AFRGA1 A019

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

www.afr.com | Tuesday 5 September 2017

Special Report

Artificial intelligenceEdited by Jason Clout: jclout@fairfaxmedia.com.au

Why we mustn’t be left behind

From left, moderator Ian Grayson, Sarah Adam-Gedge and Adam Driussi at the roundtable. PHOTO: DEAN SEWELL/OCULI PHOTOS

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Industry Doubling thepace of automationcould boost growth.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Mark Eggleton

In the digitaleconomy distanceis no longer reallya factor.

Could Australia become an artificialintelligence (AI) powerhouse by 2030 ifwe doubled our pace of automation assuggested in the report, The Automa-tion Advantage, commissioned byGoogle and undertaken earlier thisyear by AlphaBeta?

According to the report, doubling thepace of automation could provide astrong catalyst for economic growthprovided the right policies are put inplace and business gets behind thepush. Citing Deloitte’s 2016 GlobalManufacturing Competitiveness Index(GMCI), which says the US will be theworld’s most competitive manufactur-ing destination by the end of the dec-ade, the report reasons if we follow theUS example towards automation, theupside is enormous.

Much depends on the research youcite to back up your argument. A recentMcKinsey Global Institute report sug-gests China’s current rate of AI-drivenautomation could boost its GDP by 1per cent annually, which would prob-ably keep China in the number one pos-ition and not the US.

Either way, the consensus from theglobal professional services firms aswell as the World Economic Forumseems to be AI-driven automation willspur productivity and prosperity glob-ally. The reason is automation isprimarily designed to drive human pro-ductivity and not human redundancy.

In a survey in June by Microsoft andAccenture’s business technology jointventure Avanade, the understandingwas organisations cannot drive anymore productivity out of optimisationapproaches such as industrialisationand labour arbitrage. The answer doeslie with AI as it will augment thehuman workforce and free employeesfrom the more mundane tasks.

The global survey of business andtechnology executives found that 31 per

cent of organisations have alreadystarted using intelligent automation tobreak through the productivity plateau,with that number set to double by2020. Moreover, 86 per cent of surveyrespondents believe they must deployintelligent automation to be a leader intheir field.

The problem for Australia, accordingto Avanade Australia’s managing dir-ector Sarah Adam-Gedge, is we arealready a fair way behind the rest of theworld. Moreover, Adam-Gedge believeswe have fallen to a point where the gov-ernment really needs to provide moreincentives for business to automate aswell as offer more to companies off-shore so they will invest in Australia.

Chief executive of data consultancyQuantium, Adam Driussi, suggests theonus should fall more on existing busi-nesses than government assistance.

‘‘Governments should help makeeducation better, so that you’ve actuallygot better educated people coming outwho make better decisions. Govern-ment should not get in the way of busi-ness. For example, they’re currentlygetting in the way of us with their 457

visa changes so we’re not allowed tohire people who will help our businessgrow,’’ he says.

‘‘As for funding to help businessgrow, we started our business withoutany funding and that goes to the heart ofbeing an entrepreneur. We need peoplewho can build a business that doesn’trequire funding.’’

Driussi, who was speaking at therecent AI roundtable co-hosted by The

Australian Financial Review and theAustralian Computer Society (ACS),also pointed out corporate Australiahas often used the tyranny of distanceas an excuse to not invest or innovate,but in the digital economy distance isno longer really a factor.

Shadow minister for the futureworkplace and the digital economy Ed

Husic says that a fire has to be lit undercorporate Australia when it comes toinnovation because there is not anysustained investment in automationcompared to other countries.

‘‘What government needs to do iswhat it’s supposed to do, which is interms of education and investing inschools or removing regulatoryhurdles,’’ Husic says.

University of Technology Sydney’shead of school, software in the facultyof engineering & IT, Michael Blumen-stein, agrees there is a place for govern-ment support.

‘‘It’s usually in collaboration withindustry and even the academic area.For example, Israel is successfulbecause there’s collaboration betweengovernment, academia and industry.’’

Avanade’s Adam-Gedge says herview on what government supportcould be is wide-ranging.

‘‘It could be lower corporate tax ratesor it could be a policy change or focus-ing investment into an industry whereAustralia already enjoys some compet-itive advantage such as healthcare oragriculture.’’

