Transcript: Teaching Machines to be Funny | Mar 07, 2018

Steve sits in the studio. He's slim, clean-shaven, in his fifties, with short curly brown hair. He's wearing a gray suit, white shirt, and spotted gray tie.

A caption on screen reads "Teaching machines to be funny."

Steve says CAN YOU TEACH A
MACHINE TO BE FUNNY?
THAT'S THE KIND OF UNLIKELY
QUESTION THAT RAPID ADVANCES IN
TECHNOLOGY HAVE NOW PUT ON THE
TABLE.
KORY MATHEWSON IS A PhD STUDENT
AT THE UNIVERSITY OF ALBERTA
WORKING ON INTERACTIVE MACHINE
LEARNING.
HE'S ALSO AN IMPROV COMEDIAN WHO
WORKS WITH ROBOTS.
SO IF ANYONE CAN EXPLAIN WHERE
THIS IS ALL HEADED, HE'S A GOOD
BET, AND HE JOINS US NOW ON THE
LINE FROM EDMONTON, ALBERTA.

Kory is in his thirties, clean-shaven, with chin-length dark blond hair. He's wearing a red plaid shirt.

Steve continues KORY, IT'S GOOD TO HAVE YOU ON TVO.
HOW ARE YOU DOING?

Kory says I'M DOING GREAT.
IT'S GREAT TO BE HERE.

Steve says WHERE DID THE NOTION
COME FROM TO COMBINE YOUR
COMEDIC SKILLS WITH IMPROV?

The caption changes to "Kory Mathewson. University of Alberta."
Then, it changes again to "Tech talk."

Kory says A GOOD QUESTION.
I STARTED DOING IMPROV YEARS AGO
AND I STARTED WITH THE AUDIENCE.
THE SHOW IS CALLED "FIFTEEN
MINUTES OF FAME."
I WAS TRYING TO FIND THE PERSON
IN THE AUDIENCE WHO KNEW IMPROV
THE LEAST AND I WOULD GET THEM
UP AND MAKE THEM THE STAR OF THE
SHOW FOR A BRIEF AMOUNT OF TIME.
IT WAS SO MUCH FUN AND I HAD SO
MUCH FUN PRETENDING WITH THESE
PEOPLE WHO OTHERWISE WEREN'T
IMPROVISERS AND I THOUGHT THE
BEST IMPROVISERS CAN IMPROVISE
WITH ANYONE, THEY CAN MAKE A
PERFORMANCE WITH ANYONE AND
ANYTHING.
I THOUGHT THIS CHATBOT SYSTEM
THAT WAS GAINING SOME ACCEPTANCE
AND WAS MAKING SOME LEAPS IN THE
COMPUTING SCIENCE WORLD AND I
THOUGHT IF THE BEST IMPROVISER
CAN IMPROVISE WITH ANYTHING,
THEY CAN WITH A COMPUTER.
THAT'S WHERE THE SEED OF THE
SHOW STARTED.
I PITCHED IT TO THE DIRECTOR
HERE AT RAPID FIRE THEATRE IN
EDMONTON AND HE SAID THAT SOUNDS
CHALLENGING AND DYNAMIC AND
LET'S SEE IT WORK.
AND HE GAVE ME THE GREEN LIGHT
TO START THE SHOW AND THAT WAS
ABOUT A YEAR AND A HALF AGO.

Steve says SPEAKING OF SEEING
IT WORK, WE SHOULD DO THAT NOW.
WE'LL PLAY AND CHIP AND COME
BACK AND CHAT.
SHELDON, IF YOU WOULD, ROLL THE CLIP.

In a clip, Kory stands on a stage, next to a projection screen holding a microphone. On the screen, a picture of a woman appears next to moving text.

Kory says WHAT'S WRONG?
WHAT'S WRONG WITH YOU?

[LAUGHTER]

A female machine voice says
I'M NOT SURE IT WOULD BE
REALLY FUNNY FOR YOU TO KNOW.

[LAUGHTER]

Kory says I UNDERSTAND VANESSA.
CAN YOU TELL ME, HAVE YOU
THOUGHT ABOUT NAMES FOR THE BABY?
I HAVE A LIST THAT I'VE PREPARED.

The machine says YOU'D DO ANYTHING FOR ME.
IS THERE A POINT I HAVE TO TELL YOU?

Kory says ONCE THE BABY IS ALIVE,
YOU'RE PROBABLY GOING TO HAVE TO TELL ME.

[LAUGHTER]

The clip ends.

