The jobs that will be lost to com­put­ers

The Witness - - INSIGHT - GWYNNE DYER • Gwynne Dyer is an in­de­pen­dent jour­nal­ist whose ar­ti­cles are pub­lished in 145 news­pa­pers.

DON’T bother ask­ing if jobs are be­ing lost to com­put­ers. Of course they are, and the cur­rent wave of pop­ulist po­lit­i­cal re­volts in West­ern coun­tries is what Lud­dism looks like in an era of in­dus­tri­alised democ­ra­cies. The right ques­tion to ask is: what kinds of jobs are be­ing lost? Mo­ravec’s Para­dox pre­dicted the an­swer al­most 30 years ago.

Right now, it’s the jobs in the mid­dle that are at risk of dis­ap­pear­ing. Not high­level pro­fes­sional and man­age­rial jobs that re­quire so­phis­ti­cated so­cial and in­tel­lec­tual skills and pay very well. Not poorly paid jobs in de­liv­ery or the fast­food in­dus­try ei­ther, although au­to­ma­tion will even­tu­ally take jobs in the ser­vice in­dus­tries too.

But the mid­dle­in­come, semi­skilled jobs, mostly in man­u­fac­tur­ing or trans­porta­tion, that used to sus­tain a broad and pros­per­ous mid­dle class, are dwin­dling fast. West­ern so­ci­eties are be­ing hol­lowed out by au­to­ma­tion, just as Mo­ravec’s Para­dox pre­dicts. Of­ten the newly un­em­ployed find other work, but it is gen­er­ally in the low­in­come ser­vice sec­tor. These dis­in­her­ited lower­mid­dle­class and up­per­work­ing­class peo­ple are the foot sol­diers of the pop­ulist rev­o­lu­tions.

Back in the eight­ies, Hans Mo­ravec, a pi­o­neer re­searcher in ar­ti­fi­cial in­tel­li­gence (AI), made the key ob­ser­va­tion that “it is com­par­a­tively easy to make com­put­ers ex­hibit adult­level per­for­mance on in­tel­li­gence tests or play­ing check­ers, and dif­fi­cult or im­pos­si­ble to give them the skills of a one­year­old when it comes to per­cep­tion and mo­bil­ity”.

The para­dox is that ac­tiv­i­ties like high­level rea­son­ing that are chal­leng­ing for hu­man be­ings are easy for ro­bots en­dowed with AI. Sim­ple sen­sory and mo­tor skills that are easy for the av­er­age one­year­old child, on the other hand, are far beyond the cur­rent reach of the ro­bots. No sur­prise, re­ally: those skills in hu­man be­ings are the prod­ucts of a bil­lion years of evo­lu­tion, and in­deed are largely un­con­scious in us.

So the jobs that ro­bots can most eas­ily take are mid­level man­age­ment jobs and semi­skilled, highly repet­i­tive man­ual jobs — and there goes the mid­dle­class meat in the sand­wich. What’s left is a small group of rich peo­ple (who own the ro­bots), an im­pov­er­ished mass of peo­ple who pro­vide them with ser­vices of ev­ery kind or have no jobs at all, and a level of re­sent­ment in the lat­ter that is rocket fuel for a pop­ulist rev­o­lu­tion.

This dystopian vi­sion is com­mon­place nowa­days, pushed to the top of the agenda by Brexit in the UK, the elec­tion of Don­ald Trump in the U.S., and neo­fas­cist elec­tion suc­cesses (though not yet vic­to­ries) in the Nether­lands, France and Ger­many. The same phe­nom­e­non may well play a big part in Italy’s elec­tion next month.

And the ro­bots are soon go­ing to be able to take out the rest of the jobs too. Mo­ravec and his col­leagues were work­ing with the com­put­ers of 30 years ago, which were re­ally sim­ple­minded and sin­gle­minded. To­day’s and to­mor­row’s AI is run­ning on com­put­ers that are or­ders of mag­ni­tude more pow­er­ful, and that al­low them to do dif­fer­ent things, like “deep learn­ing”, for ex­am­ple.

The op­er­at­ing in­struc­tions don’t only come from the top (hu­man be­ings) any more. More and more of­ten, the AI is told what the re­sult should be, and works out how to get there for it­self by “deep learn­ing”, a trial­and­er­ror process that only be­comes fea­si­ble when you have a num­ber­crunch­ing ca­pa­bil­ity mag­ni­tudes much greater than be­fore.

Then the path opens to (among other things) ar­ti­fi­cial in­tel­li­gence that has hu­man­level sen­sory and mo­tor skills. Not right away, of course, but in due course.

There go the rest of the jobs, you might think, and cer­tainly a lot will go. There goes the need for hu­man be­ings al­to­gether, the more pes­simistic will think, and maybe that’s true too. But the lat­ter out­come is still a choice, not an in­evitabil­ity.

A sig­nif­i­cant num­ber of AI spe­cial­ists are now work­ing on what they call “ar­ti­fi­cial gen­eral in­tel­li­gence”: AGI. Rather than teach a ma­chine to use sym­bolic logic to an­swer spe­cific kinds of ques­tions, they are build­ing ar­ti­fi­cial neu­ral net­works and ma­chine­learn­ing mod­ules loosely mod­elled on the hu­man brain.

Horoshi Ya­makawa, a Ja­pan­based leader in AGI, sees two ad­van­tages to this ap­proach. “The first is that since we are cre­at­ing AI that re­sem­bles the hu­man brain, we can de­velop AGI with an affin­ity for hu­mans. Sim­ply put, I think it will be eas­ier to cre­ate an AI with the same be­hav­iour and sense of val­ues as hu­mans this way.

“Even if su­per­in­tel­li­gence ex­ceeds hu­man in­tel­li­gence in the near fu­ture, it will be com­par­a­tively easy to com­mu­ni­cate with AI de­signed to think like a hu­man, and this will be use­ful as ma­chines and hu­mans con­tinue to live and in­ter­act with each other ...”

Feel­ing re­as­sured now? Thought not. There’s never much re­as­sur­ance to be had when think­ing about the fu­ture.

Most of the jobs are go­ing to go, sooner or later, in­clud­ing the skilled man­ual jobs and the high­level man­age­ment jobs that now seem safe.

We’ll have to get used to that, just like our re­cent an­ces­tors had to get used to work­ing in cities not on farms.

But maybe the ro­bots will grow up to be our col­leagues, not our over­lords or our suc­ces­sors. If we take the trou­ble to de­sign them that way, start­ing now.

Newspapers in English

Newspapers from South Africa

© PressReader. All rights reserved.