The Economist - USA (2020-02-22)

rithmstopredicta building’senergycon-

sumptionortodetect“deepfake”videos,

with prizes sometimes exceeding $1m.

ThatisalsoFacebook’sandGoogle’swayto

makemoney.Theyhardlyeverselldata,

buttheydosellinsightsaboutwhoisthe

besttargetforadvertising.

Yetdatahavefailedtobecome“anew

assetclass”,astheWorldEconomicForum,

a conference-organiser and think-tank,

predictedin2011.Mostdataneverchange

hands,andattemptstomakethemmore

tradablehavenottakenoff.Tochangethis,

especially in Europe,manufacturers are

pushingtosecurepropertyrightsforthe

datageneratedbytheirproducts.Others

want consumers to own the data they

create,sotheycansellthemandgeta big-

gercutfromtheirinformation.

Again,economicsgetsintheway.Al-

thoughdataareoftenthoughtofasa com-

modity,corporatedatasets,inparticular,

tendnottobefungible.Eachisdifferentin

thewayit wascollected,andinitspurpose

andreliability.Thismakesitdifficultfor

buyersandsellerstoagreeona price:the

valueofeachsortishardtocompareand

changesovertime.A furtherbarriertotradingisthatthevalueofa

datasetdependsonwhocontrolsit.Whatmightsimplybedataex-

hausttoonefirmcouldbedigitalgoldtoanother.“Thereisnotrue

valueofdata,”saysDianeCoyleoftheUniversityofCambridge.

Asforpersonaldata,definingpropertyrightsistricky,because

muchinformationcannotbeattributedtooneperson.Who,for

instance,ownsthefactthata datingsitehasmatcheda couple?

Thecouplethemselves?Ortheservice?Complicatingmatters,data

haveplentyofexternalities,bothpositiveandnegative,meaning

thatmarketsoftenfail.Whyshoulda socialnetwork,say,buythe

dataofanindividualifitcanmakequiteaccuratepredictions

abouthimbycrunchingdatafromotherusers?

Althoughdataareunlikelyevertobetradedaswidelyasoil,

techfirmskeeptryingtomakethiseasier.AmazonWebServices

(aws), thecloud-computingarmofthee-commercegiantrecently

launcheda marketplacethataimstomaketradingindataaseasy

aspossible.It worksa bitlikeanonlinestoreforsmartphoneapps:

buyerssubscribetofeeds,agreetolicensingconditions,andaws

processesthepayment.

Lightstuffnotblackstuff

Astheoilmetaphorisseenasincreasinglyproblematic,thecom-

parisontosunlightorsimilarresources,suchasairandwater,has

riseninfavour.Manypeoplewhopreferthismetaphoraskifdata

donotreallylendthemselvestobeturnedintoa tradablegood,

thenwhyeventry?Woulditnotinsteadbebettertoensurethat

dataareusedasmuchaspossible?Afterall,thiswillmaximiseso-

cialwealth.Inotherwords,nobodyputsupcurtainsandtriesto

chargeforsunlight.

Thislineofargumenthasalreadygivenbirthtowhatisknown

asthe“open-data”movement.Itschampionspushorganisations

anduniversitiestogiveawaytheirdatasotheycanbewidelyused,

forinstancebystartups.Today,mostgovernments,nationalor

otherwise,boastanopen-dataproject,althoughthequalityofthe

datamadeavailablevariesgreatly.

Morerecently,companieshavestartedtopublishtheirdata,

too.Severalfirmsthatworkonself-drivingcarshavesharedsome

oftheinformationcollectedbytheirvehicles.“Forresearchersto

asktherightquestions,theyneedtherightdata,”accordingtoDra-

gomirAnguelov,principalscientistatWaymo,a firmownedbyAl-

phabet,Google’sparent,thatisoneofthecompaniesthathasdone

this.Othersareworkingontechnologytomakesuchdata-sharing

easier:Microsoftandothersoftwaremakerswillsoonstarttoim-

plementwhatit callsthe“open-datainitiative”.

Someseesucheffortsasthebeginningof anopen-source

movementfordata,muchliketheapproachthatnowruleslarge

partsofthesoftwareindustry.AndMicrosoft,inparticular,iskeen

toseethishappen.“Weneedtodemocratiseaiandthedataon

whichitrelies,”writesBradSmith,thefirm’spresidentandchief

legalofficerinhisrecentlypublishedbook,“ToolsandWeapons”.

Unsurprisingly,thispositionalsosmacksofself-interest:Micro-

softdoesnotmakemuchmoneyfromdatadirectly,butdoesfrom

toolsandservicesthathandledata.

Liketheoilcomparison,however,thedata-as-sunlightanalogy

breaksdown:opendata,too,cangoonlysofar.Forpersonaldata,

themainlimitationisincreasinglystrictprivacylaws,suchasthe

eu’s GeneralDataProtectionRegulation(gdpr), aswellastheCali-

forniaConsumerPrivacyAct(ccpa),whichwillstartbeingen-

forcedinJuly.Forcorporatedatathechecksareeconomicinna-

ture:generatinggooddataisexpensiveandtheycanrevealtoo

muchabouta firm’sproducts.“Companieswillmakeverystrategic

decisionsaboutwhatdatasetstheywillmakepublicandwhich

onestheywillkeeptothemselves,”explainsMichaelChuiofthe

McKinseyGlobalInstitute,a consultancythink-tank.

Separatingwhatcan besafelyshared

fromwhatshouldbecloselyguardedwill

betricky,buttechnologyshould,intime,

make such decisions easier. Something

called“differentialprivacy”,forinstance,

replacesonedatasetwithanotherthatin-

cludesdifferentinformation,buthasthe

samestatisticalpatterns.“Homomorphic

encryption”allowsalgorithmstocrunch

datawithoutdecryptingthem.Andblock-

chains,whicharethespecialdatabasesof

thesortthatunderliemanydigitalcurren-

Championsof

the“open-data”

movementpush

organisations

togiveaway

theirdata

The EconomistFebruary 22nd 2020 Special reportThe data economy 5

2

1