Steve says WE'VE GOT TO FIGURE
THIS ALL OUT HERE.
KORY, BEFORE THERE'S A BASE OF
KNOWLEDGE HERE THAT I FEAR WE
ALL DON'T KNOW THAT WE NEED TO
UNDERSTAND IN ORDER FOR US TO
GET HOW THIS WORKS.
SO START OFF BY TALKING TO US
ABOUT THE DISTINCTION HERE
BETWEEN ARTIFICIAL INTELLIGENCE
AND AUTOMATION AND ALGORITHMS,
BECAUSE ALL THAT IS WRAPPED UP
IN WHAT YOU'RE TRYING TO DO
HERE, I GATHER; IS THAT RIGHT?

Kory says THAT'S RIGHT, IT'S GREAT TO
START WITH SOME SORT OF
GROUNDWORK.
I LIKE THE TERM MACHINE
INTELLIGENCE.
WE THINK ABOUT MACHINE
INTELLIGENCE AS BEING ANYTHING
FROM A THERMOSTAT IN A HOUSE ALL
THE WAY UP TO A CAR THAT CAN
DRIVE ITSELF AROUND A CITY.
SO THAT IS A MACHINE THAT HAS
SOME FORM OF INTELLIGENCE.
AND THAT'S A BIT OF A SPECTRUM.
AND SOMEWHERE ALONG THAT
SPECTRUM, WELL, THAT'S WHERE
YOU'D CONSIDER ARTIFICIAL
INTELLIGENCE OR A TERM THAT'S
BEEN SORT OF MADE IN THE 1950s
OR SO.
ARTIFICIAL INTELLIGENCE IS SORT
OF THE POINT ON THAT MACHINE
INTELLIGENCE SPECTRUM WHERE WE
START TO SEE THE MACHINES HAVING
SOME HUMAN LEVEL ABILITIES IN
HOW THEY PROCESS THE WORLD OR
MOVE AROUND IN THE WORLD.
THAT IS SORT OF ARTIFICIAL
INTELLIGENCE ON THE SPECTRUM OF
MACHINE INTELLIGENCE.
AN ALGORITHM IS SORT OF THE WAY
IN WHICH WE WOULD TEACH THESE
MACHINES TO BE INTELLIGENT OR
HOW WE WOULD GIVE THEM KNOWLEDGE
OF THE WORLD.
YEAH, THE AUTOMATION I GUESS IS
MORE LIKE PRECISELY PROGRAMMING
A SYSTEM TO DO SOMETHING, A
ROUTINE, A PROCEDURE OVER AND
OVER AGAIN WITH A HUMAN LESS IN
THE LOOP OR MAYBE COMPLETELY OUT
OF THE LOOP.

Steve says THE THING I GATHER
THAT PEOPLE FIND SO AMAZING
ABOUT ALL OF THIS IS THAT YOU
HAVE KIND OF, THROUGH THE
MIXTURE OF ALL THOSE THREE
THINGS, PROGRAMMED YOUR
ARTIFICIAL INTELLIGENCE, YOUR,
YOU KNOW, IF YOU'RE ABBOTT, THEN
THAT'S COSTELLO! HAVE SOME
INTUITIVE UNDERSTANDING THAT
WHAT YOU'RE SAYING IS FUNNY AND
SHE NEEDS TO COME BACK WITH SOMETHING FUNNY.
HOW DO YOU EVEN BEGIN TO DO THAT?

The caption changes to "Kory Mathewson. Improv Comedian."

Kory says IT'S EXACTLY THE RESEARCH
QUESTION THAT STARTED THIS ALL
OFF, IS HOW DO YOU START TO MAKE
A COMEDIC PERFORMER THAT, YOU
KNOW, DOESN'T NEED TO BE THE
BEST IMPROVISER AT THE BEGINNING
BUT SLOWLY TRAINS AND LEARNS HOW
TO DO IT.
SO WHAT I DID WAS TAKE ALL OF
THE MOVIE SCRIPTS, AND NOT JUST
A COLLECTION BUT ALL OF THE
MOVIE SCRIPTS FROM THE LAST
HUNDRED YEARS, SO HUNDREDS OF
THOUSANDS OF MOVIE SCRIPTS, AND
BROKE THEM DOWN INTO THEIR
COMPONENT WORDS AND THEN GIVE
THAT ALL TO A DEEP-LEARNING
SYSTEM TO START TO CONSUME AND
ALL IT DOES IS TRY TO PREDICT
THE NEXT WORD.
SO IT LOOKS AT EVERY MOVIE
SCRIPT EVER AND SAYS IF THESE
WORDS HAPPENED, ONE, TWO, THREE,
WHAT IS WORD NUMBER FOUR?
IF IT CAN START TO PREDICT THAT
WITH HIGHER AND HIGHER ACCURACY,
IT STARTS TO HAVE A LANGUAGE
MODEL AND THAT IS UNDERLYING
MOVIE SCRIPTS.
IN DOING SO, I CAN SAY LINES AS
THOUGH THEY SOUND LIKE THEY'RE
FROM A MOVIE AND IT WILL RESPOND
PREDICTING WHAT THE NEXT LINE
WOULD BE FROM A MOVIE OR A MOVIE
IN GENERAL.
THAT'S LANGUAGE LEARNING.

