Dominic Cummings needs to learn a little more about science and blog a little less

The blog isn’t all word salad – but an awful lot of it is.

NS

Sign Up

Get the New Statesman's Morning Call email.

Dominic Cummings has posted a job advert on his blog. The text revels in his recent reading hobbies straddling computer science, forecasting, artificial intelligence, causality theory and more. 

As such, it fits neatly into his oeuvre. But it is an extraordinary document because, unlike his previous writing, this one is a job advertisement by someone in charge of the thinking behind the Prime Minister. 

Cummings wants to hire people who know and can implement some of the exciting, exotic, futuristic things he has been reading about (or would like us to think he’s been reading about). If you don’t know about any of these things, and find mention of them exotic, you might be tempted to think that this blog adds to your picture of him as that free-wheeling genius polymath; the breath of intellectual fresh air at the top of government who is going to try some new things.  What harm could it possibly do, you might ask yourself? We have tried those armies of nameless bureaucrats.

But in fact it is all very alarming, despite having the entertaining feel of a tract written for a character in Armando Iannucci’s The Thick of It

Contrary to what Cummings writes, there are not “trillion dollar bills lying on the street” at the “intersection” of anything and “the frontiers of the science of prediction”.  A lot of people have been spending decades trying to forecast a lot of things, with some success. But there are not imminent leaps forward to be had there, nor in applying what is known to government decisions or the outlook affecting them.

As Giles Wilkes, (who worked in Theresa May’s No 10 policy unit) tweeted damningly in response to the blog “you can’t machine learn your way out of the trade-offs inherent in the benefit system, SuperForecast vault over the huge economic cost of net zero, Prediction Market the Northern towns to prosperity”.

Cummings clearly wants to signal that he knows these fields and thinks them important to the task at hand. But in trying, he reveals that he has not got far enough to realise they are not going to help him do his job. 

Cummings’s job advert asks for familiarity with “theoretical computer science”. To what end?  Who in No 10 is going to read a memo explaining the takeaways from the recent theoretical computer science debates? Is he hoping to intervene in the UK computer hardware sector with something new?  Whoever applies is going to be writing programs, in Python [Cummings has no doubt read that people write programs in Python and seems to think this language would be better than others for his needs]. But programs that do what? Someone is going to write a program and depending on what comes out, we will get a policy change? Does Dominic Cummings know that you can do a lot of things on Excel too? Is that not part of the requirements because it sounds a bit too easy?

A charitable read is that the blog signals an injection of energy and hope and an invitation to join in.  

But one wonders who will take him up on it. If you are familiar with theoretical computer science, data science, AI, machine learning, computational rationality, and all the other signifiers, and you have options, why would you work for Dominic Cummings?

You might find it intriguing that he seems to have circumvented normal Human Resource processes in his blog post.  But you’d guess that he doesn’t know what he is talking about, and while it is nice to have a boss who is interested in what you do, you would assess that he had probably find out quickly that what you know isn’t going to deliver anything for him. The two or three hours it would take to explain how cat-picture-recognisers are coded [happy to oblige on this count] would probably do the job.

Going through that experience is not going to give you much to write on your CV to take to your next challenge at the “intersection” of whatever cutting edge people seek to “intersect”.

Potential applicants might ask themselves questions about how the policy formulation process is going to work.  As the pilot of VoteLeave, the organisation that won the referendum for Brexit, Cummings designed the rejection of all the accumulated wisdom we have in empirical trade economics about the effects of erecting trade barriers following Brexit. What would he make of the first memo that read “I have looked at the evidence on the effects of welfare benefits on labour supply” or any other policy area you care to name and did not conclude that everything needed to be uprooted?

One refrain in the post-Trump world is that the outputs of populist revolutionaries are to be taken “seriously, but not literally”. Reading about “trillion dollar” gains this advice is very tempting, but with the UK economy itself being about £2.2trn, it is very unlikely that there are a few changes we can make to double or treble our national income. The real significance may just be that he is signalling he has a license to pay no heed to anything or anyone that went before. The text is not written expecting to generate a dialogue about things said, just as a signal that the Prime Minister is OKwith him writing what he wants.

Cummings’s post is not all word salad.  There are things said about the paucity of expertise in the civil service.  Overlooking the irony of this message being contained in a sixth-form-level homage to sciencey things, there might well be areas where more technical expertise would produce better policy. This is a debate still worth having, despite its longevity and the amount of ink already spilled. 

Tony Yates is former professor of economics at the University of Birmingham 

Free trial CSS