Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney
-
0:17 - 0:19I have a question:
-
0:19 - 0:21Can a computer write poetry?
-
0:22 - 0:24This is a provocative question.
-
0:25 - 0:27You think about it for a minute,
-
0:27 - 0:30and you suddenly have a bunch
of other questions like: -
0:30 - 0:32What is a computer?
-
0:32 - 0:33What is poetry?
-
0:34 - 0:36What is creativity?
-
0:36 - 0:38But these are questions
-
0:38 - 0:41that people spend their entire
lifetime trying to answer, -
0:41 - 0:43not in a single TED Talk.
-
0:43 - 0:45So we're going to have to try
a different approach. -
0:45 - 0:48So up here, we have two poems.
-
0:48 - 0:50One of them is written by a human,
-
0:50 - 0:53and the other one's written by a computer.
-
0:53 - 0:55I'm going to ask you to tell me
which one's which. -
0:56 - 0:58You're not going to have long to read
-
0:58 - 1:00because we haven't got long
to do this speech. -
1:00 - 1:02Have a go, start reading.
-
1:03 - 1:07Poem 1: Little Fly / Thy summer's play, /
My thoughtless hand / Has brush'd away. -
1:07 - 1:10A I not / A fly like thee? /
Or art not thou / A man like me? -
1:10 - 1:13Poem 2: We can feel / Activist
through your life's / morning / -
1:13 - 1:17Pauses to see, pope I hate the / Non
all the night to start a great otherwise -
1:17 - 1:18Alright, time's up.
-
1:18 - 1:23Hands up if you think Poem 1
was written by a human. -
1:24 - 1:26OK, most of you.
-
1:26 - 1:29Hands up if you think Poem 2
was written by a human. -
1:30 - 1:31Very brave of you,
-
1:31 - 1:36because the first one was written
by the human poet William Blake. -
1:36 - 1:39The second one was written by an algorithm
-
1:39 - 1:43that took all the language
from my Facebook feed on one day -
1:43 - 1:45and then regenerated it algorithmically,
-
1:45 - 1:49according to methods that I'll describe
a little bit later on. -
1:49 - 1:52But most of you got that right,
it's probably a little bit easy. -
1:52 - 1:54So let's try another test.
-
1:56 - 1:58Again, you haven't got ages to read this,
-
1:58 - 2:00so just trust your gut.
-
2:00 - 2:03Poem 1: A lion roars and a dog barks.
It is interesting / and fascinating -
2:03 - 2:07that a bird will fly and not / roar
or bark. Enthralling stories about animals -
2:07 - 2:11are in my dreams and I will sing them all
if I / am not exhausted or weary. -
2:11 - 2:14Poem 2: Oh! kangaroos, sequins, chocolate
sodas! / You are really beautiful! -
2:14 - 2:18Pearls, / harmonicas, jujubes, aspirins!
All / the stuff they've always talked about -
2:22 - 2:24Alright, time's up.
-
2:24 - 2:26So if you think the first poem
was written by a human, -
2:26 - 2:28put your hand up.
-
2:29 - 2:30OK.
-
2:30 - 2:32And if you think the second poem
was written by a human, -
2:32 - 2:34put your hand up.
-
2:35 - 2:38We have, more or less, a 50/50 split here.
-
2:39 - 2:40It was much harder.
-
2:40 - 2:41The answer is,
-
2:41 - 2:45the first poem was generated
by an algorithm called Racter, -
2:45 - 2:48that was created back in the 1970s,
-
2:48 - 2:51and the second poem was written
by a guy called Frank O'Hara, -
2:51 - 2:54who happens to be one
of my favorite human poets. -
2:54 - 2:57(Laughter)
-
2:58 - 3:01So what we've just done now
is a Turing test for poetry. -
3:01 - 3:06The Turing test was first proposed
by this guy, Alan Turing, in 1950, -
3:06 - 3:07in order to answer the question,
-
3:07 - 3:10can computers think?
