Dominic Cummings: “we’re hiring data scientists, project managers, policy experts, assorted weirdos”

Link post

Do­minic Cum­mings (dis­cussed pre­vi­ously on LW, most re­cently here) is a Se­nior Ad­vi­sor to the new UK PM, Boris John­son. He also seems to be es­sen­tially a ra­tio­nal­ist (at least in terms of what ideas he’s pay­ing at­ten­tion to).

He has posted to­day that his team is hiring “data sci­en­tists, pro­ject man­agers, policy ex­perts, as­sorted weirdos”. Per­haps some LW read­ers should ap­ply.

Ex­ten­sive quotes be­low:

‘This is pos­si­bly the sin­gle largest de­sign flaw con­tribut­ing to the bad Nash equil­ibrium in which … many gov­ern­ments are stuck. Every in­di­vi­d­ual high-func­tion­ing com­pe­tent per­son knows they can’t make much differ­ence by be­ing one more face in that crowd.’ Eliezer Yud­kowsky, AI ex­pert, LessWrong etc.


Now there is a con­fluence of: a) Brexit re­quires many large changes in policy and in the struc­ture of de­ci­sion-mak­ing, b) some peo­ple in gov­ern­ment are pre­pared to take risks to change things a lot, and c) a new gov­ern­ment with a sig­nifi­cant ma­jor­ity and lit­tle need to worry about short-term un­pop­u­lar­ity while try­ing to make rapid progress with long-term prob­lems.

There is a huge amount of low hang­ing fruit — trillion dol­lar bills ly­ing on the street — in the in­ter­sec­tion of:

  • the se­lec­tion, ed­u­ca­tion and train­ing of peo­ple for high performance

  • the fron­tiers of the sci­ence of prediction

  • data sci­ence, AI and cog­ni­tive tech­nolo­gies (e.g See­ing Rooms, ‘au­thor­ing tools de­signed for ar­gu­ing from ev­i­dence’, Tet­lock/​IARPA pre­dic­tion tour­na­ments that could eas­ily be ex­tended to con­sider ‘clusters’ of is­sues around themes like Brexit to im­prove policy and pro­ject man­age­ment)

  • com­mu­ni­ca­tion (e.g Cial­dini)

  • de­ci­sion-mak­ing in­sti­tu­tions at the apex of gov­ern­ment.

We want to hire an un­usual set of peo­ple with differ­ent skills and back­grounds to work in Down­ing Street with the best offi­cials, some as spads and per­haps some as offi­cials. If you are already an offi­cial and you read this blog and think you fit one of these cat­e­gories, get in touch.

The cat­e­gories are roughly:

  • Data sci­en­tists and soft­ware developers

  • Economists

  • Policy experts

  • Pro­ject managers

  • Com­mu­ni­ca­tion experts

  • Ju­nior re­searchers one of whom will also be my per­sonal assistant

  • Weirdos and mis­fits with odd skills


A. Unusual math­e­mat­i­ci­ans, physi­cists, com­puter sci­en­tists, data scientists

You must have ex­cep­tional aca­demic qual­ifi­ca­tions from one of the world’s best uni­ver­si­ties or have done some­thing that demon­strates equiv­a­lent (or greater) tal­ents and skills. You do not need a PhD — as Alan Kay said, we are also in­ter­ested in grad­u­ate stu­dents as ‘world-class re­searchers who don’t have PhDs yet’.


A few ex­am­ples of pa­pers that you will be con­sid­er­ing:

  • [...]

  • The pa­pers on com­pu­ta­tional ra­tio­nal­ity be­low.

  • The work of Judea Pearl, the lead­ing scholar of cau­sa­tion who has trans­formed the field.


B. Unusual soft­ware developers

We are look­ing for great soft­ware de­vel­op­ers who would love to work on these ideas, build tools and work with some great peo­ple. You should also look at some of Vic­tor’s tech­ni­cal talks on pro­gram­ming lan­guages and the his­tory of com­put­ing.

You will be work­ing with data sci­en­tists, de­sign­ers and oth­ers.

C. Unusual economists

We are look­ing to hire some re­cent grad­u­ates in eco­nomics. You should a) have an out­stand­ing record at a great uni­ver­sity, b) un­der­stand con­ven­tional eco­nomic the­o­ries, c) be in­ter­ested in ar­gu­ments on the edge of the field — for ex­am­ple, work by physi­cists on ‘agent-based mod­els’ or by the hedge fund Bridge­wa­ter on the failures/​limi­ta­tions of con­ven­tional macro the­o­ries/​pre­dic­tion, and d) have very strong maths and be in­ter­ested in work­ing with math­e­mat­i­ci­ans, physi­cists, and com­puter sci­en­tists.


The sort of con­ver­sa­tion you might have is dis­cussing these two pa­pers in Science (2015): Com­pu­ta­tional ra­tio­nal­ity: A con­verg­ing paradigm for in­tel­li­gence in brains, minds, and ma­chines, Ger­sh­man et al and Eco­nomic rea­son­ing and ar­tifi­cial in­tel­li­gence, Parkes & Wel­l­man.

