Cornell University Law School
Scholarship@Cornell Law: A Digital Repository
02/&--"6"$5-475#-*$"4*0/3 "$5-47$)0-"23)*1

ere's No Such ing as a Computer-Authored
Work - And It's a Good ing, Too
James Grimmelmann
Cornell Law School+".&3(2*..&-."//$02/&--&%5
0--064)*3"/%"%%*4*0/"-602,3"4 ):13$)0-"23)*1-"6$02/&--&%5'"$15#
"240'4)& 0.154&2"60..0/3"/%4)& /4&--&$45"-201&247"60..0/3
9*324*$-&*3#205()440705'02'2&&"/%01&/"$$&33#74)&"$5-47$)0-"23)*1"4$)0-"23)*102/&--"6*(*4"-&103*40274)"3#&&/
"$$&14&%'02*/$-53*0/*/02/&--"6"$5-475#-*$"4*0/3#7"/"54)02*8&%"%.*/*342"4020'$)0-"23)*102/&--"6*(*4"-&103*402702
.02&*/'02."4*0/1-&"3&$0/4"$4 +.1$02/&--&%5
&$0..&/%&%*4"4*0/
".&32*..&-."//9&2&305$)9*/("3"0.154&254)02&%!02,/%43"00%9*/( 000-5.#*"052/"-0'
"64)&243
There's
No
Such
Thing
as
a
Computer-Authored
Work-
And
It's
a
Good
Thing,
Too*
James
Grimmelmann**
INTRODUCTION
I
would
like
to
talk
about
computer-authored
works-I
would
like
to,
except
that
they
don't
exist.
Copyright
law
doesn't
recognize
computer
programs
as
authors,
and
it
shouldn't.'
Some
day
it
might
make
sense
to,
but
if
that
day
ever
comes,
copyright
will
be the
least
of
our
concerns.
Instead,
I
will
say
something
about
why
computer
authorship
is
such
a
"bad
penny
of
a
question,"
to
use
Annemarie
Bridy's
felicitous
phrase,
even
though
it
is
so
utterly
speculative.
2
The
scholarship
pondering
the
possibility
of
computer-
authored
works
is
surprisingly
extensive,
even
though
no one
has
ever
exhibited
even
one
work
that
could
plausibly
claim
to have a
computer
for
an
"author"
in
the
sense
that the
Copyright
Act
uses
the
term.
3
*
This
Essay
is
based
on
a
talk
that
was
given
on
October
2,
2015,
at
the
Kernochan
Center
Annual
Symposium
at
Columbia
Law
School.
**
Professor
of
Law,
University
of
Maryland.
My
thanks
to
Aislinn Black,
Annemarie Bridy,
Bruce
Boyden,
Timothy
McGovern,
Christina
Spiesel,
and
the
participants
in
the
Kernochan
Center
Symposium "Copyright
Outside
the
Box"
for
their
suggestions.
This
Essay
may
be
freely
reused
under
the
terms
of
the
Creative Commons Attribution
4.0
International
license,
https://creativecommons.org/licenses/by/4.0/.
1.
See
U.S.
COPYRIGHT
OFFICE, COMPENDIUM
OF
U.S.
COPYRIGHT
OFFICE
PRACTICES
(THIRD)
§
313.2
(3d ed.
2014)
("Similarly,
the
Office
will
not
register
works
produced
by
a
machine
or
mere
mechanical process
that
operates
randomly
or
automatically
without
any
creative
input
or
intervention
from
a
human
author.")
2.
Annemarie
Bridy,
Coding
Creativity:
Copyright
and
the
Artificially Intelligent
Author,
5
STAN.
TECH.
L.
REv.
1,
1
52
(2012).
3.
Bridy,
supra
note
2,
at
¶¶
50-52;
Timothy
L.
Butler,
Can
a
Computer
Be
an
Author
-
Copyright
Aspects
of
Artificial
Intelligence,
4
COMM/ENT
L.S.
707,
739-42
(1982);
Ralph
D.
Clifford,
Intellectual
Property
in
the
Era
of
the
Creative
Computer
Program:
Will
the
True
Creator
Please
Stand
up,
71
TUL.
L.
REV.
1675,
1685-86,
1694-95
(1996); Evan
H.
Farr,
Copyrightability
of
Computer-
Created
Works,
15 RUTGERS
COMPUTER
&
TECH.
L.J.
63,
79
(1989);
Dane
E.
Johnson,
Statute
ofAnne-
Imals:
Should
Copyright
Protect
Sentient Non-Human
Creators,
15
ANIMAL
L.
15,
19-21
(2008);
Karl
F.
Milde,
Jr.,
Can
a
Computer
Be
an
"Author"
or
an
"Inventor"?,
51
J.
PAT.
OFF.
Soc'Y
378,
392-95
(1969);
Arthur
R.
Miller,
Copyright
Protection
for
Computer Programs, Databases,
and
Computer-
Generated
Works:
Is
Anything
New
Since
CONTU?,
HARV.
L.
REV.
977,
1056-72
(1993);
William
T.
Ralston,
Copyright
in
Computer-Composed
Music:
HAL
Meets
Handel,
52
J.
COPYRIGHT
SOC'Y
U.S.A.
281,
302-3
(2004-05); Pamela
Samuelson,
Allocating
Ownership Rights
in
Computer-Generated
Works,
47
U. PITT.
L.