-
3:10 - 3:12Alan Turing believed that if
a computer was able -
3:12 - 3:15to have a to have a text-based
conversation with a human, -
3:15 - 3:18with such proficiency
such that the human couldn't tell -
3:18 - 3:21whether they are talking
to a computer or a human, -
3:21 - 3:24then the computer can be said
to have intelligence. -
3:24 - 3:27So in 2013, my friend
Benjamin Laird and I, -
3:27 - 3:30we created a Turing test
for poetry online. -
3:30 - 3:31It's called bot or not,
-
3:31 - 3:33and you can go and play it for yourselves.
-
3:33 - 3:36But basically, it's the game
we just played. -
3:36 - 3:37You're presented with a poem,
-
3:37 - 3:40you don't know whether it was written
by a human or a computer -
3:40 - 3:41and you have to guess.
-
3:41 - 3:45So thousands and thousands
of people have taken this test online, -
3:45 - 3:46so we have results.
-
3:46 - 3:48And what are the results?
-
3:48 - 3:50Well, Turing said that if a computer
could fool a human -
3:50 - 3:5330 percent of the time
that it was a human, -
3:53 - 3:55then it passes the Turing test
for intelligence. -
3:56 - 3:58We have poems on the bot or not database
-
3:58 - 4:01that have fooled 65 percent
of human readers into thinking -
4:01 - 4:03it was written by a human.
-
4:03 - 4:05So, I think we have an answer
to our question. -
4:07 - 4:09According to the logic of the Turing test,
-
4:09 - 4:10can a computer write poetry?
-
4:10 - 4:13Well, yes, absolutely it can.
-
4:14 - 4:16But if you're feeling
a little bit uncomfortable -
4:16 - 4:19with this answer, that's OK.
-
4:19 - 4:21If you're having a bunch
of gut reactions to it, -
4:21 - 4:24that's also okay because
this isn't the end of the story. -
4:25 - 4:27Let's play our third and final test.
-
4:28 - 4:30Again, you're going to have to read
-
4:30 - 4:32and tell me which you think is human.
-
4:32 - 4:35Poem 1: Red flags the reason
for pretty flags. / And ribbons. -
4:35 - 4:39And wearing material / Reasons
for wearing material. / Give pleasure. -
4:39 - 4:43Poem 2: A wounded deer leaps
highest, / I've heard the daffodil -
4:43 - 4:46I've heard the flag to-day /
I've heard the hunter tell; / -
4:46 - 4:50'Tis but the ecstasy of death, /
And then the brake is almost done -
4:50 - 4:53And sunrise grows so near /
sunrise grows so near -
4:53 - 4:56That we can touch the despair and /
frenzied hope of all the ages -
4:56 - 4:58OK, time is up.
-
4:58 - 5:02So hands up if you think Poem 1
was written by a human. -
5:03 - 5:06Hands up if you think Poem 2
was written by a human. -
5:06 - 5:08Whoa, that's a lot more people.
-
5:09 - 5:12So you'd be surprised to find that Poem 1
-
5:12 - 5:15was written by the very
human poet Gertrude Stein. -
5:17 - 5:22And Poem 2 was generated
by an algorithm called RKCP. -
5:22 - 5:25Now before we go on, let me describe
very quickly and simply, -
5:25 - 5:26how RKCP works.
-
5:27 - 5:31So RKCP is an algorithm
designed by Ray Kurzweil, -
5:31 - 5:33who's a director of engineering at Google
-
5:33 - 5:35and a firm believer
in artificial intelligence. -
5:36 - 5:39So, you give RKCP a source text,
-
5:39 - 5:44it analyzes the source text in order
to find out how it uses language, -
5:44 - 5:46and then it regenerates language
-
5:46 - 5:48that emulates that first text.
-
5:48 - 5:51So in the poem we just saw before,
-
5:51 - 5:53Poem 2, the one that you all
thought was human, -
5:53 - 5:54it was fed a bunch of poems
-
5:54 - 5:56by a poet called Emily Dickinson
-
5:56 - 5:59and looked at the way she used language,
-
5:59 - 6:00learned the model,
-
6:00 - 6:04and then it regenerated a model
according to that same structure. -
6:06 - 6:08But the important thing to know about RKCP
-
6:08 - 6:11is that it doesn't know the meaning
of the words it's using. -
6:12 - 6:13The language is just raw material,
-
6:13 - 6:15it could be Chinese,
it could be in Swedish, -
6:15 - 6:20it could be the collected language
from your Facebook feed for one day. -
6:20 - 6:22It's just raw material.
