The Economist - USA (2022-01-29)

TheEconomistJanuary29th 2022 Science&technology 67

whichcantakeoverinordertolandit inan
emergency.Ithasa rangeof35kmbetween
charges,anda topspeedof130kph.
EHanghasformedpartnershipsfortrial
flightsinseveralcities,includingGuang­
zhouinsouthernChina,wherethefirmis
based.Oncea typecertificateisobtained,
thecompany plans to offercommercial
air­taxiandsightseeingflightsfromthese
places,alongwithparceldeliveries.

Crossingcontinents
Forevtols tooperatecommerciallyandat
scaleinmostcountries,threekindsofreg­
ulatoryapprovalwillberequired,saysJen­
niferTrock,whoisbasedinWashington,
dc,andleadstheaviationpracticeofBaker
McKenzie,a lawfirm.Besidesa typecertif­
icate,firmswillalsoneedproductioncerti­
fication,whichpermitsmanufacturingof
anaircrafttobeginatvolume.And,inor­
dertocarryfare­payingpassengers,anair­
line­styleoperatinglicenceisrequired.
Thoughtherearebilateralsafetyagree­
mentsbetweeneasa,caacandAmerica’s
FederalAviationAdministration(faa), the
threeagencieswhichdominatetheworld’s
aviationmarket,certificationbyonedoes
notautomaticallyapplyelsewhere,espe­
ciallyfornewdesignslikeevtols.This,
saysMsTrock,givesmakerstwooptionsif
theywishtoselltheiraircraftinanother
market.Theycansetupa localoperation
andapplyforanothercertificate,asAuto­
Flightisdoing.Ortheycansubmittoa full
technicalvalidationofthecertificatethey
haveobtainedinanothercountry,though
thismightinvolvefurthertestflights.
Thereisalsouncertaintyabouthowthe
ruleswilldifferfromplacetoplace.After
examiningsome 150 designs,easa con­
cluded that evtols were neither aero­
planes,norhelicopters,norboth.Itthere­
foredecidedtomakethema “specialcate­
gory”ofaircraft,withtheirownsetofregu­
lations.Thefaa, bycontrast,saysitcan
certifythembytailoringitsexistingrules
wherenecessary.Ithassetupa unitcalled
theCentreforEmergingConceptsandIn­
novationtoworkwithevtolcompaniesto

establishdetailedsafetyrequirements.

Nonetheless, enough regulatory pro­

gresshasbeenmadeforevtolpioneersto

remainoptimistic.InAmerica,JobyAvia­

tionhopesnextyeartobecomethefirstto

obtaintypecertificationfromthefaa. Itis

makinga dozenormoreaircraftata new

manufacturing plant inMarina, Califor­

nia, to step up its flight­testing pro­

gramme.Joby’sairtaxi,whichseatsfour

passengersanda pilot,usesmultiplero­

torstotakeoffandland.It thentiltsthero­

torsforwardwhencruising,sotheywork

likepropellers.Oneofthecompany’sback­

ersisUber,a firmthatcurrentlymakesits

moneyorganisingterrestrialtaxirides.

HotonJoby’sheelsisArcherAviation,

alsoCalifornian.Archerplansa commer­

cialride­sharingserviceusinga four­pas­

sengers­plus­pilotdesign,tostartin2024.

OneofitsinvestorsisUnitedAirlines.In

Europe,Volocopter,a Germanfirm,istest­

inga multirotorevtolinbothcrewedand

uncrewedform.It hopestoprovideair­taxi

services forthe 2024 Olympicgamesin

Paris. Lilium, another German firm, is

workingonanall­electricseven­seater.

Anditisnotonlynewfirmswhichare

gettingintothebusiness.Airbusisdevel­

opingideasforevtols.SoisBoeing,its

Americanrival.OnJanuary24th,Boeing

invested$450minWiskAero,ofSanFran­

cisco,whichisdevelopingyetanotherau­

tonomousevtol. LarryPage,a co­founder

ofGoogle,isbackingthisfirm,too.Withso

much effortandmoneygoing into this

newformofaviation,someofthesebirds

willsurelyfly.n

It’s a bird...it’s a plane...

Air-taxi investments*, $bn

Source: McKinsey

*Includes venture capital, attributed R&D

funding and announced SPAC mergers

5 4 3 2 1 0

Pre- 212019181716152014

2014

Dataprivacy

Your secret’s safe with me

D

ata arevaluable.  But  not  all  of  them

are  as  valuable  as  they  could  be.  Rea­

sons  of  confidentiality  mean  that  many

medical,  financial,  educational  and  other

personal  records,  from  the  analysis  of

which much public good could be derived,

are  in  practice  unavailable.  A  lot  of  com­

mercial data are similarly sequestered. For

example,  firms  have  more  granular  and

timely  information  on  the  economy  than

governments can obtain from surveys. But

such intelligence would be useful to rivals.

If companies could be certain it would re­

main secret, they might be more willing to

make it available to officialdom.

A  range  of  novel  data­processing  tech­

niques might make such sharing possible.

These  so­called  privacy­enhancing  tech­

nologies (pets) are still in the early stages

of development. But they are about to get a

boost from a project launched by the Unit­

ed Nations’ statistics division. The un pets

Lab,  which  opened  for  business  officially

on January 25th, enables national statistics

offices,  academic  researchers  and  compa­

nies  to  collaborate  to  carry  out  projects

which  will  test  various  pets,  permitting

technical and administrative hiccups to be

identified and overcome. 

The first such effort, which actually be­

gan last summer, before the pets Lab’s for­

mal inauguration, analysed import and ex­

port  data  from  national  statistical  offices

in  America,  Britain,  Canada,  Italy  and  the

Netherlands, to look for anomalies. Those

could be a result of fraud, of faulty record

keeping or of innocuous re­exporting. 

For  the  pilot  scheme,  the  researchers

used  categories  already  in  the  public  do­

main—in  this  case  international  trade  in

things such as wood pulp and clocks. They

thus hoped to show that the system would

work,  before  applying  it  to  information

where confidentiality matters. 

They put several kinds of pets through

their  paces.  In  one  trial,  OpenMined,  a

charity based in Oxford, tested a technique

called  secure  multiparty  computation

(smpc). This approach involves the data to

be analysed being encrypted by their keep­

er and staying on the premises. The organi­

sation  running  the  analysis  (in  this  case

OpenMined)  sends  its  algorithm  to  the

keeper, who runs it on the encrypted data.

That  is  mathematically  complex,  but  pos­

sible. The findings are then sent back to the

original inquirer. 

That inquirer thus receives its answers,

but never has access to the information on

which those answers are based. Moreover,

for extra security, the results are processed

by another pet, called differential privacy.

This  employs  elaborate  maths  to  add  a

smidgen  of  statistical  noise  to  a  result.

That  makes  the  findings  less  precise,  but

means  they  cannot  be  reverse­engineered

to reveal individual records. It also permits

the  organisation  releasing  the  findings  to

set a so­called “privacy budget”, which de­

termines the level of granularity disclosed

by the data. The result is a belt­and­braces

approach.  In  the  argot  of  the  field,  smpc

provides  input  privacy,  while  differential

privacy offers output privacy. 

The unis testing new technology that processes data confidentially