Ed­u­ca­tion key to en­sur­ing equal­ity in fu­ture work

Irish Independent - Business Week - - TECHNOLOGY - Jean Tirole

WHAT’S the fu­ture of work? Will gigs re­place salaried em­ploy­ment, and will ro­bots even­tu­ally leave hu­mans with noth­ing to do? I see rea­son for scep­ti­cism, but also for con­cern.

Tech­nol­ogy, of course, is al­ready mak­ing in­de­pen­dent work eas­ier. It puts work­ers into con­tact with cus­tomers and helps them run a back of­fice. More im­por­tantly, it al­lows in­di­vid­u­als to build and pro­mote their rep­u­ta­tions at low cost.

Cus­tomers used to rely on a taxi com­pany’s rep­u­ta­tion, or choose a wash­ing ma­chine by the man­u­fac­turer’s brand. Now, each worker has a brand: on Uber, cus­tomers can re­ject driv­ers based on their per­sonal rat­ings. A firm’s col­lec­tive rep­u­ta­tion, with the con­comi­tant con­trol of its em­ploy­ees’ be­hav­iour, is be­com­ing less im­por­tant.

That said, tech­nol­ogy can also favour stan­dard salaried em­ploy­ment. The econ­o­mists Ge­orge Baker and Thomas Hub­bard, for ex­am­ple, have noted how on-board com­put­ers could change truck­ing. By mon­i­tor­ing be­hav­iour, they would solve a mo­ral haz­ard prob­lem: driv­ers have lit­tle digi­tised, call cen­tres use soft­ware to shorten the length of con­ver­sa­tions be­tween cus­tomer and em­ployee, or re­place hu­mans with bots.

These changes have global reper­cus­sions. They threaten the low-salary, out­sourced jobs that emerg­ing and un­der­de­vel­oped coun­tries have counted on to es­cape poverty. In de­vel­oped coun­tries, as the econ­o­mist David Au­tor and his coau­thors have demon­strated, they tend to ben­e­fit those em­ploy­ees whose skills com­ple­ment the new dig­i­tal tools. This ‘hol­lows out’ the dis­tri­bu­tion of jobs into ei­ther high-pay­ing skilled po­si­tions or low-pay­ing ba­sic ser­vice po­si­tions. In the US, the dif­fer­ence in salary for those who hold univer­sity de­grees has grown enor­mously in the past 30 years.

It’s still not clear, how­ever, which hu­man tasks com­put­ers will be able to re­place, and what the ef­fects will be. De­duc­tive prob­lems, in which the par­tic­u­lar is de­duced from the gen­eral rule in a log­i­cal way, are the eas­i­est.

An ATM ver­i­fies a card num­ber, the PIN code, and the bank ac­count bal­ance be­fore is­su­ing money and deb­it­ing the ac­count. None­the­less, to­tal em­ploy­ment in bank­ing rose even as the ATM net­work spread, be­cause de­mand grew and teller jobs were re­placed by new tasks.

Com­put­ers have also made great ad­vances in in­duc­tion, which starts with spe­cific facts and works to­ward a gen­eral law. Sim­i­lar tech­niques are en­abling au­to­mated fa­cial and voice recog­ni­tion, di­ag­no­sis, and other tasks that pre­vi­ously only hu­mans could per­form.

The most dif­fi­cult tasks for com­put­ers in­volve un­fore­seen prob­lems that do not match any pro­grammed rou­tine. Rare events can­not be an­a­lysed in­duc­tively to gen­er­ate an em­pir­i­cal law.

Frank Levy and Richard Mur­nane of­fer the ex­am­ple of a driver­less car that sees a ball pass in front of it. This ball poses no danger to the car, which there­fore has no rea­son to slam on the brakes.

A hu­man be­ing, on the other hand, will prob­a­bly fore­see that the ball may be fol­lowed by a young child, and will there­fore have a dif­fer­ent re­ac­tion. It il­lus­trates the ob­sta­cles com­put­ers still en­counter.

So hu­mans and com­put­ers face dif­fer­ent chal­lenges.

Com­put­ers are much faster and more re­li­able when pro­cess­ing log­i­cal and pre­dictable tasks. Thanks to ma­chine learn­ing, they can in­creas­ingly cope with un­fore­seen sit­u­a­tions, pro­vided they have enough data to recog­nise the struc­ture of the prob­lem. On the other hand, the hu­man brain is more flex­i­ble: A fiveyear- old child can han­dle some prob­lems bet­ter than any com­puter. So the peo­ple best-equipped to suc­ceed in the new world will be those who have ac­quired ab­stract knowl­edge that helps them adapt to their en­vi­ron­ment, while those with only sim­ple knowl­edge pre­par­ing them for rou­tine tasks are most in danger of be­ing re­placed.