REV.
1185,
1192-1200
(1986);
Andrew
J.
Wu,
From
Video
Games
to
Artificial
Intelligence:
Assigning Copyright
Ownership
to
Works
Generated
by
Increasingly
Sophisticated
Computer
Programs,
25
AIPLA
Q.
J.
131,
155-57
(1997).
Most
of
these
authors
sensibly
conclude
that
computers
are
not authors,
for
now
at
least,
but
it
403
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
My
view
is
that
the
idea
of
computer-authored
works
responds
to
perceived
problems
of
deciding
who
should
be
considered
the authors
of
computer-generated
works.
The
difficulty
of
saying
who is
the
author
of
a
work
when
one
person
runs
an
unpredictably
complicated
computer
program
written
by
another
makes
it
tempting
to say
that
the
program
itself
is
the
author.
But
this
suggestion
is
unnecessary, because
the
underlying
problems
of
assigning
authorship
are
more
apparent
then
real-or
rather,
they
are
no
worse
here
than
elsewhere
in
copyright.
I
can
imagine
five
reasons
why
computer-generated
works
might
be
considered
meaningfully different
from
human-generated
works:
4
(1)
They
are
embedded
in
digital
copies.
(2)
People
create
them
using
computers
rather
than by
hand.
(3)
Programs
can
generate
them
algorithmically.
(4)
Programmers
as
well
as
users
contribute to
them.
(5)
Programs
can
generate
them
non-deterministically.
All
of
these
distinctions
are
spurious.
Old-fashioned
pen-and-paper
works
raise
all
of
the
same
issues;
there
is
nothing
new
under
the
sun.
5
These issues
have
been
with
us
since
the
Statute
of
Anne;
they
will
be with
us
as
long
as
copyright
considers
it
important
to
assign
ownership
of
a
work
to
an
"author"
who
is
causally
responsible
for
the
work's
existence.
I.
DIGITAL
COPIES
What
might
we
mean
by
a
"computer-generated
work"
in
the
first place?
The
simplest answer
is
that
since
a
computer-generated work
is
at
some
point
emitted
by
a
computer,
it
exists
in
digital
copies, as
contrasted
with traditional
works
that
exist
in
analog
copies.
A
book
is
a
physical
object
that
exists
in
the
physical
world:
it
consists
of
patterns
of
ink
on
rectangular sheets
of
paper bound
together
into
a
codex.
A
sculpture
is
a
hunk
of
marble
with
physical
properties
(most
obviously,
mass).
A
painting
is
pigment
on
canvas;
canvas
has
a
texture
and
paint
has
a
thickness. But
a
digital
copy,
one
might
say,
is
just
ones
and
zeroes;
it
exists
entirely
as
an
abstract
mathematical representation.
This
is
not
even
a
fair
characterization
of
digital
copies.
They
are
still
"copies"
is
striking
how
often
the
question has
been
asked.
The
supposed
problem
has
also
been
much
reported
on
by
experts.
COPYRIGHT
OFFICE,
SIxTY-EIGHTH
ANNUAL
REPORT
OF
THE
REGISTER
OF
COPYRIGHTS
5
(1966);
COMM'N
ON
NEW
TECH.
USES,
FINAL
REPORT
OF
THE
NAT'L
COMM'N
ON
NEW
TECH.
USES
OF COPYRIGHTED
WORKS
43-46
(1978);
OFFICE
OF
TECH.
ASSESSMENT,
INTELLECTUAL
PROP.
RIGHTS
IN
AN
AGE
OF
ELECS.
AND
INFO.
69-73
(1986);
OFFICE
OF TECH.
ASSESSMENT,
FINDING
A
BALANCE:
COMPUTER SOFTWARE,
INTELLECTUAL PROP.
AND
THE
CHALLENGE
OF TECHNOLOGICAL CHANGE
112-
13
(1992).
4.
This
is
not
to
say
that these
reasons
are
commonly
advanced
by
copyright
commentators.
The
first
two
in
particular
are
straw
men.
But
there
is
value
in
knowing
not
just
that
they
fail,
but
why.
5.
But
see
Bruce
Boyden,
Emergent
Works,
39
COLUM
J.L.
&
ARTS
377
(2016) (arguing
that
works
generated
by
unpredictable
computer
programs
raise
authorship
issues
that
are
genuinely
different
in kind);
Annemarie Bridy,
The
Evolution
ofAuthorship:
Work
Made
by
Code,
39
COLUM
J.L.
&
ARTS
395
(2016)
(giving
interesting
and
challenging
examples
of
such
programs
and
the
works
they
have
generated).
404
[39:3
2016]
THERE'S
No
SUCH
THING
AS
A
COMPUTER-AUTHORED
WORK
in
the
copyright
sense:
"material
objects
. . .
in
which
a
work
is
fixed
by
any
method
now
known
or
later
developed,
and
from
which
the work
can
be
perceived,
reproduced,
or
otherwise
communicated,
either
directly
or
with
the
aid
of
a
machine
or
device."
6
This
definition
obliterates
any
doctrinal
distinction
between
digital
and
analog
copies.
Bits
may
be
abstract and
intangible,
but
memory
chips
and
hard
drives
are
very much
"material
objects."
