Capgemini-給我畫一個未來-探索工程創(chuàng)新（英）

capgwwueng·neerin9

DRAW

ME

A

FUTURE

"Thesecretofchangeistofocusallofyourenergy,notonfightingtheold,butonbuildingthenew"

Howcanweproduceasmuchcleanenergyas

weneed?Canwetransitiontoacircular

economyandfreeourselvesfromresource

scarcity?Isthereaplaceforhumanworkersinindustry5.0?WillAIbecomeintelligentfor

real?Isthemetaverseafantasyorareality?

Thesearejustsomeofthechallenging

questionsthatengineeringandtechnology

leadersarewrestlingwith.Theanswersthey

findwillshapetheworldof2030andtheworldourchildrenandgrandchildrenwillinherit.

Engineering,thecreativeandpractical

applicationofscienceandmathematicstobringaboutadesiredchangeintherealworld,ontimeandtobudget,isexperiencingagoldenage.Thereponsibilityofthoseinleadershippositions

hasneverbeenmorecrucial.

AsengineeringandR&Dpartnertosomeofthemostinnovativecompaniesand

organizationsontheplanet,weareprivileged

toplayapartintheseincredible

transformationaljourneys.Thisbookcontains

anexcerptofsomeofthemosttransformative

movesatplayandtheassociatedchallenges

thattheworld'smostbrilliantandtalented

engineersareworkingontoday.Ouraimis

neithertoprovideanexhaustiveinventorynor

ahierarchyofalltheinnovationchallengesof

ourtime.Instead,weaimtoofferasnapshot

ofthebreadthanddiversityofthe

technologicalquestionsfacingthecurrent

generationofindustryleaders.Combining

insightandillustrationsontwenty-ninetopics,

wehopethebookwilleducate,entertainand

inspireyoutogetthefutureyouwant.

Therehasneverbeenamoreexcitingtimeto

beanengineerortechnologist,andaswesay

atCapgeminiEngineering,it'stimeforallofus

tothinkbig,bebold,andhavefun.

_FrédéricArquier

ChiefMarketingOfficer,

CapgeminiEngineering

Whybuildingthecirculareconomy isthenoblestof allengineering challenges

MostofusarefamiliarwiththethreeR's-

reduce,reuse,recycle.Buttacklingclimate

changeandbiodiversitylossneedsustogo

beyondexistingleanandgreenpracticestoamorecomprehensiveapproach,wherewasteiseliminated,resourcesarere-circulated,and

natureisregenerated.Thesearesomeoftheprinciplesofthecirculareconomy,anditis

criticaltoourfuture.

Thecirculareconomycreatesvalue,allows

development,andenablesaccesstogoodsandenergyforall.Itisnotaquestionofdegrowthbutefficiency;itisaboutproducingand

consumingdifferentlytoexploitourplanet'sfiniteresources.Nothingislost;everythingistransformed.Itisunderpinnedbyatransitiontorenewableenergiesandmaterials.

Thisneweconomicmodelimpliessharing,leasing,reusing,repairing,refurbishing,andrecyclingexistingmaterialsandproducts.

Whiletheconceptissimple,its

implementationiscomplex.Itrequires

behaviouralchange,thedevelopmentofnewdesignprinciples,andthecreationof

synergizedindustrialvaluechains.

Toscaletheircirculareconomypractices,

Buildingthisnewcirculareconomyneedsa

Redesigningproductsiswhat

organizationsmustembracecirculardesign

systemofsystemsapproachandassociated

engineerslivefor.In2008,

principlesandnewbusinessmodels.

decision-makingtools,relyingontransparent

oneofourteamsappliedfor

Organizationscanpushtheircircularinitiativesforwardbyrethinkingtheirvalueandsupply

datasharedbyacriticalmassofactorsacrossmultipleindustries,banks,andgovernmentsto

theirfirstpatent.

chainsandcollaboratingmorewithintheir

actinacommondirection.

ecosystemsandwithgovernments,lawmakers,

academics,thinktanks,suppliers,vendors,

Whilethischangecanseemoverwhelmingat

clients,andinnovativestartups.Tobuildupthe

first,itisoneofthemostexcitingandnoblest

neweconomicsatscale,theymustaligna

engineeringchallenges.Redesigningproducts

multiplicityofactors.

iswhatengineerslivefor.In2008,oneofourteamsappliedfortheirfirstpatent.Itwas

inspiredbythediscoveryofthecradle-to-cradleconceptandisoneofourproudestachievements.