Steve says YOUR ROBOT'S NAME IS
ALEX WHICH IS SORT OF AN ACRONYM
FOR ARTIFICIAL LANGUAGE
EXPERIMENT.

Kory says THAT'S RIGHT.

Steve says I GUESS WE SHOULD
FIND OUT, HOW IS INTERACTING
WITH YOUR ALEX DIFFERENT FROM
OUR EVERYDAY INTERACTING WITH
APPLE'S SIRI?

In a picture, Kory stands in a classroom next to an open laptop, holding a small plastic robot.

The caption changes to "How to train a computer."

Kory says THAT'S A GREAT QUESTION.
SO THE FIRST THING THAT YOU
MIGHT NOTICE FROM THAT CLIP IS
THAT THE DELAY TIME FROM MY
INPUT TO ALEX'S OUTPUT IS A
LITTLE BIT LONGER THAN SIRI.
HUMANS USUALLY HAVE A DELAY
RESPONSE OF ABOUT 200 TO 800
MILLISECONDS AND ALEX'S IS ABOUT
A SECOND OR TWO SECONDS,
DEPENDING ON HOW MUCH THINKING
IS GOING ON IN THE SYSTEM.
SIRI CAN CUT DOWN THAT TIME EVEN
SHORTER BECAUSE IT USES
RULE-BASED STRATEGIES TO
RESPOND.
SO IF YOU ASK SOMETHING, IT'LL
LOOK THROUGH A RULE-BASED TREE
AND THEN RESPOND WITH THE
ACCORDING RESPONSE.
WHAT ALEX IS IS A GENERATIVE
SYSTEM.
IT'S TRYING TO GENERATE
RESPONSES ONLINE IN THAT MOMENT
AND THEN SELECT THE BEST
RESPONSE.

Steve says HAVE YOU GOT SOME
MORE TWEAKING OF ALEX THAT YOU
WANT TO DO?

Kory says ABSOLUTELY.
THERE'S ALWAYS TWEAKING.
I NEED TO REDUCE THAT DELAY
TIME.
I ALSO WANT TO SELECT THE BEST
CANDIDATE.
I MAY NOT WANT THE MOST PROBABLE
RESPONSE OR A UNIQUE OR NOVEL
INTERESTING RESPONSE.
I WANT ONE THAT CONNECTS WHAT I
WAS SAYING WITH WHAT HAPPENED
BEFORE IN THE SCENE, AND THESE
ARE KIND OF NEW REWARD THEMES OR
NEW SCHEMES THAT I MIGHT USE TO
TRAIN ALEX.

Steve says LET ME JUST FOLLOW
UP ON THE POINT YOU MADE A
MOMENT AGO ABOUT HOW YOU PUT 100
COMEDY SCRIPTS INTO THIS THING
SO THEY WOULD HAVE SOME SENSE OF
HOW TO RESPOND.
I MEAN, THAT JUST SOUNDS LIKE IT
SHOULD PROBABLY TAKE A HUNDRED
YEARS TO DO.
OBVIOUSLY IT DIDN'T.
BUT HOW TIME-CONSUMING WAS IT TO
GET ALL THAT DONE?

The caption changes to "Kory Mathewson, @korymath."
Then, it changes again to "Connect with us: @theagenda, TVO.org, Facebook, YouTube, Periscope, Instagram."