Digital
copies
are
not
second-
class
citizens
of
copyright's
domain.
Of
course,
a
work
is
stored
differently
in
a
digital
copy than
in
an
analog
one.
The
text
in
a
book
is
legible
to
a
reader
who
picks
it
up
and
flips
through
the
pages,
while
the
pattern
of
ones
and
zeroes
on
a
hard
drive
is
invisible
to the
naked
eye,
even when
those ones and
zeroes
encode
the
very
same
novel.
But Congress
quite
rightly
recognized
that what matters
is
whether
the
work
"can
be
perceived"
from
a
copy
(possibly
with
"the
aid
of
a
machine
or
device"),
not
whether
it
is
directly
perceptible
when
the
copy
is
sitting
inert
on
the shelf.
The
text
of
a
book
is
also
invisible
to
humans
unless
they
take
it
down
from
a
shelf
and
look
between
its
covers.
Even
if
there
were
a
relevant
difference
between
digital
and
analog
copies
of
a
work,
it
would
be
a
property
of
the
copies,
not
a
property
of
the
work
itself.
Any
"categor[y]"
of
"authorship" cognizable
under
the
Copyright
Act
can
be
fixed
in
digital
copies
with
a
suitable
encoding.
8
Literary
works
are
made
up
of
letters,
numbers,
and
other
characters.
Those characters can
be
translated
into
binary using
standard
encodings
such
as
UTF-8.
Pictorial works
can
be
broken
down
into
pixels, and
each
pixel broken
down into
the
intensities
of
its
constituent
colors.
The
pitches
and
durations
of
notes
in
a
musical
work can
be
represented
by
numbers,
which
in
turn
can
be
represented
in
binary.
For
musical
works
with
unusual
features,
9
an
alternative
is
to
write
the
score
out
on
a
piece
of
paper
and
digitize
the image
of
the
paper,
pixel
by
pixel.
This last
trick
shows
why
the
technique
works
for
any
conceivable
work
of
authorship
protected
by
the
Copyright
Act.
"Copyright
protection
subsists
. .
.
[only]
in
works
of
authorship
fixed
in
any
tangible
medium
of
expression."
10
Any
work
subject
to
copyright
must
therefore
exist
in
some
"tangible
medium."
That's
enough.
Find
the "material
object"
in which
the
work
is
fixed
and
digitize
that
object
with enough
detail
to
capture
whatever
aspect
of
the
object
embodies
the
work.
Architectural
works
exist
in
blueprints;
find
a
copy
and
take
a
high-
resolution
photograph.
Sculptural
works
exist
in
sculptures;
find
one
and take
a
6.
17
U.S.C.
§
101
(2012).
The
phrase
"any
method
now known
or later
developed"
makes
emphatically
clear
Congress's
insistence on technological
neutrality
and
its
rejection
of
"artificial
and
largely
unjustifiable
distinctions
.
. .
under
which
statutory
copyrightability
in
certain cases
has
been
made
to
depend
upon
the
form
or
medium
in
which
the
work
is
fixed."
H.R. Rep.
No.
94-1476,
at
51
(1976).
The point
is
reiterated
in
§
102(a),
which
protects
works "fixed
in
any
tangible medium
of
expression,
now
known
or later
developed."
(emphasis
added).
7.
17
U.S.C.
§
101
(2012)
(emphasis
added).
8.
17
U.S.C.
§
102(a)
(2012).
9.
E.g.,
John
Stump,
Faerie's
Aire
and
Death
Waltz,
http://farml.static.flickr.com/148/
406258694 e65219a08c
o.jpg [https://perma.cc/9NLR-Z5MY]
(last
visited
Mar.
12,
2016).
10.
17
U.S.C.
§
102(a)
(2012).
405
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
high-resolution
three-dimensional
laser
scan.
Vinyl
purists
and
other
analog
aesthetes
may
insist
that
analog
copies
capture
nuances
that
no digital
copy
ever
can.
They
are
wrong.
Human
perception
is
limited
by
the
capacities
of
our
senses.
The
human
eye
cannot
see
details
finer
than
about
thirty
arc-seconds.
Digitize
a
painting
at
a
higher
resolution
than
a
person
with
her
nose
pressed
up against
it
can
distinguish,
and
the
resulting
copy
is
effectively
perfect,
as
far
as
she
can
tell.
The
human
ear cannot
hear
sounds
at
a
frequency
greater
than
about
twenty
kilohertz;
the
CD
audio
standard
plays
it
safe
and
goes
up
to
22.05
kHz.
An
audiophile
who
claims
she
can
hear
more
on
a
vinyl
recording
is
right
only
in
the
sense
that
record
players
add
hisses
and
pops,
or,
as
non-audiophiles
call
them,
noise.
Of
course,
some
works
exist
in
digital
copies
and
others
do
not.
But
the
works
that
do
not
might
as
well
have.
Nothing
important
about
their
status
as
copyrightable
works
would
be
lost
by
digitizing
them-and
nothing
important
would
be
gained,
either.
Digital
copies
are
convenient
for
computers
to
work
with,
and
they
can
be
remarkably
convenient
for
bringing
works
to
audiences.
(Too
convenient,
some
would
say.)