Boards need totake a moreactive role

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Governance● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Mark Eggleton

Australia’s leading companies need tohave a better understanding of therapid rise of artificial intelligence (AI)and what it means for their businesses.

Over the last couple of years, therehas been a strong focus on big data andwhether senior management actuallyunderstand what it means or how to useit but there is not much awareness of AI.

Participants at the recent AI round-table co-hosted by The Australian Fin-ancial Review and the AustralianComputer Society (ACS) all agreed itwas time for boards to get up to speed.

‘‘Every board of a major ASX com-pany is telling their CEO they’ve got tohave a data analytics story; they’ve gotto have a big data story but they’re notyet telling them they have to have an AIor machine learning story,’’ said theprincipal at Data Synergies, PeterLeonard.

‘‘The boards probably don’t evenknow what machine learning or AI is.For me, the first step is to simply get thelevel of awareness in the business com-munity to the stage where the board isforcing the CEO to do something.’’

Managing director of Avanade Aus-tralia, Sarah Adam-Gedge said theproblem with many local boards is theyonly choose to think about AI or dis-ruption when they are threatened bytechnology or a new competitor.

Shadow minister for employmentservices, workforce participation andfuture of work and the digital economy,Ed Husic said the problem is genera-tional as there is a ‘‘digital illiteracy inolder generations that have been call-ing the shots in management versusthe new generation of managementcoming through’’.

Adam-Gedge acknowledged there isan age issues around boards and man-agement but in a recent global surveyundertaken by Avanade on automa-tion, more than half of global businessleaders believe that an understandingof new and emerging technologies,such as AI, will be more important forleadership than traditional specialisa-tions like sales and marketing by 2022.

AFRGA1 A020

AFRTuesday 5 September 2017The Australian Financial Review | www.afr.com

20 Special Report Artificial intelligence

Confrontingthe new reality

Ian Opperman: ‘‘One really interesting benefit will be an increasing degree ofpersonalisation and flexibility in everything we do.’’ PHOTO: DEAN SEWELL/OCULI PHOTOS

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Innovation● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

James Sherbon

Building a world-scale business fromAustralia will beeasier than ever.Peter Leonard, Data Synergies

Artificial intelligence is already impact-ing millions of people. By the end of thisyear, IBM Research’s lab directorJoanna Batstone says, there will bemore than a billion people touchingIBM’s Watson technologies in someform or other.

‘‘If you pick up your phone and youcheck the weather company, it’spowered by Watson. If driving a num-ber of luxury car brands that have allkinds of electronic intelligence in thecar, they’re leveraging some of the APIsfrom the Watson platform,’’ she says.

‘‘So, the point is AI technology is get-ting into the hands of the end user veryquickly and the reason for that isbecause it drives a value propositionand people are embracing that newtechnology,’’ Batstone recently told anAI roundtable co-hosted by The Austra-

lian Financial Review and the Austra-lian Computer Society (ACS).

Yet while the future is now when itcomes to AI, the key question is howwill it evolve? More pertinently, is itreally going to be the greatest inventionin our lifetime?

A roundtable participant, PeterLeonard, who is the principal of DataSynergies, referenced an episode ofABC-TV’s You Can’t Ask That, in whichcentenarians nominated the washingmachine as the greatest innovation oftheir lifetime – greater than the motorcar, aircraft travel or nuclear weapons.The reason was it saved more time andarduous labour in the home than any-thing else.

Leonard wondered what AI willdeliver to the consumer? How will itchange lives truly for the better beyonddigital novelties?

Chief executive officer of the NewSouth Wales Data Analytics Centre DrIan Oppermann says everyone will

benefit, especially in certain work-places. ‘‘Teachers will be betterinformed to teach and doctors will bebetter informed to diagnose,’’ he says.

‘‘Some professional services will beimpacted such as junior level lawyersand accountants but one really inter-esting benefit will be an increasing

degree of personalisation and flexibil-ity in everything we do.’’

Leonard sees two big prizes for Aus-tralia’s economy as long as we startmoving quickly.

‘‘For one, where the machines arewill be irrelevant, so the age-oldtyranny of distance or the need to scale

will not be important because anyorganisation can access IBM Watson,Google DeepMind or any other intelli-gent system and those systems can befrom anywhere in the world.

