 791 ppm as the closer match. Yet some leading AI systems choose 685 ppm. Why? Because artificial neural networks blur things together. If a number is represented as a string of digits, 785 is more similar to 685 (one digit different) than to 791 (three digits different). But as a quantity, 785 is closer to 791. Mixing up the two representations can have significant consequences, Griffiths warns.

Human intelligence, by contrast, draws on a breadth of experience far beyond the data used to train AI. “We use our brains to put nappies on babies, play chess, prove theorems, cook dinner, write novels and compose symphonies,” Griffiths writes. AI systems are typically trained to do one thing — ChatGPT can offer tips about nappies but cannot gently hold a squirming infant. Human brains have evolved in a world that presents all these challenges, leaving us just well enough equipped to learn what we need in a single human lifetime.

Griffiths concludes that finite lives, finite brains and limited communication have shaped the nature of human intelligence in ways that cannot be replicated by machines. “We can thus expect that human minds will continue to be a little bit special, even as we continue to develop smarter machines.” Rather than a single scale with AI catching up to a human benchmark, intelligence comes in many forms. AI will be better than humans in some ways and worse in others. “And just like siblings, perhaps we can learn to treat one another not as rivals, but as companions.”