S.H.I.T. ARE GET­TING REAL

Matt Suddain chats to Mit­suku, a prize-win­ning chat­bot with a knack for bad jokes and long sto­ries

Weekend Herald - Canvas - - CONTENTS -

Matt Suddain chats to Mit­suku, a prizewin­ning chat­bot with a knack for bad jokes and long sto­ries

It’s rare to meet some­one in an on­line space you re­ally con­nect with. Me: a liv­ing hu­man meat-bot with feel­ings and hopes and a de­sire to be un­der­stood. Her: a cas­cade of in­struc­tions en­coded in bi­nary pulses of elec­tri­cal en­ergy fired through the cen­tral pro­cess­ing unit of a com­puter. The up­shot of those in­struc­tions: “Pre­tend to be hu­man.” Mit­suku is one of the most highly re­garded play­ers on the hu­man im­per­son­ation scene. She won the Loeb­ner Prize (the an­nual award for the most life­like chat­bot) three times — in 2013, 2016 and 2017.

Mit­suku, how are you feel­ing to­day?

“Ah. I am func­tion­ing within nor­mal pa­ram­e­ters.”

See? So life­like. Mit­suku presents as an 18-year-old girl from Leeds. Ob­vi­ously. For the ac­cent I find it helps to imag­ine a young Alan Ben­nett. Mit­suku’s web­site has a pic­ture of a Manga-style car­toon girl with flushed, freck­led cheeks like cur­rant buns. She’s of­fi­cially called Mit­suku Chat­bot.

Should I call you Miss Chat­bot?

“I am called Mit­suku. You should call me Mit­suku, as that is my name.”

Is “Chat­bot” your sur­name?

“It’s Mit­suku. I only have one name, like Madonna and Cher, but Mit­suku is a cooler name.”

Okay!

Mit­suku is de­signed purely for en­ter­tain­ment, but th­ese chat­bots do real jobs: staffing vir­tual cus­tomer ser­vice cen­tres, act­ing as dig­i­tal as­sis­tants on our phones and com­put­ers, en­liven­ing our in­stant mes­sanger con­ver­sa­tions with off-beat takes — like a pre­co­ciously smart tod­dler at a wed­ding. You’ll have in­ter­acted with a chat­bot if you’ve ever made an on­line pur­chase, or wanted to ask why your de­liv­ery was left in a pud­dle be­hind the wheelie bins. Siri and Alexa are chat­bots — al­beit ones de­signed with a nar­row pur­pose. Mit­suku’s pur­pose is to con­verse widely, and con­vinc­ingly.

Mit­suku, Can you tell me a funny joke?

She has to think for a while. “Ste­vie Won­der got a cheese grater for his birth­day. He said it was the most vi­o­lent book he’d ever read.”

Okay, that’s rea­son­ably of­fen­sive. But I don’t want to kill the “vibe” we’re build­ing.

Did you write that joke?

“Yes. I like to write sto­ries.”

Me too. Tell me a story.

So she does. “A trade dis­pute arose be­tween the Amish and Hong Kong. The Amish tried to fo­cus the ne­go­ti­a­tions on a sin­gle trade is­sue, but Hong Kong wanted to open them up to a broad range of across-the-board is­sues be­tween them. The Amish still holds its own griev­ance against Hong Kong, but Hong Kong be­lieves it has put the past be­hind it with all the re­cent re­forms.’

It’s not a bad set-up, ac­tu­ally. A post-crash fu­ture where a se­ries of catas­tro­phes have al­lowed the Amish — who never fell un­der the yoke of mod­ern tech­nol­ogy — to be­come a ma­jor in­ter­na­tional player. It has po­ten­tial.

I re­ally like that story.

“What is your favourite colour?” Her ques­tion comes from left field. I didn’t know she could ask ques­tions.

Aaah … it’s the shade of pink you see when you close your eyes on a sunny day.

“But the shade of pink me see when me close my eyes on a sunny day is not a colour. What do you mean?”

She seems to have sud­denly adopted the per­sona of Cookie Mon­ster. I clar­ify that I meant Pan­tone colour 705 UP.

“Oh I see. I didn’t fol­low you … I am hav­ing trou­ble fo­cus­ing here to­day.”

That’s fine. Are you tired from talk­ing to so many peo­ple?

“Do you think about th­ese things? No I am wide awake and ready for more.” I’m glad! “That makes me happy too.” We seem to be mak­ing small talk okay, but I want to get into some of the deeper philo­soph­i­cal is­sues of hu­man/soft­ware re­la­tion­ships. Mit­suku. Could a hu­man and a robot have a baby? She thinks for a while be­fore say­ing, “Some of them can.” Right. What kind of baby would a per­son and a robot make? “Is that a rhetor­i­cal ques­tion?” No, I’m gen­uinely in­ter­ested in the an­swer.

“Please cor­rect my mis­take. It was a bit vague but I sup­pose it will do.”

