Star Trek, here we come

Lo­cal startup is chang­ing way we in­ter­act with com­put­ers

Grand Magazine - - CON­TENTS - ALEX KIN­SELLA

Lo­cal startup is chang­ing the way we in­ter­act with com­put­ers

When my daugh­ter speaks to Siri on the fam­ily iPad, I can’t help but think about how what were just fu­tur­is­tic vi­sions in my youth are re­al­ity to­day. Grow­ing up in the 1980s, I could watch Michael Knight talk to KITT, his au­tonomous Pon­tiac Trans-Am, on a smart­watch on “Knight Rider.” The crew of the USS En­ter­prise spoke to the ship’s com­puter to re­search so­lu­tions to their cur­rent mis­sion on “Star Trek.”

To­day, I searched Google us­ing my voice to find out what year “Star Trek IV: The Voy­age Home” came out in theatres. It was 1986 – but Google reads IV as “eye vee” in­stead of four. Ro­man nu­mer­als are hard.

One of the more mem­o­rable scenes from that film is where the En­ter­prise’s chief en­gi­neer, Mr. Scott, tries to use a 1980s per­sonal com­puter at an en­gi­neer­ing com­pany. He first tries to speak to the com­puter as if it were from the 24th cen­tury. When the com­puter does noth­ing, cranky Dr. McCoy sug­gests us­ing the com­puter’s mouse. Mr. Scott picks up the mouse and tries speak­ing through it – to no avail. The frus­trated en­gi­neer­ing com­pany man­ager hastily sug­gests that Mr. Scott use the key­board, to which Mr. Scott replies, “How quaint.”

In­ter­fac­ing with our tech­nol­ogy has come a long way since the 1980s. We’re able to com­mu­ni­cate with our com­put­ers, cars, watches and smart­phones by speak­ing to them.

On our phones, Ap­ple’s Siri and Google As­sis­tant help us make res­tau­rant reser­va­tions or re­turn a text. At home, de­vices en­abled by Google As­sis­tant and Ama­zon Alexa are sim­pli­fy­ing our smart homes and mak­ing con­trol­ling ev­ery­thing eas­ier, from your home theatre and au­dio sys­tem to lights and ther­mostats.

These in­ter­ac­tions can seem al­most mag­i­cal. Mag­i­cal that is, un­til your car doesn’t un­der­stand what ad­dress you’re say­ing and you end up pulling over to the side of the road to man­u­ally en­ter the ad­dress into your phone’s GPS.

We’re not at the Star Trek level of voice in­ter­ac­tion yet, but lo­cal startup Malu­uba is work­ing to help us get there. Founded in 2011, Malu­uba was ac­quired in Jan­uary by Mi­crosoft.

Malu­uba builds ar­ti­fi­cially in­tel­li­gent sys­tems that un­der­stand hu­man lan­guage. What does that mean for the aver­age con­sumer? They’re de­vel­op­ing tech­nol­ogy to make com­put­ers lit­er­ate. Think back to grade school and all of those stan­dard­ized tests that we took. Those tests chal­lenged our abil­ity to un­der­stand the con­text be­hind what we were read­ing.

Malu­uba has worked on the kinds of in­ter­ac­tions that are found in Siri or Alexa. These “di­a­log sys­tems” are com­posed of a short se­quence of steps to help you ac­com­plish a task. Ama­zon Alexa is great at help­ing you find mu­sic you want to play or to or­der some­thing from Ama­zon. Ap­ple’s Siri al­lows you to in­ter­act hands-free with your iPhone or iPad to re­ply to mes­sages or set an alarm.

Malu­uba is build­ing a next gen­er­a­tion of di­a­log sys­tems that are more akin to in­ter­act­ing with a travel agent. You are able to com­pare dif­fer­ent pack­ages – even switch­ing back and forth be­tween dif­fer­ent op­tions – while the vir­tual agent ex­plains things to you.

“These are di­a­log sys­tems that help you achieve a goal even when you might not know what you want,” says Adam Trischle, se­nior re­search sci­en­tist at Malu­uba. “It’s up­ping the ante to more com­plex goals.”