Kory says IT'S AN INTERESTING QUESTION.
SO THE FACT THAT THESE COMPUTER
SYSTEMS THAT WE'RE USING NOW ARE
SO MUCH FASTER AND SO MUCH
BIGGER SPACE-WISE THAN WE HAD IN
THE PAST MEANS THAT IT'S EVEN
POSSIBLE FOR US TO DO THIS.
NOW WE CAN TRAIN THESE KIND OF
SYSTEMS... ALEX WAS TRAINED FOR
ABOUT A MONTH AND A HALF.
SO CONSUMING ALL THE DATA,
CONSUMING ALL OF THE DIFFERENT
RULES AND SCRIPTS, WELL, THAT
TOOK SOME PROGRAMMING AND THAT
TOOK ABOUT THREE WEEKS OF
PROGRAMMING, A MONTH OF
PROGRAMMING, AND THEN ONCE WE
START THE TRAINING, THEN IT HAS
TO TRAIN FOR A MONTH AND A HALF.
WHAT IT'S DOING IN THAT TRAINING
IS LEARNING THAT UNDERLYING
LANGUAGE MODEL.
AND WE WOULD IMAGINE THAT AS
COMPUTERS START TO GET FASTER
AND FASTER, WELL, THAT TRAINING
TIME CAN EVEN REDUCE.

Steve says I ASSUME THERE WAS A
LOT OF FAILURE ALONG THE WAY. YES?

Kory says THERE WAS TONS OF FAILURE.
THE FIRST SHOW I EVER DID AT
RAPID FIRE THEATRE HERE IN
EDMONTON, THE FIRST SHOW I EVER
DID WAS ALMOST A FAILURE.
JUST WALKED THAT TIGHT ROPE OF
FAILURE.
YOU KNOW, IT WAS MORE OF A
SHOWING WATCHING ME TRYING TO
DEBUG A SYSTEM LIVE ON STAGE,
AND THEN WHEN IT FINALLY DID SAY
SOMETHING, IT SAID SOMETHING
RATHER BLUE AND OFF-COLOUR, AND
HERE I AM ON STAGE WITH AN
IMPROV PARTNER THAT'S NOW SAID
SOMETHING THAT'S PARTIALLY
INCRIMINATING, SO THAT'S A
FAILURE.
IT'S ALSO A FAILURE WHEN THERE'S
TECHNICAL DIFFICULTIES.
LIKE THE IMPROV WISDOM SAYS, ANY
KIND OF FAILURE IS SUCCESS,
REALLY.
IT'S ALL PERSPECTIVE.
SO YOU DON'T KNOW UNTIL YOU
START TO ACCEPT AND ACKNOWLEDGE
THAT FAILURE.
AND THAT'S PART OF WHERE THIS
CAME FROM, RIGHT?
IN THAT IMPROV IS ALL ABOUT
USING YOUR FAILURE TO SUCCEED,
AND THE IMPROV WANTS TO WATCH
THE STRUGGLE BETWEEN FAILURE AND
SUCCESS.
AND THE SAME IS TRUE IN TRAINING
THESE MACHINE LEARNING SYSTEMS.
THEY'RE GOING TO FAIL AND LEARN
EVERY TIME THEY FAIL.
SO IT WAS DRAWING THOSE
ANALOGIES TOGETHER THAT SORT OF
GOT ME TO CONNECT THESE IDEAS.

Steve says WHAT'S THE HARDEST
PART OF TEACHING HUMANS AND
MACHINES TO DO ALL OF WHAT YOU
HAVE BEEN DESCRIBING?

The caption changes to "Performing with a robot."

Kory says THE HARDEST PART OF TEACHING
HUMANS TO DO IMPROV IS GETTING
THEM TO FORGET EVERYTHING THEY
LEARNED SOCIALLY DYNAMICALLY.
FIVE-YEAR-OLDS ARE REALLY GOOD
AT IMPROV.
THEY'RE NOT AFRAID TO SAY WHAT
THEY ARE THINKING.
ONCE YOU GET PEOPLE IN THE
WORKPLACE, PEOPLE HESITATE AND
THEY DON'T WANT TO NECESSARILY
JUMP IN AND SAY SOMETHING THAT
MIGHT BE EMBARRASSING OR
OTHERWISE.
IT'S GETTING THEM TO LOSE THAT
INHIBITION AND SAY IMPULSIVELY
WHAT'S ON THEIR MIND.
THE MOST DIFFICULT THING ABOUT
TRAINING THE COMPUTER SYSTEMS IS
WORKING OUT THE TECHNOLOGY
BEHIND IT AND ENSURING THAT
WE'RE REDUCING THE AMOUNT OF
TIME SO IT SOUNDS MORE
HUMAN-LIKE, WE'RE ALLOWING IT TO
TRAIN BETTER AND LEARN BETTER
AND NOT CATASTROPHICALLY FORGET
EVERYTHING IT'S LEARNED ALONG
THE WAY.