Willquantumcomputersmaketheimpossiblepossible?

Mostpeoplewillhaveatleastseenaheadlineortwointhenewsabout

Atpresent,quantumcomputers

quantumcomputingbreakthroughsinthelastfewyears.Thistechnology

arelimitedinthenumberof

offersthepromiseofunprecedentedcomputingperformanceforcertain

operationstheycanperform

industry-relevanttasksinsimulation,machinelearning,and

withouterrorsaccumulating,

optimization.

whichgreatlylimitstheirutility

Today,quantumhardwareisstillnascent,withsignificantengineering

challengestoovercomebeforecommercialadoption.Quantum

computersusequantumbits,orqubits,toperformcalculationsusing

quantumalgorithms,someofwhichofferanexponentialadvantageover

algorithmsonconventionalcomputers.Acentralchallengeisimproving

thequalityofthequbits.Atpresent,quantumcomputersarelimitedin

thenumberofoperationstheycanperformwithouterrorsaccumulating,

whichgreatlylimitstheirutility.Beyondquality,thereisalsoaneedto

increasescale-thenumberofqubitsthatcanbeusedforcomputing-

andthespeedofoperationalsoneedtobeincreasedtoenablethe

widespreadapplication.

Theinceptionofquantumcomputinggoesbacksomefourdecades,soitwouldbetemptingtothinkofitasatechnologyalwaysonthehorizon.Butwithtoday'ssmall-scalequantumcomputersnowavailableinthe

cloudforresearchandcommercialevaluation,andwithsustained

exponentialgrowthinthetechnologyprojectedovertheyearsahead,it's

easytoseewhymanythinkthiswillbethedecadethatquantumcomputingwillcomeofage.

Ourabilitytousemachinestocomputehasprofoundlyimpactedour

lives.Itspanseverythingfromthemoonlandingstopresent-dayAI

systemsbeingabletoanswerourquestionsalmostasconvincinglyasif

theywerehuman,andeverythinginbetween.Forallthatcomputers

toucheveryaspectofourlivestoday,wehaveyettoreachthepinnacle

ofeverythingweneedcomputerstodoforus.Wearenotevenclose.Butthecomingquantumcomputingrevolutionwillbefeltacrossall

industries.Wewillfocusonafewareasimpactingeveryone,suchasanewfoundappreciationofhealthcare.Wearefortunatethatthe

breakthroughinvaccinesandtreatmentscameattherighttimeandhad

arealimpact.Theyhadbeenyearsorevendecadesofscientificdevelopmentinthemaking.

Forallthatcomputerstouch

everyaspectofourlivestoday,wehaveyettoreachthe

pinnacleofeverythingweneedcomputerstodoforus

Butwehaveanageingpopulationandfacemanydiseasesforwhichwedonothavecuresorevenmuchinthewayofeffectivetreatments.The

drugdiscoveryprocessexploresvastnumbersofpotentialdrug

molecules,butthisissometimesnotenoughtofindapromisingleadforanewdrug.Ifcomputerscouldaccuratelysimulatetheimpactofnovel

chemistryonhumanbiology,then-incombinationwithmaturingAI

approaches-theycouldpushdrugdiscoveryintoaneweraoffinding

bettertherapeuticsmorequicklyatreducedcost.Thisabilitytosimulatedowntothelevelisoneofthepromisesofquantumcomputing,and

CapgeminiEngineeringiscurrentlyleadingresearchincollaborationwithpharmaceuticalindustrypartners.

