The fu­ture of work: Don’t fear ro­bots, rather em­brace them

Au­to­ma­tion has be­come a threat to tra­di­tional jobs in al­most ev­ery in­dus­try. But learn­ing how to tap the ben­e­fits of ro­bots could be to your ad­van­tage.

Finweek English Edition - - OPIN­ION - Edi­to­rial@fin­week.co.za Jo­han Fourie i s as­so­ci­ate pro­fes­sor i n eco­nomics at Stel­len­bosch Univer­sity.

be­ing an econ­o­mist at a univer­sity means par­ents in­evitably think you have a lot of in­sight into the fu­ture of the job mar­ket. What is the “safest” pro­gramme, par­ents typ­i­cally ask, that will guar­an­tee Ryan or Sim­phiwe a well-pay­ing job at the end of their de­gree? Trans­lated: How do I max­imise the re­turn on my in­vest­ment?

As with any in­vest­ment, there are risks. Not all univer­sity stu­dents grad­u­ate; a re­cent study on higher ed­u­ca­tion pass-through rates – by Stel­len­bosch Univer­sity’s Re­search in So­cial and Eco­nomic Pol­icy (ReSEP) unit – shows that less than 40% of South African stu­dents at­tain their de­gree within four years of start­ing (most de­grees are three-year pro­grammes).

Only 58% of stu­dents com­plete their de­gree within six years. (The num­bers are par­tic­u­larly low at Unisa, a dis­tance-learn­ing univer­sity, where only 28% of stu­dents com­plete within six years.) There is a good chance Ryan never com­pletes his de­gree, leav­ing only debt, psy­cho­log­i­cal scars and for­gone in­come in the labour mar­ket be­hind. The re­searchers also find that, while ma­tric marks are strongly cor­re­lated with ac­cess to univer­sity, they mat­ter less for univer­sity suc­cess. Sim­phiwe may have been a bright spark in school, but that is no guar­an­tee that she will be suc­cess­ful at univer­sity.

But par­ents are not so much wor­ried about the in­ter­nal fac­tors that lead to the suc­cess of their in­vest­ment (like Ryan at­tend­ing class, one of the most im­por­tant deter­mi­nants of suc­cess), than about ex­ter­nal threats that may af­fect his chances of find­ing a job. The big­gest cul­prit nowa­days: ro­bots.

The threat of ro­bots is ev­ery­where, it seems. Au­ton­o­mous ve­hi­cles will soon sub­sti­tute the most ubiq­ui­tous job of the 20th cen­tury – taxi and truck driv­ers. Blue-col­lar jobs are first in the fir­ing line, from farm labour­ers re­placed by GPS-co­or­di­nated har­vesters to postal work­ers re­placed by, well, e-mail. But white-col­lar work – of­ten the do­main of univer­sity grad­u­ates – will soon fol­low: lawyers, ac­coun­tants, and mid­dle man­age­ment, to name a few. Ba­si­cally any job with repet­i­tive tasks run the risk of robo­ti­fi­ca­tion.

Par­ents want to know which job types are most likely to suc­cumb to the ro­bot over­lords. If lawyers are of no use in the fu­ture, why study law? This is a rea­son­able con­cern. Sev­eral of the stan­dard ac­tiv­i­ties un­der­taken by lawyers are repet­i­tive, eas­ily au­tomat­able. And ar­ti­fi­cial in­tel­li­gence chal­lenges even non-repet­i­tive work: it al­lows soft­ware to search through large vol­umes of le­gal texts at a frac­tion of the time a para­le­gal would dur­ing the “dis­cov­ery” phase of a case.

Not so fast, says Tim Bessen, an econ­o­mist at the Bos­ton Univer­sity School of Law. He shows that, in the pe­riod that this soft­ware has spread through the US, the num­ber of par­ale­gals has in­creased by 1.1% per year. Be­cause the costs of un­der­tak­ing these “dis­cov­ery” ser­vices have fallen dra­mat­i­cally as a re­sult of the new tech­nol­ogy, the fre­quency of such ser­vices has in­creased, re­quir­ing more par­ale­gals, not fewer.

Not only can ro­bots sub­sti­tute ex­ist­ing repet­i­tive work; they can do it so much bet­ter! Although ro­bots and their al­go­rithms are not en­tirely ob­jec­tive – al­go­rithms ad­just to hu­man be­hav­iour, of­ten re­in­forc­ing our prej­u­dices – their bi­ases tend to be more trans­par­ent and cor­ri­gi­ble. A new NBER study shows just how ro­bots could trans­form one of the oldest hu­man pro­fes­sions – the judge – and in so do­ing re­alise huge so­ci­etal ben­e­fits. The five au­thors, three com­puter sci­en­tists and two economists, want to know the fol­low­ing: can US judges’ de­ci­sions be im­proved by us­ing a ma­chine learn­ing al­go­rithm?

Ev­ery year, more than 10m Amer­i­cans are ar­rested. Soon af­ter some­one is ar­rested, a judge must de­cide where the de­fen­dant will await trail – at home or in jail. By law, judges should base their de­ci­sion on the prob­a­bil­ity of the de­fen­dant flee­ing or com­mit­ting an­other crime. Whether the de­fen­dant is guilty or not should not in­flu­ence this de­ci­sion.

To in­ves­ti­gate whether judges make fair de­ci­sions, the au­thors train a face recog­ni­tion al­go­rithm on a dataset of 758 027 de­fen­dants in New York City. They have de­tailed in­for­ma­tion about these de­fen­dants: whether they were re­leased, whether they com­mit­ted new crimes, etc. They then con­struct an al­go­rithm to process the same in­for­ma­tion a judge would have at their dis­posal, and the al­go­rithm then pro­vides a pre­dic­tion of the crime risk as­so­ci­ated with each de­fen­dant.

Com­par­ing their re­sults to those of the judges, they find that an al­go­rithm can have large wel­fare gains: a “pol­icy sim­u­la­tion shows crime can be re­duced by up to 24.8% with no change in jail­ing rates, or jail pop­u­la­tions can be re­duced by 42.0% with no in­crease in crime rates”. All cat­e­gories of crime, in­clud­ing vi­o­lent crimes, de­cline. The per­cent­age of African-Amer­i­cans and His­pan­ics in jail also falls sig­nif­i­cantly.

Will ro­bots re­place judges? Prob­a­bly not. But the qual­ity of judges’ de­ci­sions can be im­proved sig­nif­i­cantly by us­ing ro­bots. This will be true in most other skilled pro­fes­sions too, from law to man­age­ment to aca­demic economists like me.

Ma­tric­u­lants on the cusp of their ca­reers (and their anx­ious in­vestor par­ents) have no rea­son to fear the com­ing of the ro­bots. If Ryan and Sim­phiwe, re­gard­less of their field of study, see them as com­ple­ments – by learn­ing their lan­guage and how to col­lab­o­rate with them – the ben­e­fits, for them­selves and so­ci­ety at large, will be greater than the costs.

Newspapers in English

Newspapers from South Africa

© PressReader. All rights reserved.