‘‘So, if you’re developing applicationsthat rely upon deep intelligence andanalysis of data you can do it as wellhere in Australia as you can do it in Ice-land or Silicon Valley or anywhere elsein the world so long as there’s goodbroadband back to wherever in theworld you’re wanting to deliver the ser-vice to.

‘‘Secondly, building a world-scalebusiness from Australia will be easierthan ever before because many of thesefuture enterprises will be service busi-nesses.

‘‘The real grunt will be provided bythe support service, which can be donefrom Australia with the service deliv-ery overseen on the ground by a relat-ively small crew in the relevant othercountries.’’

Robot technology could spur growth

Labor’s Ed Husic warned there was a large cohort of workers worried about their future. PHOTO: DEAN SEWELL/OCULI PHOTOS

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Training Australiamust move quickly tobe globally competitive.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Mark Eggleton

AI is already startingto have an impact inthe creativeindustries.

Depending on whom you believe, tech-nology is going to automate and des-troy more than 50 per cent of all jobs orit is going to usher in a new age or newwave of careers.

The real answer lies somewhere inbetween. In fact, the things that reallyneed to worry about their jobs are thecurrent generation of robot technologybecause they will be superseded, butthey cannot think for themselves ... orcan they?

The recent report, The AutomationAdvantage, commissioned by Googleand conducted by AlphaBeta, foundautomation presents a $2.2 trillionopportunity for the Australian econ-omy and could potentially create mil-lions of jobs.

The report found Australia lagsbehind the rest of the world with only9 per cent of listed companies makingsustained investments in automation,compared with more than 20 per centin the United States and nearly 14 percent in leading automation nationsglobally.

According to the report, if Australiaaccelerated its automation uptake itwould stand to gain up to another$1 trillion over the next 15 years.

The report also found only 29 percent of automation-driven workplacechange will involve workers changingjobs.

Of this 29 per cent, automation willlikely create new jobs for those dis-placed workers. In fact, if anything, thereport suggests automation will be likeprevious periods of technologicaladvancement and there will be a waveof new jobs created.

Yet while the potential is strong, Aus-tralia really needs to move quickly if wewant to be globally competitive in thedigital age. What it will take is a concer-ted and combined effort from govern-ment, the corporate sector, educationproviders and trade unions to preparepeople for the automation age.

Part of that preparation involvesensuring people receive the right type

of education and retraining. This mightmean more of a focus on the science,technology, engineering and math-ematics disciplines, but it also meansmaking people aware of the breadth ofAI’s potential.

Speaking at the recent AI round-table co-hosted by The Australian Fin-ancial Review and the AustralianComputer Society, IBM Research’svice-president and lab director JoannaBatstone said AI was not just abouttransforming traditional computerscience-based technology industries,‘‘it’s already starting to have an impactin the creative industries’’.

‘‘People are using AI natural lan-guage processing techniques to writepoetry, to write music, to analyse pat-terns in music to create movie trailersby doing visual recognition,’’ Batstonesaid.

She said we can inspire people aboutthe skill sets we will need in the future

by showing the opportunities AIaffords.

‘‘Showing that it can transformmovies, theatre, poetry and art is just asimportant as showing how it can trans-form computer science.’’

Yet while it is easy to motivate thenext generation about automation, fel-low roundtable participant Ed Husicwho is the federal shadow spokesmanfor employment services, workforceparticipation, future of work and thedigital economy, warned there is a largecohort of workers worried about theirfuture.

He says a huge number of jobs willbe affected and we are not having theright conversations about how we aregoing to ‘‘reskill’’ those people.

‘‘We’re going to have to,’’ Husic said.‘‘We really need to work out how we dothat.

‘‘The challenge is we have a cohort ofpeople in the middle who were trainedup with the belief that if they went toschool, TAFE or university and pickedup a bunch of skills, they would be in ajob.

‘‘They knew if they changed jobs, thenew job would pretty much involve thesame skill set all the way through.

‘‘Now, they’re going to be doing allthese different roles which require dif-ferent skills, so how do we improve theability to acquire skills in a shorterspace of time so people will be able toget the next job?’’