Healthcareisnottheonlypressingissueweface.Theneedforclimate

actionisimmediate,butthejourneywillbelong,andmanyemerging

technologieswillplayanimportantroleintheyearsahead.Oneofthe

challengesisfeedingagrowingpopulation.Fertilizersandmodern

farmingallowedustoachievefourtimesthecropyieldsatthestartof

thelastcentury.Butfertilizerisresponsiblefor3-5%ofglobalnatural

gasusage*,whichtranslatestoaround40%ofthecarbonfootprintofaloafofbread.Legumessuchaspeas,beans,andcloverhelpfixnitrogeninthesoilwithoutthisvastenergycost.Wedon'tunderstandthenaturalcatalystswhichhaveevolvedtomakethispossiblebecausetoday's

computerscannotsimulatethemwithoutapproximations.However,

quantumcomputingcouldholdthekeytounlockingthesesecretsandenableustoengineernewprocessesforsustainableagriculture.Itmayalsoimpactthedesignofsolarcellsandbatteries.

Thepotentialimpactofquantumcomputingspansmanymoreareas,

fromoptionspricinginfinancialservicestosimulationintheautomotiveandaerospaceindustries.Whetheritbepricingnewfinancialproductsandmanagingriskordevelopinganewgenerationofmaterialsthatwecannotsimulatetoday,oneday,thetechnologywillbecomeroutine.

Ultimately,quantumcomputingwillnotjustbeaboutbeingfasterormoreaccurate,butaboutmakingtheimpossiblepossible.

Quantumcomputingcould

holdthekeytounlocking

thesesecretsandenable

ustoengineernew

processesforsustainable

agriculture

*Source:InternationalEnergyAgency,Gas2020

PerAsperaadAstra*

TherocketagemayhavestartedinChinainthe12thcenturywith

cardboardandpowder,butitwasonlyin1944thattheGermanV2rocketprogrammelaunchedthefirsteverobjecttoreachthe100kmaltitude

consideredtobethelimitbetweenatmosphereandspace.Theageoforbitalflightbeganin1957withSputnik1pavingthewaytotheorbitallaunchgoldenyearsthatemergedinthe70s:aperiodofpioneers

dominatedbygovernmentalinstitutions.

Itwasin2010withSpaceXthatthelaunchworldenteredanew

commercialerawithhighproductioncadencesandafocusonreliabilityandcostmetrics.Thelaunchpriceof$k/perkilodecreasedby40%andcontinuestofall.

PrivateactorslikeSpaceXandBlueOriginareshakingthewell-

establishedplaygroundofinstitutions.Intheheavyweightcategory,

FalconHeavywillsoonbereplacedbyStarship,andNewGlennwillarrivesoonerratherthanlater.Ariane6willbelaunchedattheendof2023

withamodularandreusablefleet,theNESTSinitiativehasplansforspacetransportforupto2050,andArianeUltimateisalreadyonthedrawingboard.

Butthelaunchrevolutionraceisspeedingup.Modularandreusable

launchersarealreadyconsideredyesterday'stechnology;nanoandcube-sat(10cmx10cmx10cm)satelliteshaveappeared,andwhatwasonce

called“newspace”hasnowbeenreplacedwith“fastspace”.

Soonsolidpropulsionrocketswithbigburningplumeswillbenomorethanchildhoodmemories.AevumhasdesignedRavnX,anautonomousreusabledronethatlookslikeanairfighterandcanlaunch500kg

satellitesinLowEarthOrbit(LEO:between500kmand2,000km).Atthesametime,otherfastspacecompaniessuchasLatitudearedesigning

mini-rocketslikeZephyr,a17m-longrocketabletolaunchnanosatellites

abovea600kmaltitude.Thefirstlaunchisplannedfor2025.California-basedstart-upSpinLaunchhasspenttenyearsdevelopingacatapultthatwillenablefrequentlauncheswithlessenvironmentaldamage.

SpinLaunchconducteditsfirsttestatNewMexico'stestpadinOctober2021.