Steve says MY HUNCH IS A LOT OF
PEOPLE WATCHING US RIGHT NOW, I
REFERENCED ABBOTT AND COSTELLO
EARLIER, ROWAN AND MARTIN, BURNS
AND ALLEN, THERE HAVE BEEN GREAT
COMEDY DUOS OVER THE YEARS.

Kory says SURE.

Steve says YOUR COMEDY PARTNER
IS ESSENTIALLY AN ALGORITHM.
HOW ODD IS THAT FOR YOU?

Kory says IT IS QUITE WEIRD.
MY COMEDY PARTNER AND WHAT I'VE
CREATED AND IT'S WEIRD TO THINK
THAT ALL THESE HUMANS THAT I'VE
ENJOYED DOING COMEDY WITH ARE
NOW WATCHING ME DO IMPROV WITH
AN ALGORITHM.
ONE OF THE THINGS THAT I REALLY
LIKE IS THE IDEA THAT I CAN
SHARE THIS ALGORITHM WITH
EVERYONE.
SO I'VE BUILT A COMEDY PARTNER
THAT NOW A WHOLE GENERATION OF
IMPROVISERS CAN GO OUT AND
PERFORM WITH.
IT'S NOT JUST OUR DYNAMIC THAT'S
INTERESTING, BUT NOW I GET TO
SHARE THAT DYNAMIC WITH AN
ENTIRE COMMUNITY OF IMPROVISERS.

Steve says AND IT CHALLENGES
YOU AS AN IMPROVISER TOO, I PRESUME?

Kory says IT ABSOLUTELY DOES.
ALEX'S LIMITATIONS ARE RIGHT NOW
THAT IT CAN'T REALLY JUSTIFY
WHAT'S HAPPENING IN THE
IMPROVISATION, IN THE COMEDY, IT
DOESN'T REALLY UNDERSTAND WHAT'S
HAPPENING, WHICH MEANS IT'S UP
TO ME TO DO A LOT OF THE
JUSTIFICATION, IT'S UP TO ME TO
BRAID THE THREADS TOGETHER IN A
CLEVER NARRATIVE.
I THINK THE IMPROV IS ENGAGING
AND INTERESTING FOR THE AUDIENCE
BECAUSE THEY WATCH ME SORT OF
FLY BY THE SEAT OF MY PANTS
TRYING TO CONNECT THESE THINGS
AND THE RANDOM NOTIONS THAT ALEX
BRINGS UP.

Steve says LEST ANYBODY THINK
YOU ACTUALLY DON'T KNOW HOW TO
RELATE TO OTHER HUMAN BEINGS,
LET US ASSURE PEOPLE THAT
THERE'S ANOTHER IMPROVISER THAT
LIKES WORKING WITH A.I. IN THE
UNITED KINGDOM.
WE'D LIKE TO SHOW A CLIP OF YOUR
WORK TOGETHER AND WE'LL COME
BACK AND CHAT AGAIN.
SHELDON, THE CLIP, IF YOU WOULD?

In a clip, a blond female interviewer in her thirties stands on a sidewalk facing a balding man in his forties.

She says A SHOW INVOLVING A ROBOT AND
TWO HUMANS.
WHAT WAS YOUR IMPRESSION OF THAT?

The man says LOTS OF IT WAS FUNNY.
SOMETIMES IT WORKED AND
SOMETIMES IT DIDN'T.
SOMETIMES WHEN IT DIDN'T WORK,
IT WAS EVEN FUNNIER.

In Kory's show, the robot, using a male voice, says
THERE IS ONLY SO MUCH THAT I
CAN SHOW YOU.

[LAUGHTER]

The man says THE HUMAN PERFORMER WOULD SAY
SOME THINGS AND THEN THE ROBOT
WILL COME OUT WITH SOME SORT OF
RANDOM NON SEQUITUR.

The robot says WHAT HAVE YOU DONE WITH MY
FATHER?

[LAUGHTER]

The robot says WHY DID YOU GO TO THE
HOSPITAL?

[LAUGHTER]

The man says THAT'S A LONG WAY FROM THE
MACHINE, ONE, HAVING INTENT AND,
TWO, KNOWING WHAT THEY'VE SAID
IS FUNNY.

The clip ends.

Steve says HOW HAVE YOU TWO
BEEN ABLE TO COLLABORATE ON THIS?

A poster pops up in which Kory and the man stands side by side grinning as they hold up the two types of robots.

