AI’s re­search rut

When we think of AI as one par­tic­u­lar thing, we drag the whole field down

Jamaica Gleaner - - SOMETHING EXTRA -

MIT Tech­nol­ogy WHEN YOU pic­ture ar­ti­fi­cial in­tel­li­gence (AI), what do you see? A hu­manoid ro­bot? When you think about a real-world ap­pli­ca­tion of AI, what comes to mind? Prob­a­bly autonomous driv­ing. When you think about the tech­ni­cal de­tails of AI, what ap­proach do you name? I’m will­ing to bet it’s deep learn­ing.

In re­al­ity, AI comes in many shapes and forms. AI ma­chines go far be­yond hu­manoid ro­bots; they range from soft­ware de­tect­ing bul­ly­ing on so­cial me­dia to wear­able de­vices mon­i­tor­ing per­sonal health risk fac­tors to ro­botic arms learn­ing to feed paral­ysed peo­ple to autonomous ro­bots ex­plor­ing other plan­ets. The po­ten­tial ap­pli­ca­tions of AI are lim­it­less: per­son­alised ed­u­ca­tion, el­derly as­sis­tance, wildlife be­hav­iour anal­y­sis, med­i­cal-record mining, and much more.

Our fail­ure to ap­pre­ci­ate this spec­trum threat­ens to hold back the field. When we col­lec­tively pic­ture AI as one type of thing – whether it’s hu­manoid ro­bots or self-driv­ing cars or deep learn­ing – we’re en­cour­ag­ing the next gen­er­a­tion of re­searchers to be ex­cited ex­clu­sively about those nar­row things. If stu­dents are pre­sented with a ho­mo­ge­neous pool of AI re­search role mod­els, then it’s a self­ful­fill­ing prophecy that only stu­dents who ‘fit in’ will re­main in the field.

Since AI has en­tic­ingly broad pos­si­ble ap­pli­ca­tions, we need peo­ple with a com­pa­ra­bly broad set of ex­pe­ri­ences and world views work­ing on AI prob­lems. Wouldn’t re­search teams work­ing on AI med­i­cal ap­pli­ca­tions ben­e­fit from re­searchers trained in bi­ol­ogy? Wouldn’t teams work­ing on AI hunger re­lief ben­e­fit from re­searchers with first-hand ex­pe­ri­ence in poor coun­tries? Wouldn’t teams work­ing on AI as­sis­tive de­vices ben­e­fit from re­searchers with phys­i­cal dis­abil­i­ties?

To­day, there’s a lot of fas­ci­nat­ing work go­ing on in AI, but we’re also kind of in a rut. We’ve tended to breed the same style of re­searchers over and over again – peo­ple who come from sim­i­lar back­grounds, have sim­i­lar in­ter­ests, read the same books as kids, learn from the same thought lead­ers, and ul­ti­mately do the same kinds of re­search. Given that AI is such an all-en­com­pass­ing field, and a giant part of our fu­ture, we can’t af­ford to do that any­more.

Newspapers in English

Newspapers from Jamaica

© PressReader. All rights reserved.