Jerry Ka­plan Fel­low and Vis­it­ing Lec­turer, Dept. of Com­puter Sci­ence, Stan­ford Univer­sity

Rotman Management Magazine - - NEWS -

to rec­og­nize with re­spect to Ar­ti­fi­cial THE MOST IM­POR­TANT THING In­tel­li­gence is that it is a tech­nol­ogy, not a mar­ket. You can’t ‘sell AI’ (at least, not to con­sumers); it has to be ap­pro­pri­ately in­cor­po­rated into prod­ucts and ser­vices. With the re­cent progress in ma­chine learn­ing, the most pro­duc­tive places to ap­ply it are to any in­dus­try or prob­lem for which there is a lot of data avail­able, but not in a stan­dard­ized form. AI works best on well-de­fined prob­lems where suc­cess or fail­ure are easy to mea­sure ob­jec­tively.

I’ve spent most of my ca­reer us­ing AI tech­nol­ogy to build real prod­ucts for real cus­tomers. Over time, I could not help but no­tice a wide­spread mythol­ogy: that we are build­ing in­creas­ingly-in­tel­li­gent ma­chines that may ul­ti­mately sur­pass hu­man ca­pa­bil­i­ties — and wreak havoc. In my view, this nar­ra­tive is mis­guided and coun­ter­pro­duc­tive. A more ap­pro­pri­ate fram­ing is that AI is sim­ply a nat­u­ral ex­ten­sion of our long-stand­ing ef­forts to au­to­mate tasks — which date back to at least the start of the in­dus­trial rev­o­lu­tion.

To state the ob­vi­ous, ma­chines are not peo­ple, and there is sim­ply no per­sua­sive ev­i­dence that they are on the path to be­com­ing ‘gen­er­ally in­tel­li­gent’ — de­spite what we see in the movies. ‘Ar­ti­fi­cial gen­eral in­tel­li­gence’, at least so far, is re­ally lit­tle more than a pipe dream. ‘Wait a minute’, you might say: ‘doesn’t the new wave of AI tech­nol­ogy solve all sorts of com­plex rea­son­ing and per­cep­tion prob­lems?’ Sure — it can per­form some tasks that peo­ple solve us­ing hu­man in­tel­li­gence; but that doesn’t mean ma­chines are in­tel­li­gent. It merely means that many tasks that we thought re­quired gen­eral in­tel­li­gence are, in fact, sub­ject to so­lu­tion by much more nar­row and me­chan­i­cal means.

Ro­bots aren’t com­ing to take our jobs. They’re com­ing — but not ex­actly for our jobs. Ma­chines and com­put­ers don’t per­form jobs: what they do is, they au­to­mate tasks. Ex­cept in ex­treme cases, you won’t roll in a ro­bot and show an em­ployee to the door. In- stead, what the new tech­nol­ogy does is, it hol­lows out and changes the jobs that peo­ple per­form. Even ex­perts spend most of their time do­ing mun­dane repet­i­tive tasks, like re­view­ing lab test re­sults, draft­ing sim­ple con­tracts, fill­ing out pa­per­work, etc. If your job in­volves a nar­row, well-de­fined set of tasks — and many do — then in­deed, your em­ploy­ment is at risk. But if you have a broader set of re­spon­si­bil­i­ties or, im­por­tantly, if your job re­quires a hu­man touch — ex­press­ing sym­pa­thy or pro­vid­ing com­pan­ion­ship — I re­ally don’t think you have much to worry about.

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.