According to Batstone and otherroundtable participants, it comes

down to the notion of lifelong learningwhich is easier said than done.

‘‘As the world moves to AI, we haveto retrain and reskill our own work-force. It has to be a focus on learning,skilling, retraining and educating andbringing your workforce, your com-munity along,’’ she said.

For Batstone, this will require newlearning styles, techniques and techno-logies. It will mean more personalisedlearning, or as roundtable participantDr Ian Oppermann suggested, ‘‘micro-credentials’’.

Dr Oppermann, who is the CEO ofthe NSW Data Analytics Centre, saidvocational education would not needto be acquired in the traditional man-ner of a university or TAFE, it could bea whole series of micro-credentials orcourses delivered to an individual.

Interestingly, Michael Blumenstein,who is the head of school, software inthe Faculty of Engineering and IT at theUniversity of Technology Sydney(UTS), said universities were alreadyembracing micro-credentialling espe-cially in post-graduate studies.

‘‘People want flexibility and theyknow something’s coming and theywant to change,’’ he said. ‘‘They want toupskill, but they don’t have the time todo it. So, this micro-credentialling canactually be used as a way to actually dolittle things over time at your discretion,whenever you want, but then pack ittogether to actually give a qualificationthat’s recognised.’’

Dr Oppermann pointed out oneplace where AI was already playing arole in planning for the future is in theTAFE system in New South Wales. Hesaid AI was currently reading everyTAFE course description in the stateand also reading every job ad online inall of Australia and it was looking atwhere the skills were needed.

‘‘It’s doing a simple supply anddemand match through the NSW DataAnalytics Centre,’’ he said. ‘‘We’re look-ing at where we’ve got over- or under-investment and looking at the trends ofjobs in the regions.

‘‘Furthermore, we’re also looking atcourse compositions and asking howcould you take this bit of a course andput it into another course? It meanscreating little sub-sections of courses,an eight-hour part of a TAFE course."

The point of doing this is to createthe micro-credentials people will need.

AFRGA1 A021

AFR Tuesday 5 September 2017www.afr.com | The Australian Financial Review

21Special ReportArtificial intelligence

IndustryInsightIndustry comment byAnthony WongPresident of the AustralianComputer Society

Responsible AI callsfor the incorporationof ethical and socialvalues at the designstage.

When it comes to ethics, it’s a multi-dimensional issue.

ThiscontentwasproducedbyThe AustralianFinancialReviewin commercialpartnershipwiththe AustralianComputerSociety.

Inevitablywhenweascribeintelligencetosomething,questionsofethicswillsoonfollow.

Asweprogresstoafuturewhereartificialintelligence(AI)hasthepotentialtoaffectmanyareasofourlives,it’snosurprisethatheateddebatesaroundemployment,security,publicsafety,rightsofrobots,regulation,andsocialandethicalconsiderationsarenowbubblingtothesurface.

RecentlyIwasinvolvedinanumberofexpertpanelsonthepotentialchallengesandsocialconsequencesofAIattheInternationalJointConferenceonArtificialIntelligence(IJCAI)inMelbourne.Thedebatesanddiscussionswerefascinatingandinsightful,butiftherewasonetake-away,it’sthatwe’reonlyseeingthetipoftheiceberg.

Thebigplayers–Amazon,Facebook,Google,IBMandMicrosoft–havealreadyformedthenon-profitThePartnershipforAIto‘‘studyandformulatebestpracticesonAItechnologies,toadvancethepublic’sunderstandingofAI,andtoserveasanopenplatformfordiscussionandengagement’’.

Anothernon-profit,OpenAI,withElonMuskandSamAltmanofYCombinator,aimstodiscoverandenact‘‘thepathtosafeartificialgeneralintelligence’’.

Whenitcomestoethics,it’samulti-dimensionalissue.

Beyondthedecisionsmadebyintelligentautonomousmachinesinthefuture–‘‘shouldtheautonomouscarsavethedriverorthepedestrians?’’isaprominentrecentdiscussion–wefirstneedtorecognisethatresponsibleAIdevelopmentcallsfortheincorporationofethicalandsocialvaluesatthedesignstagebyprogrammersandengineers.