NewZealand'sRocketLab,foundedin2006,hasdevelopedElectron,arocketcapableofputtinga300kgsatelliteintoLEOforacostofUS$7.5million.Electronincorporatesmanydesigninnovations,including3D

printingandhigh-densitybatteries,topoweritselectropumps,whichfeedpropellantintoitscombustionchambers.RocketLabhas500

employeesandhaslaunchedmorethan100satellites.Ithasrecentlyannouncedthedevelopmentofanewandmorepowerfulrocket:

Neutron.

Willthelauncherofthefuturebemonostagewithlowcarbonimpact,

withanuclearcell,catapulted,hydrogen,orhybrid?Willitself-destruct

orcomeback?Therewillbenewlaunchmethods,newlaunchers,andnoburningplumesinthefuture.Thatwillbethevisioninlessthanadecade.Researchisnotonlyfocusingonsystemsandproducts,butalsoon

propulsion.

Thereisnoone-size-fits-allsolution.Weexpectanad-hocmixofdifferentapproachesdependingontheorbit(LowEarthOrbit,Medium,orbeyond)andforhuman-crewedoruncrewedmissions.Forexample,usinga

catapultlauncherfornanosatellitesbutnotforhuman-crewedmissions.Forlong-distancemissionslikeMars,solutionsintheworksinclude

preparedhubsorgatewaysthatcanbeusedforreloadingorspacetugsthatwillputtheobjectonthecorrecttrajectory.

Thespaceracehaschangedsignificantlyinthe21stcentury.Now,playersfocusonspeed,affordability,sustainability,andreliabilityratherthanthenumberofachievements.Theoldsayingperasperaadastraistrueofthenewspacebusiness.Evenifthespaceconquestisarduous,wewillreachthestars,andmaythebestrocketwin.

…AdAstra

*Throughhardshiptothestars

Howdoyousecurecommunicationsinthe Quantumage?

Morethanever,communicationhasbecomeacriticalaspectofourdailylives.Fromthesimplestofbusinessesrelyingonanonlinepresencetoprosper,topeoplemakingpurchasesatasupermarket,tofriendsand

familycommunicatingthroughsocialmedia,tremendousinformationtransferoccurseverysecond.Protectingthisever-increasingamountofinformation,oneofthemostvaluableresourcesintheworld,isever

morecriticalandcomplex.

Public-keyinfrastructureisusedtodistributekeysthatprotectalmostalldatatransferredovertheinternet.Thecryptographythatunderpinsthesecurityofthisinfrastructureisbasedonmathematicalproblemsthat

are,forallpracticalpurposes,infeasibletosolve,evenwithsubstantialcomputationalcapabilityandtime.However,quantumcomputers

threatentomakethecurrentkeydistributiontechniquesobsolete.Sobetterwaysofkeepinginformationsecureareneeded.Muchoftoday’sencrypteddataisvulnerabletointercept-now-decrypt-laterattacks,

potentiallyjeopardizingbusinessandgovernmentalsecrets.

Thelawsofquantumphysicsallowfornewcommunicationtechnologieswithenhancedsecurityproperties.Quantumkeydistribution(QKD),themostimportantcurrentcandidateapproach,enablesdistantpartiesto

producesecretkeysbysharingquantumsystemswithspecificproperties.

Unlikepublic-keybasedtechniques,wheresecurityisderivedfrom

computationalhardnessconjectures,inQKD,theuncertaintyprinciple–whichlimitsonetoaccessonlycertaininformationaboutphysical

systems;andentanglement–atypeofphysicalcorrelationthatonlyexistsinthequantumworld,providesecurityevenagainstanall-

powerful(computationally)adversary.

Reliablysharingquantumsystemsoverlongdistancesisthemostcriticalchallengeinthecurrentquantumcommunicationtechnologylandscape.Quantumsystems,typicallyintheformofphotons,areprecisely

manipulatedintermsofpolarization,phase,etc.,beforetransmission

throughquantumcommunicationchannelsconsistingofopticalfibersforshortranges,andfreespacesatelliteslinksforlongdistances.Devices

likequantumrepeaters,placedinbetweencommunicatingparties,willplayanessentialroleindevelopinglong-rangequantumchannels.