But
the
contingent
fact
of
whether
a
work
exists
in
digital
copies,
analog
copies,
or
both
does
not
affect
its
status
as
a
work
of
authorship:
the
expression
it
contains
is
exactly
the
same
"regardless
of
the
nature
of
the
material
objects
. . .
in
which
[it
is]
embodied.""
II.
DIGITAL
WORKS
The
next-simplest
answer
to
what
counts
as
a
computer-generated
work
is
one
that
was
created
using
a
computer.
Some
poets
write
longhand;
others
type.
Again,
we
might
posit
that
there
would
be
some
relevant
differences,
as
a
class,
between
works
using
computers
and
works
created
entirely-without
computers.
Again,
we
would
be
wrong.
Every
kind
of
copyrightable
work
can
be-and
regularly
is-created
using
computers.
Take
the
list
of
categories
in
§
102(a).
There
is
specialized
software
to
create
every single
one
of
them:
(1)
"literary
works":
Microsoft
Word,
Apple
Pages,
Scrivener
(2)
"musical
works":
Finale,
Sibelius,
Noteflight
(3)
"dramatic
works":
Microsoft
Word
again,
Final
Draft
(4)
"pantomimes
and
choreographic
works":
Dance
Designer,
DanceForms
(5)
"pictorial,
graphic,
and
sculptural
works":
Adobe
Photoshop,
Photo
Booth,
Adobe
Illustrator,
Affinity
Designer,
SketchUp,
AutoCAD
(6)
"motion
pictures
and
other
audiovisual
works":
Final
Cut
Pro,
Maya,
Adobe
Flash
(7)
"sound
recordings":
Garage
Band,
Pro
Tools
(8)
"architectural
works":
AutoCAD
again,
Chief
Architect,
SoftPlanl
2
11.
17
U.S.C.
§
101
(2012).
12.
17
U.S.C.
§
102(a)
(2012).
406
[39:3
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
world,
the
digital
creative
process
yields
a
digital
copy
on a
computer.
This
is
not
something
new
about
digitally
created
works.
It
merely
makes
explicit
a
causal
relationship
that
has always
inhered
in
creativity:
a
work,
as
fixed
in
a
copy,
is
the
result
of
what
an
author
does.
III.
ALGORITHMIC
CREATION
A
third
way
in
which
computer-generated
works
might
be
thought
to
be
uniquely
challenging
for
copyright
is
that
they
involve
algorithmic
creation.
Rather
than
being
produced
intuitively
through
a
process
whose
details
are
not
available
for
inspection,
they
are
the
outputs
of
a
process
whose
steps
are
precise
and
explicit.
This
is
not
a
distinction
that
ought
to
make
a
difference.
True,
rote
rule-
following
does
not
look
like
copyrightable
creativity;
"slavish
copying"
is
not
originality.'
3
But
creativity
can
also
inhere
in
a
creator's
selection
of
the
rules
she
will
follow.
She
can
introduce
copyrightable
authorship
through
her
choice
of
rules
or
through
her
choice
among
variations
permitted
by
the
rules,
and
as
a
general
matter
these
two
kinds
of
authorship
are
equivalent.
A
work
does
not
know
whether
it
was
created
freehand
or
by
following
rules.
The
use
of
rules
at
all
is
simply
the
choice
to
split
the
creative
process
into
two
stages
rather
than
one.
The
inputs-whatever
it
is
that
we
mean
by
"creativity"
or
"expression"
or
"authorship"-are
indistinguishable,
and
so
is
the
output-a
fixed
copy
of
the
work.
If
an
author,
for
her
own
convenience,
decides
to
automate
some
of
the
steps
by
programming
a
computer,
copyright
should
not
look
any
less
generously
upon
her.
Copyright
does
not turn
up
its
nose
at
printmakers
who
engrave
a
single
copper
plate
to
make
an
edition
of
500
prints,
or
on
novelists
who
use
movable
type
rather
than
writing
each
letter
by
hand.
Their
creative
work
is
substantially
complete
by
the
time
they
ink
a
plate
or
fish
sorts
out
of
a
type
case;
to
complain
that
these
steps
are
too
mechanical
to
support
a copyright
is
to
go
looking
for authorship
in
all
the
wrong
places.
This
remains
true
even
when
an
author
acts
without
conscious
attention
to
each
detail;
it
is
not
that her
hand
is
unguided,
but
rather
that
it
is
guided
by
neuronal
firings
in
her
brain
in a
way
she
does
not
consciously
attempt
to
direct.
In
this
sense,
all
authorship
is
algorithmic.
To
say
that
an
author
creates
intuitively
is
simply
to
say
that
neither
she
nor
we
have
ready
access
to
the
algorithm
she
follows.
14
Indeed,
the
distinction
between
an
algorithm
and
its
output
is
often
irrelevant
for
copyright
purposes.
One
of
the
most
fundamental
insights
of
computer
science
is
the
idea
that
a
program-software-is
just
another
form
of
data.
15
That's
what
13.
Bridgeman
Art Library,
Ltd.
v.
Corel
Corp.,
25
F.
Supp.
2d
421,
426-27
(2d
Cir.
1998).
14.
See
generally
Bridy,
supra
note
2;
Alan
L.
Durham,
The
Random
Muse:
Authorship
and
Indeterminacy,
44
WM.
&
MARY L.
REv.
569
(2002).
15.