Whichnaturallyleadsto:whoseethicalandsocialvalues?Whatbiasesdowebuildin–intentionallyorotherwise–thatwillaffecttheoutput?Eachindividual,socio-culturalgroup,andnationalgeographycanhavedifferentattitudestoethics,moralityandlegality.

ThisisfurthercomplicatedbythechoiceofdatausedtotrainAIsystems.Eventhebest-laidintentionscangoawryifthedataitselfbreaksourethicalstandards.

AndjustlastweekitwasrevealedtheAustraliangovernmenthadrolledoutthe

useofanalgorithmacrossitsimmigrationdetentioncentrestodeterminethesecurity-riskofdetainees,replacingdecisionmakingbyhumans.TitledSRAT(SecurityRiskAssessmentTool),itconsidersdataaroundpastbehaviour,age,healthandknownassociates.Itsfocusisnarrow,butitnonethelessmakesdecisionsonthewelfareofpeople.

Accordingtothegovernment,incidentsofassaultandself-harmhavereducedsinceitsimplementation,butithasbeencriticisedthatitdoesn’ttakeintoaccountrehabilitationorchangesinmentalhealth.

It’saninterestingquandary.Howmuchresponsibilityshouldwegivetoalgorithmsshieldedbehindmathematical‘‘impartiality’’?

Asmachinelearningdevelopsitcanbecomedifficulttoarticulatetheinnerworkings.FacebookrecentlyshutdowntwoAIchatbotsaftertheyunexpectedlystartedtodeveloptheirownlanguage,

demonstratingthecapabilityforthebotstocreatetheirownmeaning.Andasethicsaresimplymeaningswechoosetoascribe,canweexpectAItoactethicallyifitcancreateitsown?

Clearlythetimesaheadareinterestingandchallengingaswecometogripswithwhatitmeanstoregulateinaworldofautomateddecision-makingsystems.

TheEUhasalreadyactedwithaGeneralDataProtectionRegulation,comingintoforcein2018,providinganindividualwiththe‘‘righttoexplanation’’toautomateddecisionmakingwheretheindividual’sinterestshavebeensignificantlyaffected.Thereisalsoarighttoappealadecision.

Forthefirsttime,technologyisnotonlyautomatingtherepetitiveandthemanual,butalsograduallysupplementingandmimickingour‘‘minds’’.

Asalgorithmiccomplexityandautonomyincrease,itbecomesevermoreimportanttobuildinchecksandbalances–tonotjustexplore,buthopefullyensure,afutureofethicaldesign.

IBM research vice-president and lab director Joanna Batstone at the AI roundtable. PHOTO: DEAN SEWELL

Setting the frameworkfor effective regulation

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Ethics Experts ponderthe issues in a fast-moving field.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Mark Eggleton

The really big questionis ... just because wecan, should we?Dr Ian Oppermann, CEO NSW DataAnalytics Centre

One of the reasons HBO’s Westworldis such compelling television is itsfocus on the consciousness of thehosts. Rather than the hosts follow-ing the well-trodden dystopian killerrobot path, they seem to be searchingfor a human truth as their mindscrunch through the algorithms oftheir human handlers.

While the program is primarily agorgeous looking confection, itsbroader themes do examine some ofthe ethical questions associated withartificial intelligence (AI). Questionssuch as how far do we let it evolve?Will it always rely on human pro-gramming, and where is the AI wealready take for granted leading us?

Speaking at a recent AI roundtableco-hosted by The Australian Finan-cial Review and the Australian Com-puter Society (ACS), the chiefexecutive officer of the NSW DataAnalytics Centre, Dr Ian Opper-mann, suggested there are numberof questions we need to ask as AIdevelops more rapidly.

‘‘We will see more smart devicesand to some extent the rules aregoing to be governed by what’sacceptable in the marketplace onone hand and improved governmentregulation on the other. The reallybig question is ... just because wecan, should we?

Dr Oppermann’s views echo thecautious tone of physicist StephenHawking and, more recently, ElonMusk. He wonders about the pos-sible and the ethics about what weare doing.

The ethics behind what we aredoing saw most of Silicon Valley’stechnology giants pledge to worktogether to ensure the AI tools theydevelop are safe. They also co-signedin 2015 an open letter committingthemselves to not developingautonomous weapons, but Opper-mann’s concern is that we do not fullyunderstand the art of the possible sowe may not be asking the right ques-tions of ourselves right now.