Theintegrationofthesetechnologiesintoexistingnetwork

infrastructurehasalreadybegun.Moreover,researchisbeingdone

towardsquantum-basedservicesbeyondQKD,suchasE-voting,secretsharing,digitalsignatures,andsecuremultipartycomputation(apotentcryptographictoolthatprovideswaysformultiplepartiestoperform

jointcomputationsontheirprivateinputs,withaguaranteethattheirindividualinputsstayhiddenduringthecomputation).

Everyadvancegetsus

closertoaninterconnectedworldinwhichallour

communicationsare

protectedthrough

quantum-generatedkeys

Everyadvancebringsusclosertoaninterconnectedworldinwhichallourcommunicationsareprotectedthroughquantum-generatedkeys.Once

wehaveaccesstosuchquantuminternet,withtheadvancesoftraditionalcommunicationsystemsandtheadventoftheInternetofThings,wewillliveinaworldwhereallpeopleandtheirdevicesareconstantly

communicatingamongthemselvessecurely.Ironically,thisrobustaddedlayerofsecurityisprovidedbythetiniestofparticles,whichisabeautifulthingtothinkabout!

Whatisstoppingusfromhaving flyingcars?

Twosignificantchallengesaredisruptingthe

aviationsector:theurgentneedforzero

emissionsandtheunstoppableshifttowards

connectingcitiesandunservedregionslackingpropergroundinfrastructure(railorroads).Weareenteringthemostexcitingchapterin

aviationsincethedawnofthejetage.Itwillbemarkedbyeye-catchingandrevolutionary

alternativestoaerialmobility.

Fordecades,wehavedreamedabouthoppingbetweencitieswithpersonal“flyingcars.”Butmovingfromavisiontoaviablealternativetocurrenttransportationsolutionsisabigstep.ThefundamentaltechnologiestoenableinnercityhopsinsmallVTOL(verticaltakeoffand

landing)aircraftmustbereadytoscaleand

operatesafely.Combustionengineswere

nevergoingtounderpinaircraftthatcouldrunsilentlyoratapricetagthatwouldappealtothemasses.Helicoptershaveflownabove

citiesfordecades,butthehighnoisepollution,

greenhouseemissions,andoperatingcostspreventedthemfrombecomingviable

alternativesforeverydaycommutes.

Today’sacceleratedgrowthinbattery

technologies,coupledwiththeadventof

distributedelectricpropulsion,autonomous

software,andenhancedpassenger

experiences,haveenabledthere-designof

aerialvehiclesfromsquareone.Wefinallyhavethetechnologiestodeliveralmostsilent,zero-emission,andaffordableairtaxioperations.

Theelectricaerialrevolutionisinprogress.Inadecade,itwillbecommonplacetoseeflyingairtaxishoppingfromcitytocityandunmanneddronesdeliveringgroceriesinminutes.

MostadvancedmanufacturersarerunningtrialsandtestingmultipleelectricVTOL

architecturesandapplications,gathering

valuabledataonsystemreliabilityandflightperformance.

Nonetheless,the“road”aheadislongand

complex.Safetystandardsforcommercial

aircraftarestringent,andweexpectthemtobeevenstricterforurbanairmobility.The

mostcriticalhurdlemanufacturersmust

overcomeincertifyingtheaircraftisaprocessthatdemonstrateshowtheaircraftmeets

EASA/FAAsafetyrequirements.Eachaircraft’scomponentmustbedesignedand

manufacturedtocomplywiththesestandards.

Unlikeconventionalaircraftorhelicopters,

novelelectricVTOLaircrafthavenumerous

innovationsanddesignchoicesincompatible

withexistingregulations.Toaccelerate

conventionalcertificationprocesses,additionalconsiderationsmightarise,suchas:

—

Whatwillittaketodemonstratethatthesenewaircraftarchitecturesareatleastassafeas

commercialaviation?