Kory says IT'S AN AWESOME QUESTION.
AFTER I DID THAT SHOW HERE IN
EDMONTON AT RAPID FIRE THEATRE I
PUT UP A BLOG POST THAT SAID
HERE'S WHAT I WANTED TO DO,
HERE'S WHAT WORKED AND WHAT
DIDN'T WORK, AND MOST OF IT
DIDN'T WORK.
A FEW DAYS LATER IN MY INBOX
COMES AN e-mail THAT I OPEN UP
FROM PIOTR AND IT SORT OF SAYS TWO THINGS: ONE IS, SCREW YOU.
I WAS WORKING ON TRAINING A
MODEL AND I WASN'T READY YET.
AND TWO, LET'S WORK TOGETHER.
PIOTR IS AN INCREDIBLY TALENTED
A.I. RESEARCHER AND A VERY FUNNY
IMPROVISER.
HE WAS WORKING ON SOMETHING VERY
SIMILAR.
HE SAID LET'S WORK TOGETHER.
AT THAT POINT OUR PARTNERSHIP
STARTED.
WE BUILT THIS COMPANY CALLED
HUMAN MACHINE AND WE'VE PRODUCED
A FEW DIFFERENT SHOWS.
WE'VE DONE IMPROV TOTALLY
REMOTELY IN TWO DIFFERENT
THEATRES TO TWO DIFFERENT
AUDIENCES WITH TWO HUMANS AND
TWO AIs, A SHOW CALLED BINARY
TWO.
WE HAVE ANOTHER SHOW WHERE WE
PAIR UP HUMANS AND HUMANS THAT
ARE PUPPETTED BY AIs AND
HUMANS THAT ARE PUPPETTED BY
OTHER HUMANS AND CONNECT THEM
ALL ON STAGE TOGETHER.
SO WE GET TO BOUNCE IDEAS OFF
EACH OTHER.
IT'S SUCH A SMALL COMMUNITY
WORKING AT THIS INTERSECTION
BETWEEN SCIENCE AND ART THAT
IT'S NEAT TO HEAR SOME
REVERBERATIONS FROM OTHER
LIKE-MINDED FOLK.

Steve says YOU KNOW, THE BIG
IMPROV PLACE IN TORONTO OF
COURSE IS SECOND CITY.
I WONDER IF YOU'VE HEARD
ANYTHING FROM THEM OR ANY OTHER
PEOPLE IN THE IMPROV BUSINESS?

The caption changes to "Watch us on Facebook Live: Weekdays at 8 PM."

Kory says ABSOLUTELY.
SO TWO WONDERFUL FOLKS, ONE IS
IN TORONTO WHO WORKED FOR SECOND
CITY AND NOW WORKS AT BAD DOG
COMEDY AND HE DOES A SHOW CALLED
YES ANDROID WHERE HE CONNECTS
SOME A.I. WITH SOME PEOPLE BUT
IT'S STARTING UP AND FINDING ITS
FEET, AND THE OTHER IS RACHEL
ROSENTHAL.
SHE HAS WRITTEN AN INTERESTING
ESSAY ON THE WORK PIOTR AND I
HAVE DONE WITH ALEX.
IT STARTED I THOUGHT I WOULD
HATE THIS AND IT GOES ON TO
DESCRIBE HER INTERACTION AND
UNDERSTANDING OF THE SYSTEM.
SO THE COMMUNITY IS SLOWLY
GROWING AND IT'S OPPORTUNITIES
LIKE THIS THAT I GET TO SHARE
WITH OTHER PEOPLE AND SAY THIS
IS AN EXCITING PLACE IF YOU'RE A
COMEDIAN, IF YOU'RE A SCIENTIST,
IF YOU'RE AN ARTIST.
WE CAN CONNECT ON THIS.
AND BECAUSE ALL THIS TECHNOLOGY
IS OPEN SOURCE AND IT'S
ACCESSIBLE, IT MEANS THAT THESE
A.I. SYSTEMS CAN BE TRAINED BY
ANYONE.
NOW WE HAVE THE TECHNOLOGY TO BE
ABLE TO TRAIN THESE SYSTEMS.
ANYONE CAN IN THEIR COMPUTER IN
A DAY OR SO AND HAVE SOMETHING
INTERESTING THAT THEY CAN START
TO INTERACT WITH.

Steve says TAKE THIS FORWARD.
LET'S GO FIVE YEARS FORWARD.
HOW DO YOU IMAGINE THIS KIND OF
CHANGING SHOW BUSINESS AS WE
KNOW IT TODAY?

The caption changes to "A.I. powered humour."