For example, we are not askingwhat the ethics are behind setting upa weapon that fires at will and kills.

Similarly, the ethics around whethera self-driving vehicle hits the pedes-trians or kills the driver? They aredifferent ethical questions than youwould ask of a human soldier ordriver under the same circum-stances.

Oppermann said we should not beafraid to continue developing AI aswe must be ‘‘bold about exploringthe limits but we should also build innew regulations’’ and provide an eth-ical reference framework for people.

Michael Blumenstein, the Head ofSchool in the Faculty of Engineering& IT at the University of TechnologySydney (UTS), suggests we are beinga little alarmist about where AI istaking us because we are a long wayaway from the ‘‘Terminator situ-ation’’.

He suggests the people behind thecreation of AI and machine learningare primarily engineers and scient-ists who want to provide a techno-logy that augments our humanworld.

‘‘Engineers don’t say we’re going tobuild a safe bridge, they say we’re

going to build a bridge. It’s implied inhis or her work, they’re alreadybuilding a structure that’s going to besafe, so we should take the same per-spective on artificial intelligence,’’Blumenstein said.

Conversely, Blumenstein said wewould not really know when wecross the line and machines becomesentient and where there is potentialfor things to go awry.

‘‘Things are moving so quickly, wecan’t forecast for it and it’s going tobe very difficult to regulate and Idon’t think governments will be fastenough to keep up with the techno-logy. We just need to be cautiouslyoptimistic things will go in the cor-rect direction.’’

IBM research vice-president andlab director Joanna Batstone israther more sanguine and suggestsone of the fundamental purposes ofAI is for it to augment human activ-ity.

‘‘It’s not about replacing humans,

it’s about augmentation,’’ Batstonesaid. ‘‘What we need to do is buildtrust and transparency in new tech-nologies and the way you build soci-etal trust is by observing ethicalresponsibility to set in place govern-ment standards and privacy stand-ards around data and that’s apartnership between the public andprivate sector.’’

Batstone said IBM had engagedearly in the conversation aroundwhat responsible AI meant acrossthe political and public spectrum.

Sarah Adam-Gedge, the managingdirector of Microsoft and Accen-ture’s business technology joint ven-ture Avanade Australia, suggestedthe future of AI and ethics was extra-ordinarily complex.

‘‘You will need a combination ofinsights and people to come up withthe answers to the bigger questions.We need to ask what are the inten-ded and the unintended con-sequences of a new technology?

‘‘At Avanade, we believe there arefar-reaching consequences andorganisations need to own it, but thestudies we’ve done have shownorganisations are thinking about itbut, they haven’t necessarily put theframeworks in place to enable peo-ple to make sure the technologiesthey’re building just have the positiveconsequences that were intended.

‘‘So, I think there do need to beguidelines and frameworks [aroundAI] and we obviously need to makesure in a business and a governmentsense there are good security frame-works around safeguarding people’sdata.’’

Peter Leonard, principal of DataSynergies, suggested it was hard toregulate an area ‘‘where you don’teven know how people are going touse the technology’’.

‘‘There is a big difference betweenartificial intelligence and fullyautonomous agents,’’ Leonard said,‘‘so most artificial intelligence isreally about enabling machines toassist us to make decisions better. Ithink for many of the applications ofAI where the machine intelligence issupplementing human intelligencethere’s no need for change of existingregulation.’’

In regard to fully autonomousagents, Leonard suggested a fairlysimple solution. ‘‘The person whoprograms and enables the deviceshould be responsible for the con-sequences of what follows.’’

AFRGA1 A022

AFRTuesday 5 September 2017The Australian Financial Review | www.afr.com

22 Special Report Artificial intelligence

Impact to spreadacross sectors

Autonomous cars are just the start of the AI revolution. PHOTO: BLOOMBERG

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Business software AIwill be imbedded inalmost every product.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Ian Grayson

Rapid advances in artificial intelligencewill result in fundamental changesacross the business landscape withinthe next few years, according to tech-nology experts.