—

Canadigitalizedcertificationworkflowmaintain

therequiredtraceabilityfromregulatory

airworthinesstoevidencecompliance,enablingautomatedVTOLtestcasegenerationand

systemstrade-offanalysis?

—

Candata-drivenmachinelearningautomatetestscenariogeneration?Canevidencegenerationbeacceleratedbyconductingreal-worldtestsinasdata-richamanneraspossible?

Weareconvincedthatby2030,flyingin

electricVTOLaircraftwillbepartofourdaily

habits;thischangewillbeahugestepforwardtowardfullsustainabilityforthewholeaviationindustry.Combinedwithautonomous

transportation,advancedairmobilitywillsoonenablegreener,connectedcitieswheretrafficcongestionisamerememory.

Movingfromcombustionenginestoelectricpropulsionisanessentialsteptoward

sustainability,andwearewellonthepathto

gettingthere.Electricairtaxisareonlythefirst

stepforamuchwiderjourneythatwillseeaviationreachitsemissionsgoalsand

eventuallybringbackcommercialsupersonictravel.However,thistimesupersonicwillbeelectric.

Canweturnoffthelightsin factories?

“Lightsout”manufacturingisusedfor

manufacturingprocesseswherefactoriesrunautonomouslywithouthumanintervention.Thetermisquiteliteral,withproduction

occurringfreeofhumannecessitieslike

lightingorheating,ventilation,andair

conditioning(HVAC).FamousSci-Finovelist

PhilipK.Dickfirstreferredtoitinhis“Autofac”shortstoryin1955.

Therehaslongbeenapromiseofasmart.

everything-can-dofuture.Ontheotherhand,it’shardformanymanufacturerstolook

aroundanexistingfacilityandimaginethatitcouldonedaybetransformedintoafactoryofthefuture–whetherlights-outorotherwisedata-intensivetoreachnewproduction

efficienciesandqualityofoutput.

But,withtheadventofArtificialIntelligenceandthedigitalizationoftheshopfloor,thisislikelytomakeitswayfromfictiontorealityinthecomingyears.However,althoughan

increasingnumberofstepsandmanufacturingprocesseshavebecomethewaytoturnthe

lightsoffisstillfullofchallengesandpitfalls.

Thenumberonetechnologicalcornerstoneinreachingthisstageofcompleteproduction

automationrevolvesarounddata.Datais

everywhereinthefactory:fromCAD/CAM,

PLM,andsimulationsystemstosensorsattheendofthemanufacturingline.Butmuchofthisdataisunstructured.Toturnbigdatainto

actionablesmartdata,itmustbeaggregatedandcontextualized.Thisstructureconnectsdatapointswiththeirmeaningsothatdigitaltoolscanbeusedto

detecttrendsandidentifyproblemareas.

Insightslikeareasofhighdowntimefora

specificmachine,toomanyoperatorerrors,

productbuildorsupplyissues,longleadtimesduetooperator-intensivemonotonouswork,

oranoverabundanceofnon-valueadd

activitiesbecomevisiblewhenmanufacturingdataisharnessed,organized,andevaluated.

Dataisnotbeingusedtoitsfullpotentialbutinisolation,though.Inmanyfactories,variousdepartmentsworkonsystemsthatarenot

interconnected.Thismeans,forexample,thataqualityissuedetectedinawork-in-progress(WIP)onthefactoryfloorisdifficulttoconnecttoavariationinthepropertiesofanincomingcomponentorrawmaterialthatmayhave

causedtheWIPproblem.

So,thejourneytolights-outandotherhigh-

efficiencyproductionschemesbeginswith

somethingalreadypresentinmanufacturingoperations:data.Connectingitacrosssystemsandwithmeaningfulconceptsoftentakestheformofadigitaltwin:aninteractive,virtual

representationofaphysicalproductor

process.Digitaltwinsenableengineerstoexplorealternativeapproachesandwhat-ifscenariosbeforetheytaketheplungewithsignificantcapitalinvestments.Digital

manufacturingtoolsandthedigitaltwinare

designedtoreplicateautonomousprocessesinthevirtualrealm,wheretheycanbeoptimizedbeforedeployment.

