What Makes You So Special?

There is a seductive belief that I think most people hold, usually unconsciously, that being human makes us special. The reach of this belief is broad. It gets applied to bodies and minds, our place in the animal kingdom and universe, and even to the specific moment in time that we happen to be living through.

It’s used also as a measure of oneself against everyone else. And I mean that literally. Everyone else in the entire world, existing in the past, present and future. Main character syndrome is what we call the sense a person has that the story is about them. Expanding it more broadly, the larger version is the feeling that, at very least, the story is about us. People. Humans.

I don’t subscribe to this type of thinking. The idea that any of us, individually or as a species, are inherently exceptional relative to everything else seems absurd to me. This isn’t an essay about nihilism or solipsism. I don’t think nothing matters, or that people are worthless. I just can’t find a rational basis for believing there is some fated reason why all of us, or any one of us, should be at the centre of everything. If I were Sam Altman or Elon Musk, I might read the evidence differently.

We are not the strongest being on Earth, or the quickest, or the most numerous, or the most individually resilient. But we are the cleverest, and that is why the planet came to be arranged around us. If you had to point to one thing that people hold onto as the reason we are special, intelligence would be it.

Machines have been displacing humans for a long time. Industrialisation brought with it machines that were stronger and faster than human labourers, taking jobs from people working in farming and manufacturing by doing the work better. Pride and livelihood were affected for a great many people, but the threat never extended to the idea that humans were the cleverest things around. Machines that write, reason and argue, increasingly better than we do, touch a different nerve.

In 1951, Alan Turing gave a BBC talk on whether computers could think. His script notes that it is “customary, in a talk or article on this subject, to offer a grain of comfort, in the form of a statement that some particularly human characteristic could never be imitated by a machine.” He gave the examples of writing good English, being influenced by sex appeal, and smoking a pipe. He then declined to agree: “I cannot offer any such comfort, for I believe that no such bounds can be set.”

Now, 75 years later, his refusal looks well placed, on the first of those examples at least. People still search for the grain of comfort, and still offer it, but it will not hold still.

Different versions of the comfort are expressed all the time. A model can write more code than a person, and better, but it cannot take a step back and think, “That isn’t as clean as I’d have written it.” Even where that’s true, the claim being made is tiny. The capability is entirely conceded, and the best psychological defence a person can mount is to look for a mildly dissatisfying feature, unworthy of outward expression because it carries no functional value.

Each time the capability expands, the claim retreats to something different or smaller. The retreat is rarely admitted, and the new position is usually held as confidently as the last one.

Sometimes the comfort doesn’t retreat at all. I recently heard it said, as fact, that ChatGPT will confidently predict the winner of an election because it is required to give a helpful answer. I asked it straight away. It replied clearly that there was no way for it to know who would win the next US presidential election. Something like it happened once. In June 2024, a month before the UK general election, Sky News asked ChatGPT who had won, and it said Labour. OpenAI called it a bug, and by the time Sky published, the answer had changed. Old claims keep circulating for ordinary reasons. I suspect this one has also lasted because it is a grain of comfort that people want to take.

People raise complaints about em dashes and sycophancy in the same way and for the same reason. Sycophancy was a real problem, and not all models manage it the same way. But in my experience, what is currently available at the frontier is a long way from the models those complaints were based on. People will differ in how much they flatter you, too, and that also isn’t a comment on our species.

Em dashes are a style matter worthy of a mournful comment from me. Until recently, they were seen in almost every email I wrote. Now, they’re nowhere to be found. How I want to write hasn’t changed. No, the reason I stopped using them is because of how the em dash gets read now. I don’t want a recipient of my email to think it’s the product of unscrutinised AI. If my reader believes that human input is better, then it matters to me that my writing comes across as human. Unfortunately, that’s at odds with my natural writing style, and so I’ve changed it. I miss using em dashes, and the irony is that in trying to seem more human, I’m normalising my writing style to something that is less identifiably me — a real human.

Claims like these travel well, and the reason why is simple to me. If you’re willing to seriously consider the capabilities of AI, it is completely rational to be anxious. Turing saw this in 1951. He thought the danger of being superseded was remote, but “not astronomically remote”, and “certainly something which can give us anxiety”. He also thought that many people opposed the idea of a thinking machine not for any rational reason, “but simply because they do not like the idea.” Commentators who insist that the capabilities aren’t real, or aren’t as great as they appear, are offering words of comfort. It’s no mystery why people gravitate towards them, internalise them, and repeat them. It’s nicer than the alternative on both an individual and an existential level.

But there are two anxieties here that I want to treat differently from one another.

The first relates to the consequences: what AI will do to the way we work, our finances, our privacy, our security, our political systems, and to where power resides. I share that anxiety. It is rational. Denying the capabilities of AI does not make the consequences disappear. The only thing that it will do is leave you less prepared.

The second anxiety relates to our standing: whether humans are special, and whether we are still at the top of anything if intelligence cannot be the foundation of a claim along those lines. A belief in human exceptionalism soothes this anxiety, but that doesn’t make it true.

Are we diminished by what these systems can do? In many respects, yes. Intelligence enabled us to function as the dominant species of our planet, and it’s why we have most of the things that we imbue with material and intangible value: jobs, money, influence over our world and our future. The emergence of a tool that performs better at the things that could previously only be done using human intelligence casts doubt over our worth as contributors to the world we’ve built. No belief about ourselves can change that.

Human exceptionalism adds another complication on top: that our intelligent contributions are why we matter at all. After you make the claim that humans are exceptional because of special characteristics, losing in a contest measuring those characteristics becomes an existential challenge. The arguments in favour of the view that humans possess those characteristics can attempt to defend against the second anxiety, but they will do little to defend against the consequences of AI. Very few will keep their job because they can privately note a different way of achieving the same outcome.

You can accept all of that and still argue that the belief in human exceptionalism has an important job to do. If people aren’t special, then why would we value human life or work at all? I can’t alleviate that worry. Much of how we are valued at work, in what we are paid or whether we are heard, depends on our performance. I believe that AI will continue to shift how human performance is valued when there is an alternative. But that is an argument about what we stand to lose, not evidence that we are special. A belief can be useful and still not be true, and even if this one is useful in some places for the moment, the fact that it is giving way makes it an unreliable thing to build on.

This doesn’t tell any of us how we should actually measure our worth, or what we should feel about the transition we’re living through.

Turing had no grain of comfort to offer in 1951, and I can’t find one now. I’m anxious, too, about work and who will hold the power. Something will have to carry us through this period. I don’t know what that will look like yet, but it can’t be the belief that humans are the apex intelligence of the universe, destined to dominate in perpetuity. That belief has always rested more on magical thinking than on anything truly objective and measurable.

Anyway, sorry for the ad break. It’s time for your next scene.