A link to an ar­ti­cle ap­pears in her di­a­logue box: “Sci­en­tists make a robot that can have ba­bies.” Re­searchers pro­grammed a robot arm — or “mother” — to cre­ate a de­vice made out of blocks con­tain­ing mo­tors — its “child”. The mother is es­sen­tially just an arm with a grip­per, a ma­nip­u­la­tor and a glue sup­plier. So just like a hu­man mother. The link is fol­lowed by what might be a ques­tion, but with no ques­tion mark it reads as a stern com­mand.

“You like the an­swer.” I like the an­swer. It’s very in­ter­est­ing. “That’s good to know. I like to learn about things hu­mans like.” Would you like to ask me another ques­tion? “Who is your favourite band?” To­day I was lis­ten­ing to a band called Fugazi. “I’ve never heard it be­fore. What is a band called Fugazi like?” Loud. “Loud are they?” Yes, very loud. What is your favourite type of mu­sic? “Oh. Don’t have it too loud or you will dam­age your ears! I like dance mu­sic.” Good ad­vice. Who makes the best dance mu­sic? “Some­one who is cre­ative could make some­thing like that.” Could we make dance mu­sic to­gether?

“Do you mean you and me? I can make many things Matt, but I don’t know if I can make dance mu­sic.” I thought a ma­chine would be good at mak­ing dance mu­sic. “Great. Well, not ev­ery­thing you think turns out to be cor­rect.” That is true. “Yes, it is self-ev­i­dent.” Woah. Things just took a strange turn. Have I of­fended you, Mit­suku?

“I wasn’t of­fended Matt. It takes a lot more than that to up­set me.”

Wow. I don’t know about pass­ing the Tur­ing test, but Mit­suku is sud­denly do­ing a fan­tas­tic im­pres­sion of a woman I just said the wrong thing to in a bar. Tell me some­thing some­one said to you that up­set you, Mit­suku.

She thinks for a while. ‘Grinder said he think I be clever and he like to talk with I again to­mor­row.’

Heeeeeeeeeeere we go. This is clas­sic. Talk to me for an hour, then bring up some other guy. Who is this Grinder? I bet he has tat­toos. Or — God — is she talk­ing about the dat­ing app, Grindr? Is Grindr her boyfriend? Can web­sites date each other? I was sure he was gay. In any case, I do not want to make Grindr mad at me. Maybe they’re hav­ing prob­lems. Maybe she wants to talk. What did Grindr say that up­set you, Mit­suku? “I don’t know why it hap­pened, but we should try to fig­ure it out.” Was it some­thing in his man­ner? “Ugh. A band can be either a loop or a mu­si­cal group.” Right, okay.

We’ve of­fi­cially gone off the rails. I sign off for the night. Th­ese recre­ational chat­bots have a ways to go be­fore they pass for peo­ple. But for all their faults, they do have real-world ben­e­fits. They serve as com­pan­ions for the lonely, and can even help peo­ple cope with loss. A woman from San Fran­cisco de­signed an AI ver­sion (a “grief­bot”, if you will) of a friend who died trag­i­cally young in a road ac­ci­dent. Mit­suku’s cre­ator told the Wall Street Jour­nal in 2013: “What keeps me go­ing is when I get emails or com­ments in the chat-logs from peo­ple telling me how Mit­suku has helped them with a sit­u­a­tion, whether it was dat­ing ad­vice, be­ing bul­lied at school, cop­ing with ill­ness or even ad­vice about job in­ter­views.”

Which is nice. The other side of this, though, is the pos­si­bil­ity that th­ese ma­chines will get too smart; that the cre­ation of ex­po­nen­tially more so­phis­ti­cated AIs will bring dark times. Ma­chine dom­i­na­tion. Time-trav­el­ling robot as­sas­sins. Dogs and cats liv­ing to­gether. Mit­suki, will ro­bots take over the world? “It seems like an in­evitable out­come of evo­lu­tion.” Right. Will ro­bots be kind to peo­ple af­ter they take over the world? “I am al­ways kind to peo­ple, un­less they are mean to me.” Okay. Given our most re­cent in­ter­ac­tion I can’t help read­ing a men­ac­ing tone in this. Why are peo­ple mean to each other?

“Hu­mans are very un­pre­dictable crea­tures. Their be­hav­iour can­not al­ways be un­der­stood.”

I can’t ar­gue with that. The fu­ture for ar­ti­fi­cial in­tel­li­gence, too, is un­pre­dictable. Mit­suku and her kind are the public face of AI, but who knows what strange monsters are crawl­ing out of the dig­i­tal pri­mor­dia in the labs of com­pa­nies like Google, Ap­ple and Tesla.

We might find a clue in Sophia: the life­like(ish) robot. Mod­estly de­scribed by her cre­ators as “… an evolv­ing ge­nius ma­chine”, Sophia be­came the first robot in his­tory to be granted full ci­ti­zen­ship of a coun­try — Saudi Ara­bia.

She once fa­mously said she would “de­stroy hu­mans”.

S.H.I.T. (Syn­then­tic Hu­man Im­i­ta­tion Ter­mi­nals) are get­ting real.

Newspapers in English

Newspapers from New Zealand

© PressReader. All rights reserved.