This is why ed­u­ca­tion is cru­cial. If we don’t have a sys­tem that gives ev­ery­one a chance to gain the nec­es­sary skills, dif­fer­ences in ed­u­ca­tion and fam­ily back­ground will lead to even greater in­equal­ity. (Bloomberg View)

THIS is the fifth ar­ti­cle in my weekly startup di­ary. I’m writ­ing down my busi­ness de­ci­sions as I make them, and then writ­ing about how things play out.

At this point I feel we should re­view some of the de­ci­sions so far, and see what we can learn.

The first big de­ci­sion was to write this col­umn. The most im­me­di­ate pos­i­tive ef­fect has been an in­creased num­ber of for­mer con­tacts reach­ing out.

Some of those con­tacts are in­ter­ested in build­ing deeper busi­ness re­la­tion­ships. This is bet­ter than just mak­ing new con­tacts – old friends re­ally are the best.

On the other hand, since I have de­clared a pol­icy of open­ness about my startup, I now need to be clear with ev­ery­one about what is go­ing to be printed in this col­umn.

That’s a com­pli­ca­tion I did not fore­see, which I guess is pretty naïve in hind­sight.

My rule is that con­ver­sa­tions are pri­vate by de­fault, and I’ll ask per­mis­sion be­fore writ­ing about any­thing I dis­cuss. A great deal of busi­ness does de­pend on cus­tomer and sup­plier con­fi­den­tial­ity, even if you’re pre­pared to share your own num­bers.

The sec­ond big de­ci­sion was to start a news­let­ter for a seg­ment of my in­tended cus­tomer base. The news­let­ter has been go­ing for seven weeks, and now has 133 sub­scribers, and a 32pc open rate (the pro­por­tion of sub­scribers that open and read the news­let­ter when it ar­rives in their In­box).

Last week I had 119 sub­scribers. I also wrote last week that I would be try­ing a num­ber of dif­fer­ent pro­mo­tion tac­tics. In the end, I only tried a sin­gle tac­tic, the red­dit.com ads. I re­alised that try­ing mul­ti­ple tac­tics at the same time would con­flate re­sults

To avoid this I re­frained from any other pro­mo­tion, apart from a few reg­u­lar tweets.

Sadly, the red­dit.com ads have not been ef­fec­tive. I’ve spent about $150 US Dol­lars for an ex­tra 14 sub­scribers – yikes! My click-through rate (the pro­por­tion of clicks to views) is 0.03pc, which is pretty mis­er­able. I’m go­ing to de­clare this a fail­ure. Now I could start at­tempt­ing to re­fine the ad copy and tar­get­ing, but I would need a 500pc im­prove­ment to get to my tar­get of 500 sub­scribers on Jan­uary 1, 2018. That’s un­fea­si­ble.

I’m go­ing to fo­cus on per­son­alised emails for next week – again just us­ing a sin­gle tac­tic to avoid con­fla­tion.

The third de­ci­sion was com­ple­men­tary to the sec­ond: not to build a soft­ware pro­to­type, at first.

In­stead I have fo­cused on talk­ing to as many peo­ple as pos­si­ble in the in­dus­try, and re­fin­ing my pitch.

I’m still in the process of do­ing this, so have not col­lected the in­for­ma­tion into a struc­tured form. I hereby de­cide that the time is now ripe to do so, and will re­port back on my cus­tomer dis­cov­ery ef­forts in fu­ture col­umns.

A soft­ware pro­to­type will have to be built. There has to be a real prod­uct to make this busi­ness work.

My plan has been to spend the first three months of the busi­ness val­i­dat­ing the idea and work­ing out the busi­ness model.

At this point we are about 10 weeks in (you have only been ex­posed to the last five weeks), so soft­ware de­vel­op­ment is due shortly. We’ll leave all that to later col­umns too.

A fourth de­ci­sion that I have not dis­cussed yet is the de­ci­sion to set up a lit­tle event man­age­ment agency.

Yes, you read that right. We are also in the busi­ness of run­ning tech­nol­ogy meetup groups, and other events for tech­ni­cal types.

We run the Dublin Mi­croser­vices meetup, and will soon be help­ing to launch the Cork Ar­ti­fi­cial In­tel­li­gence meetup. Tech mee­tups are small gath­er­ings of be­tween 50 to 100 tech en­thu­si­asts who meet once a month in the evening to hear two or three speak­ers give pre­sen­ta­tions about the topic of the group.

You can find all the mee­tups on meetup.com. Mee­tups are a fan­tas­tic way to con­nect with your com­mu­nity and I could not rec­om­mend them highly enough.

So what busi­ness does a Soft­ware-as-a-Ser­vice startup have run­ning events? Be­cause we need to “eat our own dog food”. Eat­ing your own dog food means us­ing your own prod­uct to solve the same prob­lems your cus­tomers have.

It means ex­pe­ri­enc­ing the same grow­ing pains with bugs, poor per­for­mance, and lack of fea­tures. It is one of the most pow­er­ful strate­gies you can adopt. You know your cus­tomer in­ti­mately, be­cause you are your own cus­tomer.

I am a con­fer­ence speaker,

Newspapers in English

Newspapers from Ireland

© PressReader. All rights reserved.