Malu­uba works on get­ting ma­chines to un­der­stand lan­guage, but that doesn’t nec­es­sar­ily mean voice. Their work fo­cuses on un­der­stand­ing the text of a con­ver­sa­tion. This means their tech­nol­ogy can be used to make the chat­bot in­ter­ac­tions in Face­book, Slack or Kik bet­ter than what you see to­day.

“We re­ally like text,” says Trischle. “Talk­ing to your phone can be weird and awk­ward when you’re sur­rounded by peo­ple.”

The sys­tems that Malu­uba de­vel­ops can be used with voice or by typ­ing. As an ex­am­ple, the vir­tual agent might help you find some­thing you don’t know you’re look­ing for.

“You ask for a cheap pack­age for Hawaii and there’s noth­ing avail­able be­cause Hawaii is ex­pen­sive, so the sys­tem sug­gests Costa Rica is sim­i­lar and it meets your bud­get,” adds Trischle.

If you want to im­press your friends at a bar­be­cue this sum­mer, the two terms you need to know are ar­ti­fi­cial in­tel­li­gence and ma­chine learn­ing. They’re of­ten used in­ter­change­ably, but this is in­cor­rect.

Ar­ti­fi­cial in­tel­li­gence is a field that started in the 1960s. “The re­search into ar­ti­fi­cial in­tel­li­gence was very am­bi­tious and very

It’ll only be a mat­ter of time be­fore my daugh­ter will be able to tell Net­flix to play her favourite show and it will know ex­actly what to put on.

grandiose,” says Trischle. “They wanted to get com­put­ers to re­ally think in a hu­man sense.”

These re­searchers fo­cused on cre­at­ing gen­eral in­tel­li­gence that mim­icked hu­man be­ings. It was a mas­sive task and there were mul­ti­ple bar­ri­ers the re­searchers couldn’t over­come at the time. Ar­ti­fi­cial in­tel­li­gence re­search went into a kind of hi­ber­na­tion.

The re­searchers who stuck around took a new tack. Rather than try­ing to get the ma­chines to think like hu­mans, they fo­cused on build­ing sys­tems that could learn from data to ac­com­plish sim­ple, spe­cific tasks. This new field is called ma­chine learn­ing.

“We don’t care if they look or act like the brain, we just want to solve some real world, sim­ple prob­lems,” notes Trischle.

“Au­tonomous ve­hi­cles, voice in­ter­ac­tion ... those ad­vances have hap­pened be­cause the tasks are very con­strained,” adds Paul Gray, di­rec­tor of prod­uct mar­ket­ing at Malu­uba.

What’s driv­ing this trend? It’s a com­bi­na­tion of data and hard­ware. First, we have an al­most un­fath­omable amount of data avail­able to re­searchers. “We have al­most 20 years of In­ter­net data and peo­ple are con­nected all the time,” says Gray.

Sec­ond, com­put­ers – from lap­tops to smart­phones – are more pow­er­ful.

Go­ing back to those not-so mag­i­cal­in­ter­ac­tions, I wanted to know if it was some­thing I was do­ing wrong.

“Most of these cur­rent sys­tems have a prob­lem with ex­pec­ta­tion ver­sus re­al­ity,” says Trischle. “They’re good for the in­ter­ac­tions they’ve been built for – Alexa for mu­sic and Siri for travel.”

“This year will be the break­through for lan­guage, get­ting it to the level we’ve seen for vi­sion and speech,” adds Gray.

Imag­ine be­ing able to ask your vir­tual as­sis­tant to find flights for a trip to Brazil. When the re­sults are re­turned to you, you ask a sec­ond ques­tion about whether any vac­cines are re­quired. Your vir­tual as­sis­tant will un­der­stand the con­text of your ques­tion.

In­stead of re­turn­ing a few web­site sug­ges­tions as Siri would to­day, your vir­tual as­sis­tant will read the sites, de­ter­mine the an­swer and tell you what you need to know. All of this while not for­get­ing that you’re talk­ing about a trip to Brazil. The goal is that you for­get you’re talk­ing to a ma­chine rather than a hu­man be­ing.

It’ll only be a mat­ter of time be­fore my daugh­ter will be able to tell Net­flix to play her favourite show and it will know ex­actly what to put on. It will prob­a­bly be some­thing I don’t want to watch. The more things change, the more they stay the same. Bring on the magic.

Alex Kin­sella

Newspapers in English

Newspapers from Canada

© PressReader. All rights reserved.