Alan
Turing
made
this
point
in
the
article
that
created
the
field
of
computer
science;
he
gave
an
explicit
construction
for
writing
out
any
possible
program
as
a
string
of
symbols.
See
A.M.
Turing,
408
[39:3
409
2016]
THERE'S
NO
SUCH
THING
AS
A
COMPUTER-AUTHORED
WORK
makes
general-purpose
computers
possible;
we
can
load
new
programs
on
them
in
the
same
simple
way
we
load
new
data.
Take
an
example:
Microsoft
Word
is
a
program.
It
is
represented
digitally:
as
a
sequence
of
ones
and
zeroes.
Those
bits
are
data;
you
install
that
data
from
a
CD
or
by
downloading
it.
In
turn,
your
computer
interprets
that
data
a
sequence
of
instructions,
executing
them
one
by
one
and
thereby
causing
Word
to
do
what
it
does.
In
the
same
way,
a
Word
document
is
also
data:
a
digital
sequence
of
bits
stored
on
a
hard
drive
somewhere.
Now
for
the
crucial
move:
a
Word
document
is
also
a
program,
one
that
tells
Word
itself
what
characters
to
display
and
how
to
format
them.
Your
computer
does
not
draw
any
fundamental
distinction
between
the
bits
that
make
up
Word
and
the
bits
that
make
up
a
Word
document.
Both
are
data;
both
are
programs.
Bits
are
bits.
We
could
regard
an
MP3
file
as
a
sequence
of
sound
data,
or
we
could
regard
it
as
a
program
for
producing
sound.
Both
descriptions
are
correct;
every
digital
fixation
of
a
work
is
a
program
for
generating
that
work.
And
since
every
work
can
be
fixed
digitally,
every
work
can
be
generated
by
a
program.
The
point
is
even
more
apparent
when
we
think
about
the
actual
creative
process.
Consider
again
the
sequence
of
steps
that
analog
Annelise
and
digital
Dennis
follow
to
create
pictorial
works.
Annelise's
creative
acts
look
intuitive
and
inscrutable,
while
Dennis's
creative
acts
leave
behind
a
digital
trail:
the
precise
sequence
of
steps
that
he
followed,
as
recorded
by
Adobe
Illustrator.
It
keeps
track
of
them
in
case
Dennis
hits
"undo";
the
only
way
to
enable
him
to
retrace
his
steps
is
to
leave
binary
breadcrumbs
along
the
route
of
his
creative
journey.
But
now
suppose
that
Dennis,
having
repeatedly
hit
"undo"
until
he
is
back
at
a
blank
canvas,
decides
he
likes
what
he
made
after
all,
and
starts
hitting
"redo"
until
he
has
the
finished
illustration
again.
In
other
words,
Illustrator's
saved
undo/redo
information
provides
a
precise
set
of
instructions
for
recreating
Dennis's
work
from
scratch,
one
that
causes
Illustrator
to
repeat
what
it
did
when
Dennis
was
driving-
resulting
in
exactly
the
same
image.
Dennis's
sequence
of
creative
acts
is
a
program
for
generating
the
work,
and
it
is
also
a
copy
of
the
work,
which
means
that
so
are
Annelise's
creative
acts.
All
creativity
is
also
algorithmic
in
the
sense
that
we
could
encode
the
work
as
a
program
making
completely
explicit
what
the
creator
did
to
produce
it.
IV.
SEQUENTIAL
CREATION
The
next
potential
challenge
posed
by
computer-generated
works
is
that
people
frequently
use
computer
programs
written
by
others,
which
raises
the
question
of
how
to
allocate
ownership
of
the
resulting
works
between
programmer
and
user.
It
is
easy
to
give
examples
where
one
or
the
other
has
a
better
claim.
On
the
one
hand,
where
the
program
is
Finale
and
the
work
is
a
string
quartet,
the
user
is
the
author
of
that
musical
work.
Finale's
programmers
have
no
more
of
a
copyright
On
Computable
Numbers,
with
an
Application
to
the
Entscheidungsproblem,
42
PROC.
LONDON
MATH.
Soc.
230
(1936).
Turing's
insight
is
all
the
more
remarkable
for
having
been
formulated
before
actual
digital
computers
existed
to
run
those
programs.
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
claim
than
the
penmaker
would
if
a
composer
had
written
the
quartet
out
by
hand.
On
the
other
hand,
where
the
program
displays
a
fifteen-second
animation
of
fireworks
whenever
the
user
pushes
the
space
bar,
the
programmer
is
the
author
of
that
audiovisual
work.
The
user
has
no
copyright
claim
because
"the
program
would
have
generated
the
same
output
no
matter
which
human
user
caused
the
output
to
be
generated."
6
It
seems,
therefore,
that
the
problem
is
to
distinguish
computer
users
who
are
genuine
authors
from
users
who
merely
push
a
button.
But
this
is
not
a
problem
unique
to
computers.
Even
in
the
offline
world,
there
are
buttons
and
there
are
buttons.
The
user
who
pushes
a
button
on
a
music
box
to
start
it
playing
is
not
an
author;
the
user
who
pushes
a
button
on
a
camera
to
take
a
photograph
is.
Trying
to
allocate
copyrights
between
computer
programmer
and
computer
user
is
the
same
kind
of
task
as
trying
to
allocate
them
between
thing-maker
and
thing-user.