-
6:22 - 6:25And nevertheless, it's able
to create a poem -
6:25 - 6:28that seems more human
than Gertrude Stein's poem, -
6:28 - 6:30and Gertrude Stein is a human.
-
6:31 - 6:35So what we've done here is,
more or less, a reverse Turing test. -
6:36 - 6:41So Gertrude Stein, who's a human,
is able to write a poem -
6:41 - 6:44that fools a majority
of human judges into thinking -
6:44 - 6:47that it was written by a computer.
-
6:47 - 6:50Therefore, according to the logic
of the reverse Turing test, -
6:50 - 6:52Gertrude Stein is a computer.
-
6:52 - 6:54(Laughter)
-
6:55 - 6:56Feeling confused?
-
6:56 - 6:58I think that's fair enough.
-
6:58 - 7:02So far we've had humans
that write like humans, -
7:02 - 7:05we have computers that write
like computers, -
7:05 - 7:08we have computers that write like humans,
-
7:08 - 7:11but we also have,
perhaps most confusingly, -
7:13 - 7:15humans that write like computers.
-
7:16 - 7:18So what do we take from all of this?
-
7:19 - 7:21Do we take that William Blake
is somehow more of a human -
7:21 - 7:23than Gertrude Stein?
-
7:23 - 7:26Or that Gertrude Stein is more
of a computer than William Blake? -
7:26 - 7:27(Laughter)
-
7:27 - 7:29These are questions
I've been asking myself -
7:29 - 7:31for around two years now,
-
7:31 - 7:33and I don't have any answers.
-
7:33 - 7:35But what I do have are a bunch of insights
-
7:35 - 7:38about our relationship with technology.
-
7:39 - 7:42So my first insight is that,
for some reason, -
7:43 - 7:46we associate poetry with being human.
-
7:46 - 7:49So that when we ask,
"Can a computer write poetry?" -
7:49 - 7:51we're also asking,
-
7:51 - 7:53"What does it mean to be human
-
7:53 - 7:55and how do we put boundaries
around this category? -
7:56 - 7:59How do we say who or what
can be part of this category?" -
7:59 - 8:03This is an essentially
philosophical question, I believe, -
8:03 - 8:05and it can't be answered
with a yes or no test, -
8:05 - 8:06like the Turing test.
-
8:07 - 8:10I also believe that Alan Turing
understood this, -
8:10 - 8:13and that when he devised
his test back in 1950, -
8:13 - 8:16he was doing it
as a philosophical provocation. -
8:18 - 8:23So my second insight is that,
when we take the Turing test for poetry, -
8:23 - 8:26we're not really testing
the capacity of the computers -
8:26 - 8:29because poetry-generating algorithms,
-
8:29 - 8:34they're pretty simple and have existed,
more or less, since the 1950s. -
8:35 - 8:38What we are doing with the Turing
test for poetry, rather, -
8:38 - 8:43is collecting opinions about what
constitutes humanness. -
8:44 - 8:46So, what I've figured out,
-
8:46 - 8:49we've seen this when earlier today,
-
8:49 - 8:52we saw that William Blake
is more of a human -
8:52 - 8:53than Gertrude Stein.
-
8:53 - 8:56Of course, this doesn't mean
that William Blake -
8:56 - 8:57was actually more human
-
8:57 - 9:00or that Gertrude Stein
was more of a computer. -
9:00 - 9:05It simply means that the category
of the human is unstable. -
9:06 - 9:08This has led me to understand
-
9:08 - 9:11that the human is not a cold, hard fact.
-
9:11 - 9:14Rather, it is something
that's constructed with our opinions -
9:14 - 9:17and something that changes over time.
-
9:17 - 9:22That is to say, the category of the human
is unstable. -
9:23 - 9:27So my final insight is that
the computer, more or less, -
9:27 - 9:31works like a mirror
that reflects any idea of a human -
9:31 - 9:32that we show it.
-
9:33 - 9:34We show it Emily Dickinson,
-
9:34 - 9:37it gives Emily Dickinson back to us.
-
9:37 - 9:39We show it William Blake,
-
9:39 - 9:41that's what it reflects back to us.
