Criticism Is Easy

Much has been written about the Dunning–Kruger effect, in every language, until its name became a ready-made argument in any quarrel on the internet. I had no intention of writing about it. What would I add? But over the past two days a great deal of criticism has been aimed at LEAP, and at HUMAIN in particular, by people who work neither in technology nor in investing in it. That is what made me write this piece. This is not a defence of the conference or of the company. Both can be criticised, and both should be. The piece is about a single question: what does the critic actually know about the thing he is criticising?

The Editorial of 1903

On 9 October 1903, an editor at The New York Times wrote an editorial about flight. Two days earlier, the flying machine built by the scientist Samuel Langley had fallen into the Potomac in front of the reporters. The American War Department had spent fifty thousand dollars on it. The editor wrote that flight might be achieved one day, but only after somewhere between one million and ten million years of work by mathematicians and mechanics. The editorial was headed “Flying Machines Which Do Not Fly.”

Sixty-nine days later, on 17 December, the Wright brothers flew at Kitty Hawk.

The editor was not ignorant of the news. He had seen a failed experiment on which the government had spent a great deal of money, and he concluded that the whole field was hopeless. What he did not know was that two brothers who repaired bicycles in the town of Dayton had built a small wind tunnel, measured lift again, and corrected the tables that Lilienthal and others had relied on. He judged an entire field from a single piece of news.

What is the Dunning–Kruger effect?

In 1999, David Dunning and Justin Kruger of Cornell University published a study in which they tested students on logic, grammar and the appreciation of humour, and then asked each student to estimate where he ranked among his peers. The students whose scores fell in the bottom quarter placed themselves above average. The students who did best placed themselves lower than they deserved.

The explanation the study offers is a simple one. The skill you need in order to do a piece of work is the same skill you need in order to judge how well it has been done. Someone who does not know the rules of grammar cannot see his own grammatical mistake. The old logicians had a name for this condition: compound ignorance, al-jahl al-murakkab, to be ignorant of a thing and ignorant that you are ignorant of it. It is not the same as simple ignorance. A person in simple ignorance knows that he does not know, so he either asks or says nothing.

One important qualification: the effect is smaller than it is usually made out to be. Several researchers have gone back to the study’s data, among them Edward Nuhfer in 2017 and Gilles Gignac in 2020. They found that part of the famous curve is accounted for by two statistical phenomena that have nothing to do with conceit: regression to the mean, and the general tendency of people to rate themselves above average. The effect is real, but it is weaker than the popular picture of it. That is why it should not be used as an accusation in an argument. Anyone who charges an opponent with the Dunning–Kruger effect is assuming that he himself is the expert, and that is an assumption that needs evidence.

Confidence Does Not Transfer Between Fields

The problem with the criticism of LEAP and HUMAIN is not ignorance itself. It is the transfer of confidence from one field into another. The critic may well be a successful man in his own field: a lawyer, a writer, a doctor, a businessman. In that field he learned to trust his own judgement, and then he used the same confidence in a field he never studied. Worse still is when he is none of the above. Just a novice learner in any field. Anyway, this is Twitter!

I know this from medicine. If a software engineer wrote on Twitter that my reading of a bone marrow sample was wrong, no one would pay him any attention. But I watch a doctor write confidently about the viability of data centres and the price of GPUs. I watch a writer pass verdict on an artificial intelligence investment worth tens of billions, without having read anything about supply chains or the economics of compute. He has no information. He has an impression: that the noise is loud and the promises are louder.

An impression is not criticism. Criticism starts from inside the subject: from a figure, a contract, a particular technology. An impression starts from outside it: from a general feeling about the noise. That is the difference between an editor at The New York Times and an aeronautical engineer who knows that Langley’s crash says nothing about the wind tunnel in Dayton.

What Is a Critic Entitled to Ask?

To be fair, let me put the strongest case the critics have. A conference held at state expense that announces deals worth billions of dollars is a fair thing to ask about. How many of those deals turned into signed contracts? How many of them became working projects? A national artificial intelligence company that buys tens of thousands of chips is also a fair thing to ask about. Is there domestic demand to absorb this computing capacity, or will it be leased abroad? And is human expertise being built at the same speed as the buildings? These are legitimate questions, and necessary ones. Specialists should be putting them under their own names.

But every one of these questions takes knowledge before you can even frame it, and more knowledge still before you can answer it. The one who has that knowledge asks. The one who does not have it mocks, because mockery needs no knowledge at all.

Dunning himself proposed a single remedy: feedback. You learn the skill, and your capacity to judge it rises along with it. Anyone who wants to criticise an artificial intelligence company should read at least one financial report, and should understand what “sovereign compute” means before passing verdict on it. Anyone who will not do that should ask instead of judging.

To weigh this on the scale of the Qurʾān: I do not claim that the Qurʾān was revealed to explain an experiment at Cornell University. But the Qurʾān laid down a rule about speech that comes before psychology: “Do not pursue what you have no knowledge of. Hearing, sight and heart — each of these will be asked about” (Q 17:36). The verse does not forbid ignorance, for ignorance is no sin. It forbids following a thing and ruling on it while you do not know it.

In Sūrat Yūnus there is a description of the man who rejects before he understands: “Rather, they denied what they had not encompassed in knowledge” (Q 10:39). The denial came before the knowledge was encompassed, and that is the order of compound ignorance exactly. Another verse gives the way out: “Ask the people of knowledge, if you do not know” (Q 16:43).

A saying is attributed to al-Shaʿbi, one of the Successors: “‘I do not know’ is half of knowledge.” I have not verified the chain of transmission for it, but its meaning is plain enough. Knowing the limits of your own knowledge is part of knowledge.

The Correction of 1969

In 1920, The New York Times published another editorial. It mocked the scientist Robert Goddard for saying that a rocket could work in a vacuum, and said he lacked the knowledge that is taught in high schools. On 17 July 1969, while Apollo 11 was on its way to the Moon, the paper published a short correction admitting that it had been wrong.

The first correction took sixty-nine days. The second took forty-nine years. In both cases the verdict came from an office that had seen neither the wind tunnel nor the rocket. Today we face the same choice: to pass judgement on what we do not know and wait for the correction, or to ask before we judge, or to keep silent, for silence is wisdom.

✦ ✦ ✦

This is a translation of an essay written in Arabic. Quotations from the Qurʾān are rendered for sense; the Arabic page carries them in the original.

© 2026 Husain Alkhaldy — published in the site's essays.