These
are
problems
of
sequential
creation:
A
and
B
both
make
contributions
toward
a
work
over
time,
one
after
the
other.
Who
owns
the
work?
It
depends.
There
are
at
least
six
possible
legal
outcomes:
Infringing
copy:
If
B
makes
only
de
minimis
modifications
to
A's
work,
then
B
is
not
an
author
in
the
copyright
sense.
Copyright
in
the
resulting
work
continues
to
be
owned
solely
by
A.
If
I
take
your
short
story,
format
it
entirely
in
small
caps,
and
then
hit
"Save
As"
to
store
a
copy,
I
am
an
infringer,
not
an
author.
Unlawful
derivative
work:
If
B
modifies
A's
work
enough
to
contribute
copyrightable
authorship,
then
B
is
the
author
of
a
derivative
work.
The
copyright
status
of
this
work
depends
on
whether
B
has
A's
permission,
or
makes
a
transformative
fair
use,
or
has
some
other
statutory
authorization
to
prepare
a
derivative
work.
17
If
B
uses
A's
work
"unlawfully,"
then
A
owns
the
copyright
in
the
portions
copied
from
her
preexisting
work,
while
B's
incremental
contributions
are
uncopyrightable.
If
I
do
substantial
editorial
work
on
your
short
story
without
your
permission-rearranging
scenes,
adding
new
expository
dialogue,
and
eliminating
an
unnecessary
character-
neither
of
us
may
be
able
to
claim
a
copyright
in
these
tweaks.
Lawful
derivative
work:
If,
on
the
other
hand,
B
creates
a
derivative
work
but
does
so
"[]lawfilly,"
then
A
owns
the
copyright
in
the
portions
copied
from
her
preexisting
work,
while
B
owns
the
copyright
in
her
incremental
contributions.
If
you
gave
me
permission
to
revise
your
story
and
copyright
the
revisions-or
if
my
changes
so
alter
the
story's
meaning
as
to
make
it
a
biting
commentary
on
the
original-I
will
own
a
copyright
in
my
brilliant
revisions
and
you
will
own
a
copyright
in
your
inferior
original.
Joint
work:
Another
possibility
in
the
case
where
the
resulting
work
reflects
both
A
and
B's
contributions
is
that
they
mutually
intend
that
their
contributions
"be
merged
into
inseparable
or
interdependent
parts
of
a
unitary
whole,"
1
8
making
them
the
co-
16.
Samuelson,
supra
note
3,
at
1206-07.
17.
17
U.S.C.
§
103(a)
(2012).
18.
17
U.S.C.
§
101
(2012).
410
[39:3
411
2016]
THERE'S
No
SUCH
THING
AS
A
COMPUTER-AUTHORED
WORK
owners
of
a
joint
work.
19
If
you
and
I
are
bandmates,
you
lay
down
a
vocal
track,
and
I
then
add
a
drum
line
and
guitar
riffs,
the
resulting
sound
recording
is
most
likely
a
joint
work.
Sole-authored
work:
If
B
modifies
A's
work
or
discards
portions
of
it
to
the
point
that
the
resulting
work
is
no
longer
substantially
similar
to
A's,
and
B
also
makes
copyrightable
contributions
of
her
own,
then
B
is
the
sole
author
and
sole
copyright
owner
of
the
resulting
work.
If
I
write
a
screenplay
in
Microsoft
Word,
I
alone
own
the
copyright,
because
the
screenplay
is
not
substantially
similar
to
Word.
No
copyright:
Finally,
if
B
discards
A's
work
without
adding
anything
copyrightable
of
her
own,
the
output
simply
fails
to
be
a
copyrightable
work
at
all.
Neither
A
nor
B
is
a
copyright
owner;
anyone
is
free
to
copy
it.
If
I
open
up
Microsoft
Word
and
immediately
hit
"print,"
neither
Microsoft
nor
I
has
a copyright
in
what
is
on
the
page
that
emerges
from
the
printer,
because
the
page
will
be
blank.
These
six
possibilities
range
over
every
combination
of
possible
ownership
by
A
and
B.
The
resulting
work
could
be
owned
solely
by
A
(in
whole
or
part),
by
A
and
B
together
(jointly
or
in
separate
portions),
solely
by
B,
or
by
no
one.
The
problem
of
allocating
ownership
between
programmers
and
users
is
just
a
special
case
of
this
more
general
and
very
familiar
problem.
Both
contribute
to
the
final
form
of
the
work.
The
programmer
goes
first,
and
then
the
user,
but
more
than
that
we
cannot
say
without
knowing
about
the
particular
program
and
the
particular
user
input.
The
resulting
work
might
be
virtually
identical
to
the
program,
or
it
might
be
virtually
identical
to
the
user's
input,
or
it
might
be
similar
to
both,
but
identical
to
neither,
and
we
will
have
to
inspect
the
expression
in
the
program,
the
input,
and
the
work
to
say
for
sure
which
is
the
case.
The
user
might
arrange
segments
of
expression
supplied
by
the
programmer,
as
when
a
Garage
Band
user
splices
together
loops
and
samples
that
come
with
the
program.
The
user
might
apply
complicated
programmer-supplied
transformations
to
her
own
expression,
as
when
a
Photoshop
user
applies
filters
that
come
with
the
program.