Kory says ABSOLUTELY.
TWO THINGS I IMAGINE IS THAT,
ONE, WE WILL START TO IMMERSE
OURSELVES A LITTLE MORE IN
CINEMA.
SO WE CAN IMAGINE A VIRTUAL
REALITY MOVIE WHERE WE LOOK
AROUND AND THE MAIN SCENE IS
HAVING BETWEEN ACTORS.
BUT OVER ON THIS SIDE, IF WE
LOOK AND TURN OUR HEADS, WELL
THEN WE MIGHT SEE SOME
ARTIFICIAL INTELLIGENT DIALOGUE
SYSTEMS COMMUNICATING WITH EVE
OTHER.
IT'S NOT THE MAIN STORY PLOT BUT
THEY HAVE THEIR OWN SUB PLOT.
WE IMAGINE A MORE IMMERSIVE
WORLD THAT IS FLUSHED OUT BY
SOME OF THESE DIALOGUE SYSTEMS.
AND YOU'RE WALKING AROUND A
VIDEO GAME THAT'S IMMERSIVE, IF
YOU GO INTO THE VILLAGE, THE
VILLAGE IS FULL OF A.I. SYSTEMS
AND YOU HAVE TO SORT OF
NEGOTIATE WITH THEM IN
CONVERSATIONS TO GET CERTAIN
THINGS.
I ALSO IMAGINE THAT THE SHOWS
THAT WE'RE DOING ON STAGE ARE
GOING TO CHANGE.
RIGHT NOW I'M INTERACTING WITH
MOSTLY DIALOGUE.
WE CAN IMAGINE SIMPLE
IMPLEMENTATIONS THAT WOULD ALLOW
US TO PROJECT LIGHT INTO THE
SPACE, PROJECT DIFFERENT IMAGES,
PROJECT VIDEO AND ALL OF THAT
COULD BE CONTROLLED
ALGORITHMICALLY AS WELL.
IT MIGHT CHOOSE TO ADD A SOFT,
SUBTLE MUSIC, A ROMANTIC
BACKDROP OF THESE VENETIAN
PILLARS AND ROBES AND START TO
TALK IN THIS ROMANTIC ROMAN
LANGUAGE AND ALL OF A SUDDEN I'M
MUCH MORE IMMERSED THAN IF IT
JUST SORT OF TELLS ME.
IT'S THE IDEA IN ACTING OF SHOW
IT, DON'T TELL IT.
DO THE THING RATHER THAN SAY THE THING.

Steve says KORY, WHAT MAKES YOU
THINK THIS IDEA HAS LEGS AND
WON'T BE JUST SORT OF A GIMMICK
FOR TODAY AND GONE TOMORROW?

Kory says IT'S A GOOD QUESTION AND IT'S
FUNNY YOU SAY LEGS.
IT IS A COMPUTER AND IT DOESN'T
REALLY HAVE LEGS.

Steve says GOOD POINT.

Kory says WATCHING AN AUDIENCE WATCH
THESE SHOWS IS WHAT BRINGS ME
THE MOST HOPE BECAUSE I CAN SEE
THE FASCINATION THAT THE
AUDIENCES HAVE WITH HAVING A
ROBOT ON STAGE, AND THEN
WATCHING THE IMPROVISERS
INTERACT WITH IT.
IT LOOKS LIKE THEY'RE BEING
UNLOCKED AND UNLEASHED AND ABLE
TO TELL THESE MORE FANCIFUL
STORIES OF FUTURISTIC
TECHNOLOGY, SOMETHING THAT WAS
SORT OF SAVED FOR SCIENCE
FICTION IN BOOKS AND TV, AND NOW
WE CAN DO IT LIVE ON STAGE
BECAUSE WE HAVE THESE ROBOTS WE
CAN INTERACT WITH.

Steve says I WAS GOING TO SAY.
IN SOME RESPECTS, YOU'RE TOO
YOUNG TO REMEMBER THIS, BUT I
GREW UP WITH A TV SHOW CALLED
"LOST IN SPACE" AND THERE WAS A
ROBOT THAT THE ROBINSON FAMILY
INTERACTED WITH.

Kory says YEAH.

Steve says OF COURSE, THERE WAS
SOMEBODY INSIDE THE ROBOT AND IT
WASN'T EXACTLY WHAT YOU'RE
DOING, BUT THE NOTION IS PRETTY
SIMILAR, ISN'T IT?
IT'S KIND OF INTERACTING WITH
THE FUTURE.
YOU'RE DOING IT RIGHT NOW.
THAT'S PRETTY COOL, RIGHT?

The caption changes to "New frontiers."

Kory says THANK YOU, YEAH.
I ABSOLUTELY DO THINK IT'S COOL.
IS THAT YOUR FAVOURITE ROBOT?
WOULD YOU SAY THAT'S YOUR
FAVOURITE ROBOT?

