The Turing Test


PC, Xbox One

These tests, we’re told, are be­yond the ca­pac­ity of ma­chines; they can only be solved by a hu­man brain. Well, per­haps not this one. Oh, we get there in the end – it turns out that this isn’t an ex­am­ple of need­lessly finicky puz­zle de­sign so much as a re­minder to look up oc­ca­sion­ally – but it’s enough for us to con­cede that maybe we don’t rep­re­sent the ideal of in­tel­li­gent be­hav­iour to which Turing’s pa­per re­ferred.

Such stick­ing points are, thank­fully, rare, which says much for the el­e­gant es­ca­la­tion of chal­lenge and com­plex­ity in Bulk­head In­ter­ac­tive’s de­signs – not least since it’s all ac­com­plished with­out a word of as­sis­tance. These co­nun­drums are built around the trans­fer­ral of con­trol, as you use a gun-shaped En­ergy Ma­nip­u­la­tion Tool (EMT) to vac­uum up and launch en­ergy orbs from one power source to an­other.

At first, that means noth­ing more tax­ing than find­ing a door that’s held open and re­mov­ing its en­ergy source to power an­other that lies be­tween you and the exit. Then the num­ber of doors in­creases while your sup­ply of orbs shrinks. Some orbs re­side within cubes that must be trans­ported man­u­ally to their des­ti­na­tion. It’s only re­ally this phys­i­cal el­e­ment that would seem to dis­al­low ma­chines from tak­ing part; oth­er­wise these all ap­pear to be sin­gle-so­lu­tion logic prob­lems. The most ef­fi­cient op­tion tends to be the only one.

This cen­tral me­chanic fea­tures to the end, but as you pass from one area to the next, it’s steadily ac­com­pa­nied by more new el­e­ments. Pur­ple and green orbs lack the con­sis­tent power of their blue coun­ter­parts, flick­ing on and off – though that makes them ideal for ob­jects you don’t nec­es­sar­ily want to re­main in place. Reds, mean­while, lack stay­ing power, sup­ply­ing just enough en­ergy for you to cross a light bridge, per­haps, but run­ning out of juice within sec­onds of you reach­ing the other side. There are levers to move gi­ant mag­nets and re­po­si­tion walk­ways, pres­sure plates to weigh down, and beams of light con­trol­ling hy­draulic plat­forms whose path you may need to tem­po­rar­ily ob­struct. To this end and more, you’ll later be able to in­ter­face with a ro­bot ally and CCTV cam­eras – as long as you main­tain an un­bro­ken line of sight – both of which can trig­ger elec­tronic switches that can’t be man­u­ally ac­ti­vated.

But wait, weren’t these tests de­signed for hu­man brains? We ap­pear to have been mis­led. Then again, that’s kind of the point: EMT no longer seems such a func­tional name when you con­sider the wider mean­ing of that M. From the out­set it’s clear your mis­sion was never likely to be straight­for­ward. As as­tro­naut Ava Turing in­ves­ti­gates an ISA re­search fa­cil­ity on Jupiter’s moon of Europa, os­ten­si­bly to look for its miss­ing oc­cu­pants, the warn­ing signs are ob­vi­ous as soon as you touch down. Take, for ex­am­ple, your seem­ingly friendly AI guide, Tom. He is cour­te­ous and po­lite as he and Ava con­verse be­tween each puz­zle room, but a cou­ple of hours in you’re al­ready half ex­pect­ing him to break into Daisy, Daisy. In truth, the hints of a quiet, un­der­ly­ing malev­o­lence come a lit­tle too soon, but your un­ease grows as you lis­ten in to au­dio logs of ear­lier ex­changes be­tween Tom and the ISA crew. There’s some­thing si­mul­ta­ne­ously child­ish and chill­ing as Tom con­cludes an ar­gu­ment by say­ing, “You’re not bet­ter than me,” even though his ar­gu­ment is rooted in the logic of his pro­gram­ming. “It is not a threat,” he in­sists in an­other, in a way that deeply im­plies the op­po­site.

Even so, the story doesn’t quite go where you’re ex­pect­ing and the weighty philo­soph­i­cal, eth­i­cal and moral quan­daries that come into play as the real rea­son for your mis­sion crys­tallises are in­tel­li­gently han­dled. If oc­ca­sion­ally it feels as if you’re be­ing lec­tured by a stu­dent ma­jor­ing in thought ex­per­i­ments, and the di­a­logue is too on-the-nose, there are pas­sages that res­onate – notably an ob­ser­va­tion about the con­trol­ling in­flu­ence of so­cial stim­uli on our de­ci­sion-mak­ing. And while you’ll have an idea of the story’s des­ti­na­tion well be­fore it gets there, the writ­ers con­found ex­pec­ta­tions in the game’s cli­mac­tic mo­ments, lead­ing play­ers to as­sume they’ll be forced to make a fi­nal choice, be­fore re­plac­ing it with an­other. Its am­bi­gu­i­ties raise a num­ber of ques­tions, not least in invit­ing you to con­sider your own role. Who, ex­actly, is in con­trol here? There’s plenty to think about, then, but the more you do, the wider the cracks grow in its aus­tere façade. Tom, we’re told, can’t solve these puz­zles, be­cause he’s not per­mit­ted to think lat­er­ally. And yet rarely are you given the op­por­tu­nity to con­sider a co­nun­drum from a cre­ative new an­gle. Only once, in the later stages, are you called to re­act against the game’s con­di­tion­ing, to think about an ob­ject in a dif­fer­ent way to solve an out­wardly sim­plis­tic stumper. And on a foun­da­tional level, the fact ten of these rooms lie be­tween each area of the fa­cil­ity is one leap of logic that’s never re­solved: it’s like ne­go­ti­at­ing the most lu­di­crously ex­tended and con­vo­luted se­cu­rity pro­ce­dure ever de­vised.

That’s symp­to­matic of a wider dis­con­nect be­tween sys­tems and story. As a two-han­der in­volv­ing a test sub­ject and an enig­matic AI, it’s dif­fi­cult not to draw com­par­isons with Por­tal. Valve’s game found a way to em­bed its nar­ra­tive within the walls of its chal­lenges; here, con­ver­sa­tions be­tween Tom and Ava stop and only restart once a sec­tor is com­pleted, with a late-game bar­rage of au­dio logs to fill in the gaps. Longer, more elab­o­rate tests kill the pac­ing, too – just as the truth seems to be get­ting closer, you’re forced to spend 20 min­utes in a sin­gle room, solv­ing a slightly more com­plex vari­a­tion on the kind of puz­zle you’ve been tack­ling for sev­eral hours. There’s prom­ise in The Turing Test’s con­stituent parts, but con­sid­ered as a whole, it fails the im­i­ta­tion game.

He is cour­te­ous and po­lite, but a cou­ple of hours in you’re al­ready half ex­pect­ing him to break into Daisy, Daisy

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.