You will see in these pa­pers an in­ter­sec­tion of:

  • von Neu­mann’s foun­da­tion of game the­ory and ‘ex­pected util­ity’,

  • main­stream eco­nomic the­o­ries,

  • mod­ern the­o­ries about auc­tions,

  • the­o­ret­i­cal com­puter sci­ence (in­clud­ing prob­lems like the com­plex­ity of prob­a­bil­is­tic in­fer­ence in Bayesian net­works, which is in the NP–hard com­plex­ity class),

  • ideas on ‘com­pu­ta­tional ra­tio­nal­ity’ and meta-rea­son­ing from AI, cog­ni­tive sci­ence and so on.

If these sort of things are in­ter­est­ing, then you will find this pro­ject in­ter­est­ing.

It’s a bonus if you can code but it isn’t nec­es­sary.

D. Great pro­ject man­agers.

If you think you are one of the a small group of peo­ple in the world who are truly GREAT at pro­ject man­age­ment, then we want to talk to you.


It is ex­tremely in­ter­est­ing that the les­sons of Man­hat­tan (1940s), ICBMs (1950s) and Apollo (1960s) re­main ab­solutely cut­ting edge be­cause it is so hard to ap­ply them and al­most no­body has man­aged to do it. The Pen­tagon sys­tem­at­i­cally de-pro­grammed it­self from more effec­tive ap­proaches to less effec­tive ap­proaches from the mid-1960s, in the name of ‘effi­ciency’. Is this just an­other way of say­ing that peo­ple like Gen­eral Groves and Ge­orge Muel­ler are rarer than Fields Medal­lists?


E. Ju­nior researchers

In many as­pects of gov­ern­ment, as in the tech world and in­vest­ing, brains and tem­per­a­ment smash ex­pe­rience and se­nior­ity out of the park.

We want to hire some VERY clever young peo­ple ei­ther straight out of uni­ver­sity or re­cently out with with ex­treme cu­ri­os­ity and ca­pac­ity for hard work.


F. Communications

In SW1 com­mu­ni­ca­tion is gen­er­ally treated as al­most syn­ony­mous with ‘talk­ing to the lobby’. This is partly why so much pun­ditry is ‘nar­ra­tive from noise’.

With no elec­tion for years and huge changes in the digi­tal world, there is a chance and a need to do things very differ­ently.


G. Policy experts

One of the prob­lems with the civil ser­vice is the way in which peo­ple are shuffled such that they ei­ther do not ac­quire ex­per­tise or they are moved out of ar­eas they re­ally know to do some­thing else. One Fri­day, X is in charge of spe­cial needs ed­u­ca­tion, the next week X is in charge of bud­gets.


If you want to work in the policy unit or a de­part­ment and you re­ally know your sub­ject so that you could con­fi­dently ar­gue about it with world-class ex­perts, get in touch.


G. Su­per-tal­ented weirdos

Peo­ple in SW1 talk a lot about ‘di­ver­sity’ but they rarely mean ‘true cog­ni­tive di­ver­sity’. They are usu­ally bab­bling about ‘gen­der iden­tity di­ver­sity blah blah’. What SW1 needs is not more drivel about ‘iden­tity’ and ‘di­ver­sity’ from Oxbridge hu­man­i­ties grad­u­ates but more gen­uine cog­ni­tive di­ver­sity.

We need some true wild cards, artists, peo­ple who never went to uni­ver­sity and fought their way out of an ap­pal­ling hell hole, weirdos from William Gib­son nov­els like that girl hired by Bi­gend as a brand ‘di­v­iner’ who feels sick at the sight of Tommy Hilfiger or that Chi­nese-Cuban free run­ner from a crime fam­ily hired by the KGB. If you want to figure out what char­ac­ters around Putin might do, or how in­ter­na­tional crim­i­nal gangs might ex­ploit holes in our bor­der se­cu­rity, you don’t want more Oxbridge English grad­u­ates who chat about La­can at din­ner par­ties with TV pro­duc­ers and spread fake news about fake news.

By defi­ni­tion I don’t re­ally know what I’m look­ing for but I want peo­ple around No10 to be on the look­out for such peo­ple.

We need to figure out how to use such peo­ple bet­ter with­out ask­ing them to con­form to the hor­rors of ‘Hu­man Re­sources’ (which also ob­vi­ously need a bon­fire).



As Paul Gra­ham and Peter Thiel say, most ideas that seem bad are bad but great ideas also seem at first like bad ideas — oth­er­wise some­one would have already done them. In­cen­tives and cul­ture push peo­ple in nor­mal gov­ern­ment sys­tems away from en­courag­ing ‘ideas that seem bad’. Part of the point of a small, odd No10 team is to find and ex­ploit, with­out wor­ry­ing about me­dia noise, what Andy Grove called ‘very high lev­er­age ideas’ and these will al­most in­evitably seem bad to most.

I will post some ran­dom things over the next few weeks and see what bounces back — it is all up­side, there’s no down­side if you don’t mind a bit of noise and it’s a fast cheap way to find good ideas…

H/​T ioannes_shade