Inanutshell,thetechnologiesalreadyexisttoautomatefurtherthemanufacturingofgoods,butarewereadyasasociety?

Aswitheverysuchdigitaltransformation

movement,thepoweroftoolsisnottheonlynorthemostcriticalsuccessfactorofchange.Culturalchangeinsidecompaniesandbroadersocietyisfarmoreessentialandistwo-sided.

Ontheonehand,automationwillreplacecurrentproductionjobsandlessqualifiedworkers,threateningoursocialclasseswithdisplacementandoursocietieswithunrest.

Ontheotherhand,adoptingsuchlights-outfactorieswillcreatetheneedfornewjobs

aroundmanufacturing.Awholefamilyofnewpositionsandskillsislikelytoemergeinthe

design,planning,control,monitoring,and

maintenanceofsuchautonomousproductionsites.

Howwillcompaniesandpublicpowers

accompanythisshifttowardsaneweconomicorganization?Asinmanycases,technologywillsucceedatscaleifitisperceivedasatoolforhumans,notifitresultsinathreattothem.

ThefutureofAI: superdeep learningorrealworldreasoning?

ArtificialIntelligenceandMachineLearningarenotnewfields;they’vebeenaroundlonger

thanmostpeoplerealize.Rightnow,though,they’reatahugelycriticaljuncture,withthenextdecadepossiblybeingoneofthemostsignificantintheirhistory.

Fromwatchingthemediaoverthelastfew

years,you’dbeforgivenforthinkingthatAI

wasnowasolvedproblem,astherehavebeenmanystoriesofimpressiveAIsystems

outperforminghumans.Ifthat’strue,though,wherearealltherobots?Whereisthesuper-intelligentAIthatwillsolvealltheworld’s

problems?

ThemoreboringtruthisthatAIhassolved

someverynarrowproblemsinverynarrow

ways.Theseareextremelyimpressivefeatsofengineeringbut,forthemostpart,arenot

intelligentandrarelysurvivefirstcontactwiththeinfinitecomplexityoftherealworld.ThishasledtoalacklusteradoptionofAIacross

industry,withthesecleverbutfragileAI

solutionsdeployedonlyinnarrowlydefinedlow-riskusecases.

Thiscommercialrealitycontrastsstarklywiththehigh-profilepicturepaintedinthepopularmedia,whichseemstoshowimpressivefeatsofintelligence,fromapparentlysentientAI

languagemodelstogallery-worthygenerativeartworks.However,theseremarkablefeatsarejusttheresultofadecadeofrapidprogress

usingartificialneuralnetworks.Inrecentyears,wehavemasteredtheartofcreatinghuge

machinelearningmodelswhichcansowhollyandprofoundlyidentifythepatternspresentinspecificdatasetsthattheycanpredicttheir

outputsormanipulatetherelationshipsinthedataatwilltoproduceendlesscreative

variantsfromit.

Thisrealizationthatscalealonecanproduce

intelligent-lookingresultshasbeenoneofthedrivingfactorsinAIoverthelastdecade.Thisphilosophyisbehindoneofthetwomajor

factionsemergingwithinAI,wheresome

peoplebelievethattrueintelligencewill

emergeifonlywecouldmakethemodelsbigenough.Othersbelievetheselargemodelsarenothingmorethanglorifiedcurvefitting–a

statisticalmimicryofintelligencethatcan’t

extrapolatebeyonditstrainedexperience.Thissecondfactionbelievesthattrueintelligencecanonlycomefromsystemsthatexplicitly

reasonaboutreal-worldconcepts(so-called“symbolmanipulation”),developingatruegeneralintellectratherthanjustcopying

patternsofbehaviorfromotherintelligentbeings.

Thedebateragesonbetweenthetwocamps,buttherealityisthatbothareprobablyright

tosomeextent,an