The
user
might
provide
her
own
expressive
elements
and
stitch
them
together
using
an
underlying
framework
supplied
by
the
programmer,
as
when
an
Adobe
Flash
Professional
user
makes
an
animation.
The
fact
that
some
of
these
examples
of
computer-assisted
creation
are
algorithmic
as
well
as
sequential
still
does
not
make
them
genuinely
new
in
kind.
Offline
sequential
creation
can
also
be
algorithmic.
A
Spirograph
enables
one
person
to
make
a
design
according
to
a
pattern
specified
by
another.
So
do
dress
patterns,
recipes,
and,
for
that
matter,
blueprints.
Sometimes,
the
author
of
the
instructions
is
regarded
as
the
sole
author
of
the
output;
architects
own
copyright
in
built
buildings,
not
the
contractors
who
actually
do
the
construction
work.
Sometimes,
there
is
room
for
the
rule-follower
at
copyright's
table:
if
a
"distinguishable
variation"
is
enough
for
a
copyright
in
a
work
created
to
the
plan
of
another,
I
own
the
paint-by-numbers
canvas
I
fill
in,
even
if
I
only
color
inside
19.
17
U.S.C.
§
201(a)
(2012).
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
the
lines.
2
0
We
have
not
thought
that
these
cases
raise
distinctive
and
difficult
issues
of
the
sort
that
would
lead
us
to
treat
a
Spirograph
disc
as
the
"author"
of
the
resulting
loops
and
whirls,
or
that
an
origami
pattern
owns
the
crane.
V.
NONDETERMINISTIC
CREATION
The
last
way
in
which
computer-generated
works
are
potentially
different
is
that
the
same
program
can
be
run
repeatedly
with
identical
inputs
to
generate
different
works.
21
Take
the
99-cent
"Mozart's
Dice
Game"
app
for
iOS
by
the
developer
Rodskagg.
22
Every
time
the
user
taps
the
"random"
button,
Mozart's
Dice
Game
strings
together
pseudorandomly-selected
measures
of
music
to
produce
a
new
piano
minuet
and
trio
in
a
late
19t
century
style.
It
is
unappealing
to
say
that
the
act
of
tapping
the
button
is
an
act
of
authorship
justifying
a
copyright
for
the
user.
But
the
fact
that
the
minuets
are
generated
pseudorandomly
also
makes
it
unappealing
to
treat
the
programmers
as
the
authors,
since
they
did
not
compose
any
particular
minuet.
This
again
is
a
hard
problem,
but
again
it
is
not
a
new
one.
The
"Mozart's
Dice
Game"
app
is
nothing
more
than
a
computer-implemented
version
of
an
actual
dice
game
for
composing
music.
This
Musikalisches
Wirfelspiel
was
published
in
1792
and
apocryphally
attributed
to
Mozart.
People
were
composing
music
using
nondeterministic
processes
long
before
there
were
computers;
there
is
a
rich
history
of
of
artistic
creation
incorporating
variable
elements.
2
3
Indeed,
there
is
a
rich
history
of
creation
that
is
sequential
and
algorithmic
as
well
as
nondeterministic:
the
people
"composing"
music
at
home
by
rolling
dice
were
doing
so
according
to
instructions
supplied
by
the
game's
designer.
It
is
helpful,
I
think,
to
look
at
how
copyright
deals
with
a
related
case:
the
one
in
which
the
same
person
both
creates
and
uses
the
nondeterministic
process.
Consider
a
composer
who
rolls
dice
to
choose
notes
according
to
a
scheme
of
her
own
devising.
24
She
is
a
perfectly
plausible
candidate
for
copyright
ownership,
for
three
possible
reasons.
One
is
to
say
that
she
is
an
author
because
of
her
creativity
in
defining
the
process
itself.
By
describing
a
mapping
from
the
die
rolls
to
musical
notes,
she
exercises
creative
control
over
some
aspects
of
the
composition;
20.
Alfred
Bell
&
Co.
v.
Catalda
Fine
Arts,
191
F.2d
99,
102
(2d
Cir.
1951).
21.
In
an
earlier
draft,
I
referred
to
"random"
rather
than
"nondeterministic"
creation.
But
see
Timothy
McGovern,
Copyright
Law
Would
Be
Easy
If
It
Wasn'tfor
All
the
Damn
Creativity,
HERDING
BATS
(Dec.
23,
2015),
http://herdingbats.blogspot.com/2015/1
2
/copyright-law-would-be-easy-if-it-
wasnt.html
(critiquing
my
usage
of
"random").
As
McGovern
explains:
Randomness
indicates
either
a
deterministic
process
occurring
at
a
scale
too
fine
to
measure,
or
a
provably
random
process
like
radioactive
decay,
where
the
distribution
of
discrete
events
can
be
probabilistically
predicted
over
time,
but
within
any
given
timeframe
their
occurrence
or
non-
occurrence
is
unpredictable
(cite:
I
asked
a
physicist!).
As
computer
scientists
know,
although
we're
working
on
randomness
generators,
basically
the
best
we
can
get
at
this
point
is
pseudorandomness.
22
Rodskagg,
"Mozart's
Dice
Game,"
http://www.dicegame.rodskagg.com/
[https://perna.cc/S49K-X499].
23.
See
generally
Durham,
supra
note
14.
24.