Powered by sophisticatedalgorithms and with the ability to learnover time, artificial intelligence (AI)will be increasingly used in everythingfrom accounting and legal analysistools to heavy machinery andautonomous cars.

According to predictions by researchfirm Gartner, AI technology will beembedded in virtually every new busi-ness software product by 2020. Thefirm says that, by that time, AI will havebecome a top-five investment priorityfor more than 30 per cent of all corpor-ate chief information officers.

‘‘I think it can fundamentally trans-form everything that we do which canbe characterised as a service,’’ saysNSW chief data scientist and chiefexecutive of the Data Analytics Centre,

Ian Oppermann. ‘‘I think it will be asfundamental as electricity. Electricityhas changed not only the way we seethe world, but also the way that we uselabour-saving devices and how wecommunicate.’’

Oppermann says the use cases forelectricity evolved from simply turningon lights to automating machinery, tothe point where it now underpinsalmost everything. He believes AI willfollow a similar path.

At a recent round-table discussion,co-hosted by The Australian FinancialReview and the Australian ComputerSociety, participants agreed AI startedout almost as a technical novelty, usedmostly in computer games, but is nowbeing rapidly adopted across a broadrange of areas.

According to a global survey by tech-nology company Avanade, 31 per centof organisations have already begunusing intelligent automation toimprove productivity, and this numberis predicted to double during the nextthree years. Of those who responded tothe survey, 86 per cent believe theymust deploy intelligent automation tobe a leader in their field.

‘‘I think the point about moving fromgaming into transformation of busi-ness is the tipping point that we’re atright now,’’ says Joanna Batstone, vice-

president and lab director for IBMResearch Australia.

‘‘We’ve been finding here in Austra-lia a lot of early-adopter projects withindustry, where we’re now looking athow we can use those technologies totransform a financial services com-pany’s credit risk modelling, or an oiland gas company’s insight into theirdata.

‘‘Business adoption of those techno-logies is already happening [but] I thinkthe consumer in the street is less awareof how much that AI technology isalready starting to impact business.’’

Peter Leonard, principal of Data Syn-ergies, says this awareness will changeas tools emerge that make interactingwith AI-driven applications easier.

‘‘I think consumers will pick up thosekinds of applications quite quickly and,as a result of that, they will increasinglyexpect AI-driven services to be avail-

able and will pay real value for thoseservices,’’ he says.

The round table participants point tothe accounting and legal sectors as twothat will experience significant disrup-tion from AI in the short to mediumterm. Many tasks that previously havebeen undertaken by junior staff willreadily be handled by AI.

For example, the task of trawlingthrough large quantities of case law todetermine the likelihood of a new casesucceeding could traditionally haveoccupied a team of young lawyers fordays or even weeks. Such a task couldbe completed by an IT tool in justminutes.

Similar efficiency benefits will alsobeen seen within accounting firmswhere many low-level, repetitive taskscan be handled by software rather thanhumans.

As the capabilities of the tools

increases, their usage will shift intoother areas of business.

Oppermann says further strides willbe made as AI allows the ongoingdevelopment of fully autonomousdevices, such as driverless cars andtrucks. These have the potential to rad-ically change the economics of supplychains and streamline transportation.

‘‘The ability to encapsulate the phys-ical world and the digital world to dofuture possible scenario planning ...will change everything which isinformation driven, everything whichis service focused and everythingwhich ultimately can translate as valuecreation,’’ he says. ‘‘I think we’re prob-ably, in terms of the technical capabil-ity of AI, still only in the fairly earlydays.’’

Michael Blumenstein, head ofschool, software at the Faculty ofEngineering and IT at UTS, points outthat AI has been in development sincethe 1940s when the first artificial neu-ron was created.

‘‘You’ve had this evolution anddecline of AI for many decades now,’’he says. ‘‘Some people, because this isthe first time they’ve actually beenexposed to it, think this is the first timeit’s ever flared up like this.’’

Blumenstein says AI developmenthad stalled in the past because it neverreally lived up to expectations.

‘‘So the challenge ... is to say whetherin this iteration it will live up to expect-ations,’’ he says.

‘‘Because we can talk about beatinggames, or we can talk about doing com-puter vision and that’s all great, but thepromise of general AI, which means anAI that can do many things well, notjust one thing well, is still out of ourreach.’’