WHAT AR­TI­FI­CIAL IN­TEL­LI­GENCE ISN’T ... AND IS

Australian Transport News - - Operations + Strategy | Artificial Intelligence -

Ar­ti­fi­cial in­tel­li­gence (AI) is not new. The term was coined in 1956 by John McCarthy, a Stan­ford com­puter sci­ence pro­fes­sor who or­gan­ised an aca­demic con­fer­ence on the topic at Dart­mouth Col­lege in the sum­mer of that year.

The field of AI has gone through a se­ries of boom-bust cy­cles since then, char­ac­terised by tech­no­log­i­cal break­throughs that stirred ac­tiv­ity and ex­cite­ment about the topic, fol­lowed by sub­se­quent pe­ri­ods of dis­il­lu­sion­ment and dis­in­ter­est known as ‘AI Win­ters’ as tech­ni­cal lim­i­ta­tions were dis­cov­ered.

Ar­ti­fi­cial in­tel­li­gence can be de­fined as hu­man in­tel­li­gence ex­hib­ited by machines; sys­tems that ap­prox­i­mate, mimic, repli­cate, au­to­mate, and even­tu­ally im­prove on hu­man think­ing.

Through­out the past half-cen­tury, a few key com­po­nents of AI were es­tab­lished as es­sen­tial: the abil­ity to per­ceive, un­der­stand, learn, prob­lem solve, and rea­son. Count­less work­ing def­i­ni­tions of AI have been pro­posed over the years but the uni­fy­ing thread in all of them is that com­put­ers with the right soft­ware can be used to solve the kind of prob­lems that hu­mans solve, in­ter­act with hu­mans and the world as hu­mans do, and cre­ate ideas like hu­mans.

In other words, while the mech­a­nisms that give rise to AI are ‘ar­ti­fi­cial’, the in­tel­li­gence to which AI is in­tended to ap­prox­i­mate is in­dis­tin­guish­able from hu­man in­tel­li­gence.

In the early days of the sci­ence, pro­cess­ing in­puts from the out­side world re­quired ex­ten­sive pro­gram­ming, which lim­ited early AI sys­tems to a very nar­row set of in­puts and con­di­tions.

How­ever, since then, com­puter sci­ence has worked to ad­vance the ca­pa­bil­ity of AI-en­abled com­put­ing sys­tems. Board games have long been a prov­ing ground for AI re­search, as they typ­i­cally in­volve a fi­nite num­ber of play­ers, rules, ob­jec­tives, and pos­si­ble moves. This es­sen­tially means that games – one by one, in­clud­ing check­ers, backgam­mon, and even Jeop­ardy! to name a few – have been taken over by AI.

Most fa­mously, in 1997, IBM’s Deep Blue de­feated Garry Kas­parov, the then reign­ing world cham­pion of chess. This tra­jec­tory per­sists with the an­cient Chi­nese game of Go, and the de­feat of reign­ing world cham­pion Lee Sedol by Deep­Mind’s Al­phaGo in March 2016.

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.