THINK

Some say AI is an ex­is­ten­tial threat. Oth­ers think it can only be a force for good.

HWM (Malaysia) - - CONTENTS - By Koh Wanzi

AI: Men­ace or no? Con­sol­i­dat­ing An­droid auto-up­dates

Why Win­dows 10 S is still not main­stream ready

Watch: The Mist

Ev­ery­one has some­thing to say about AI these days. “They’re go­ing to take away all our jobs!” pan­icked voices cry out stri­dently. “One day, they’re also go­ing to be­come smarter than peo­ple and kill us all!”

That sounds like the mad ram­blings of a con­spir­acy the­o­rist who watched one Ter­mi­na­tor movie too many, un­til you re­al­ize that Tesla and SpaceX CEO Elon Musk sits firmly in that camp. Musk isn’t tak­ing to the streets and yelling about a ro­bot up­ris­ing, but he has re­ferred to AI as hu­man­ity’s “big­gest ex­is­ten­tial threat”.

He’s also said that ef­forts to de­velop smarter in­tel­li­gences are akin to “sum­mon­ing the de­mon”, so there’s lit­tle doubt that he views AI as a very real threat.

More re­cently, Rus­sian pres­i­dent Vladimir Putin weighed in, say­ing point­edly that “Who­ever be­comes the leader in this sphere will be­come the ruler of the world.”

Pre­dictably, this prompted Musk to of­fer up an­other nugget of apoc­a­lyp­tic wis­dom, and he took to Twit­ter warn­ing that com­pe­ti­tion for AI su­pe­ri­or­ity at a na­tional level would be the most likely cause of a third World War. Tools, not ri­val en­ti­ties When the news cy­cle is pe­ri­od­i­cally dom­i­nated by head­lines of su­per­s­mart pro­grams like Al­phaGo beat­ing the world’s best play­ers at Go, sup­pos­edly a game that is dif­fi­cult for com­put­ers to grasp, it’s easy to get swept up in the nar­ra­tive that AI is quickly sur­pass­ing us.

In some re­spects, that is true. Al­phaGo is clearly a prodi­gious Go player. And even in a highly skilled pro­fes­sion like medicine, re­searchers are work­ing on AI that can help with di­ag­noses of con­di­tions like di­a­betic retinopa­thy, cus­tom­ize treat­ments, and make sense of pa­tient data.

The prob­lem with this line of rea­son­ing is how amor­phous the term AI has be­come. Self-driv­ing cars? Im­age recog­ni­tion in Google pho­tos?

Alexa and Siri? That’s all be­ing lumped to­gether un­der the same term.

But all these pro­grams still serve very nar­row pur­poses, and are a long way from what we’d con­sider gen­eral AI, which would be ca­pa­ble of per­form­ing things that hu­mans do and pos­sess abil­i­ties across mul­ti­ple do­mains.

AI is trend­ing heav­ily to­ward solv­ing spe­cific tasks, likely due to the dif­fi­culty of de­vel­op­ing an ar­ti­fi­cial gen­eral in­tel­li­gence, and the more im­me­di­ate util­ity of train­ing it to be su­per good at a nar­row spec­trum of things. At this point, AI are tools rather than bud­ding in­tel­li­gences.

We’re still a long way off from be­ing able to have an ac­tual con­ver­sa­tion with Siri, and de­spite that slightly creepy in­ci­dent where Face­book AI chat­bots de­vel­oped their own lan­guage that re­searchers could not un­der­stand, there are more press­ing con­cerns than ma­li­cious AI.

When the ma­chine is a bet­ter worker

The first thing that comes to mind is job ob­so­les­cence. To be clear, we’re talk­ing about a sce­nario decades into the fu­ture, but it’s still a more ur­gent is­sue than the hypothetical cre­ation of su­per-smart AI.

A 2017 pa­per pub­lished by re­searchers at Ox­ford and Yale, which sur­veyed AI re­searchers on how soon they ex­pected ma­chines to be bet­ter at all tasks than hu­mans, gives a 50-per­cent chance of this hap­pen­ing in just 45 years. And in 120 years, all jobs may be au­to­mated, ac­cord­ing to the pa­per.

Low-skilled jobs will be the first to go, so the most trou­bling prospect is how AI will po­ten­tially widen the al­ready yawn­ing gap be­tween the haves and haves-not.

Newspapers in English

Newspapers from Malaysia

© PressReader. All rights reserved.