-
9:41 - 9:43We show it Gertrude Stein,
-
9:43 - 9:45what we get back is Gertrude Stein.
-
9:46 - 9:48More than any other bit of technology,
-
9:48 - 9:54the computer is a mirror that reflects
any idea of the human we teach it. -
9:55 - 9:57So I'm sure a lot of you have been hearing
-
9:57 - 10:00a lot about artificial
intelligence recently. -
10:01 - 10:04And much of the conversation is kind of,
-
10:05 - 10:06Can we build it?
-
10:07 - 10:10Can we build an intelligent computer?
-
10:10 - 10:12Can we build a creative computer?
-
10:12 - 10:14What we seem to be asking over and over
-
10:14 - 10:17is can we build a human-like computer?
-
10:18 - 10:19But what we've seen just now
-
10:19 - 10:22is that the human
is not a scientific fact, -
10:22 - 10:25that it's an ever-shifting,
concatenating idea -
10:25 - 10:28and one that changes over time.
-
10:28 - 10:31So that when we begin
to grapple with the ideas -
10:31 - 10:33of artificial intelligence in the future,
-
10:33 - 10:35we shouldn't only be asking ourselves,
-
10:35 - 10:37"Can we build it?"
-
10:37 - 10:39But we should also be asking ourselves,
-
10:39 - 10:42"What idea of the human
do we want to have reflected back to us?" -
10:42 - 10:45This is an essentially philosophical idea,
-
10:45 - 10:48and it's one that can't be answered
with software alone, -
10:48 - 10:53but I think requires a moment
of species-wide, existential reflection. -
10:53 - 10:54Thank you.
-
10:54 - 10:57(Applause)
- Title:
- Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney
- Description:
-
If you read a poem and feel moved by it, but then find out it was actually written by a computer, would you feel differently about the experience? Would you think that the computer had expressed itself and been creative, or would you feel like you had fallen for a cheap trick? In this talk, writer Oscar Schwartz examines why we react so strongly to the idea of a computer writing poetry -- and how this reaction helps us understand what it means to be human.
--
Poetry test #1
Poem 1
Little Fly
Thy summer’s play,
My thoughtless hand
Has bush’d away.Am not I
A fly like thee?
Or art not thou
A man like me?Poem 2
We can feel
Activist through your life’s
Morning
Pauses to see, pope I hate the
Non all the night to start a
great otherwiseI’ll snake swirling
Vastness guess
Totally mental hamsters if I
Know I put on a year a crucial
Absolutely.Poetry test #2
Poem 1
A lion roars and a dog barks. It is interesting
and fascinating that a bird will fly and not
roar or bark. Enthralling stories about animals
are in my dreams and I will sing them all if I
am not exhausted and weary.Poem 2
Oh! kangaroos, sequins, chocolate sodas!
You really are beautiful! Pearls,
harmonicas, jujubes, aspirins! All
the stuff they’ve always talked aboutstill makes a poem a surprise!
These things are with us every day
even on beachheads and biers. They
do have meaning. They’re strong as rocks.Poetry test #3
Poem 1
Red flags the reason for pretty flags.
And ribbons.
Ribbons of flags
And wearing material
Reason for wearing material.
Give pleasure.
Can you give me the regions.
The regions and the land.
The regions and wheels.
All wheels are perfect.
Enthusiasm.Poem 2
A wounded deer leaps highest,
I’ve heard the daffodil
I’ve heard the flag to-day
I’ve heard the hunter tell;
‘Tis but the ecstasy of death,
And then the brake is almost done,
And sunrise grows so near
sunrise grows so near
That we can touch the despair and
frenzied hope of all the ages.This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at http://ted.com/tedx
- Video Language:
- English
- Team:
- closed TED
- Project:
- TEDxTalks
- Duration:
- 11:04
Maricene Crus commented on English subtitles for Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney | ||
TED Translators admin edited English subtitles for Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney | ||
TED Translators admin edited English subtitles for Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney | ||
TED Translators admin edited English subtitles for Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney | ||
TED Translators admin edited English subtitles for Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney | ||
TED Translators admin edited English subtitles for Can a computer write poetry? | Oscar Schwartz | TEDxYouth@Sydney |
Maricene Crus
3:12.06
to have a to have a text-based → to have a text-based
Thank you!