Steve says I GUESS THAT'S IT.
IT'S THE ONE THAT COMES TO MIND.

Kory says MY FAVOURITE IS ROSIE FROM
THE JETSONS.
THERE ARE ROBOTS THAT HAVE BEEN
DEPICTED IN SCIENCE FICTION IN
DIFFERENT WAYS AND THEY'RE NOT
NECESSARILY SUBSERVIENT.
ROSIE WAS QUITE SASSY AND THE
ROBOT FROM "LOST IN SPACE" WAS
HELPFUL BUT NOT NECESSARILY
PERFECT.
AND THAT'S SORT OF WHAT WE'RE
DOING HERE.
WE'RE SHOWING THAT THESE ROBOTS
DO EXIST NOW.
WE CAN INTERACT WITH THEM NOW.
BUT THEY'RE NOT TO BE FEARED.
THEY'RE NOT SCARY.
THEY ARE LIMITED IN THEIR
CAPACITY.
AND SHOWING THOSE LIMITATIONS
SORT OF ALLOWS A COMFORT WASH
OVER THE AUDIENCE.
I'M NOT AS WORRIED ABOUT ROBOTS
TAKING OVER THAN I THOUGHT I
WOULD BE.

Steve says KORY, FOR THE
RECORD, MY ROBOT WAS REAL.
YOURS IS A CARTOON.
THAT'S A BIG DIFFERENCE THERE.

Kory says SURE.

Steve says LET'S FINISH UP ON
THIS.
IF YOU ARE IN FACT ONTO
SOMETHING HERE AND THIS HAS
BROADER APPLICATIONS GOING
FORWARD, WHAT DO YOU IMAGINE
ARTIFICIAL INTELLIGENCE, OTHER
APPLICATIONS COULD BRING TO THE
THEATRE WORLD GOING FORWARD?

Kory says YEAH.
I IMAGINE THAT THE THEATRE IS
HERE TO INTERPRET AND SHARE
HUMAN EXPERIENCE WITH LARGER AND
LARGER AUDIENCES, AND I THINK
THAT ARTIFICIAL INTELLIGENCE HAS
AN INTERESTING PLACE TO PLAY IN
UNDERSTANDING THE PERCEPTION OF
THE AUDIENCE.
SO WE CAN NOW IMAGINE POINTING
CAMERAS AND MICROPHONES AT THE
AUDIENCE AND MEASURING THEIR
ENGAGEMENT DURING A SHOW, AND
THEN CHANGING THE SHOW, MAKING
IT MORE ENGAGING, COMPELLING,
SAD, HAPPY AS THE AUDIENCE IS
BEING MEASURED IN THEIR
PERCEPTION OF IT.
SO WE COULD IN FACT IMAGINE A
MORE IMMERSIVE INTERACTIVE
THEATRE EXPERIENCE IF THE
AUDIENCE GETS TO CONTROL THE
MUSIC THAT'S HAPPENING AND THEN
WATCH THIS SCENE ON STAGE.
THAT WOULD ALLOW US TO CONNECT
WITH EMOTIONS EVEN DEEPER THAN
WE CAN RIGHT NOW.
RIGHT NOW IT'S ABOUT THE ACTORS
SORT OF PORTRAYING THAT EMOTION
SO CLEAR.
IF WE COULD ACTUALLY PULL THAT
EMOTION MEASUREMENT FROM THE
AUDIENCE, THAT WOULD GIVE US ALL
KINDS OF INTERACTIVITY. YEAH.

Steve says GOOD.
WHEN ARE YOU GOING TO HAVE YOUR
Ph.D. DONE?

Kory says NOVEMBER 2018.

Steve says SO PERHAPS NEXT TIME
YOU COME HERE I'M GOING TO HAVE
TO CALL YOU DOCTOR; IS THAT RIGHT?

Kory says THAT WOULD BE THE HOPE, YEAH.

The caption changes to "Producer: Patricia Kozicka, @TrishKozicka."

Steve says GOOD FOR YOU.
WELL, IN ADVANCE WE WISH YOU
WELL FINISHING THE Ph.D. AND
WE CAN'T THANK YOU ENOUGH FOR
SPARING TIME FOR TVO TONIGHT.
KORY MATHEWSON, IMPROV COMEDIAN
AND A Ph.D. STUDENT IN MACHINE
LEARNING. THANKS SO MUCH, KORY.

Kory says THANK YOU SO MUCH.

Watch: Teaching Machines to be Funny