Cf
Durham,
supra
note
14,
at
602-03
(discussing
John
Cage's
compositional
practices).
412
[39:3
413
2016]
THERE'S
No
SUCH
THiNG
AS
A
COMPUTER-AUTHORED
WORK
that
is
enough,
even
if
other
aspects
are
literally
out
of
her
hands.
Although
this
approach
runs
into
some
difficulties
in
defining
substantial
similarity
and
when
the
process
is
repeated,
its
basic
claim
is
conceptually
sound.
25
Transposing
this
reasoning
back
into
cases
of
sequential
creation
suggests
that
the
composer
would
still
be
an
author
even
if
someone
else
rolls
the
dice,
and
similarly
the
programmer
of
a
nondeterministic
algorithm
can
be
an
author
of
its
outputs
even
if
someone
else
pushes
the
button.
A
second
reason
the
composer
can
be
an
author
is
that
she
adds
non-procedural
creativity
when
she
writes
out
the
notes
onto
a
sheet
of
staff
paper;
her
handwriting
is
hers
in
the
same
way
that
a
painter's
brushstrokes
are.
This
rationale
reaches
only
the
way
she
writes
the
notes,
not
the
notes
themselves,
but
it
too
is
enough
to
justify
a
copyright.
In
sequential-creation
cases,
this
reasoning
suggests
that
the
Musikalisches
Wirfelspiel
player
who
writes
out
a
dice-generated
minuet
longhand
is
also
pro
tanto
an
author.
This
kind
of
authorship
drops
out
entirely
from
computer-generated
works:
the
imperfections
and
variations
introduced
by
a
copyist's
hand
are
absent.
Hitting
"save"
to
store
a
digital
copy
of
a
freshly-
generated
minuet
introduces
no
new
authorship.
The
third
reason
is
more
openly
pragmatic.
We
could
admit
that
a
composer
who
plays
at
dice
does
not
control
their
fall,
any
more
than
Jackson
Pollock
controlled
the
fluid
dynamics
of
his
paint
splatters.
They
do
not
"owe
their
origin
to
an
act
of
authorship"
26
-but
we
are
nonetheless
willing
to
sweep
them
into
the
composer's
copyrights.
For
one
thing,
there
are
no
competing
claimants.
Either
the
composer
owns
the
copyright
in
these
aspects
of
the
work,
or
no
one
does.
Giving
her
a
copyright
does
not
deprive
other
authors
of
their
due.
For
another
thing,
there
is
no
great
need
for
public
access
to
this
particular
outcome.
Giving
the
composer
a
copyright
in
the
expression
resulting
from
this
particular
roll
of
the
dice
means
only
that
anyone
enamored
of
the
idea
just
needs
to
roll
their
own.
The
composer
ends
up
with
exclusive
rights
over
only
an
infinitesimally
small
sliver
of
the
possibility
space;
the
random
elements,
being
random,
are
exceedingly
unlikely
to
fall
the
same
way
again.
Treating
the
composer
as
the
author
in
law,
even
if
she
is
not
the
author
in
fact,
is
a
harmless
fiction-and
one
that
avoids
the
difficult
task
of
distinguishing
the
aspects
of
a
work
that
result
from
random
processes
inside
an
author's
body
from
the
aspects
that
result
from
random
processes
outside
of
it.
27
The
photographer
who
takes
a
picture
of
a
chaotic
ocean
swell
contributes
authorship
by
choosing
where
to
point
the
camera
and
when
to
click
the
shutter
button,
but
she
ends
up
with
a
copyright
on
patterns
of
waves
and
spray
that
are
in
no
sense
her
expression.
25.
See
generally
Dan
L.
Burk,
Method
and
Madness
in
Copyright
Law,
2007
UTAH
L.
REv.
587
(2007);
Alan
L.
Durham,
Speaking
of
the
World:
Fact,
Opinion
and
the
Originality
Standard
of
Copyright,
33
ARIZ.
ST.
L.J.
791
(2001);
James
Grimmelmann,
Three
Theories
of
Copyright
in
Ratings,
14
VAND.
J.
ENT.
&
TECH.
L.
851
(2011).
26.
Feist
Publ'ns,
Inc.
v.
Rural
Tel.
Serv.
Co.,
499
U.S.
340,
347
(1991).
27.
See
Durham,
supra
note
14,
at
573
(questioning
extent
to
which
this
distinction
is
maintainable).
I
owe
the
helpful
terminological
distinction
between
authorship
in
fact
and
authorship
in
law
to
Annemarie
Bridy.
Bridy,
supra
note
2,
at
¶¶
61-68.
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
But
these
simple
pragmatic
fictions
break
down
when
one
person
carries
out
a
nondeterministic
process
designed
by
another.
With
multiple
humans
in
the
picture,
an
"attribute
ownership
to
the
human"
rule
becomes
ambiguous.
This
is
not
to
say
that
the
rule
becomes
unusable:
we
could
for
equally
pragmatic
reasons
say
that
the
programmer
owns
a
given
nondeterministic
algorithmic
sequential
work,
or
that
the
user
does,
or
both.
It
is
just
that
we
must
confront
the
full
sixfold
diversity
of
sequential
creativity
cases,
which
means
that
no
rule
can
be
both
simple
and
pragmatic.
This
problem
was
latent
in
offline
examples
like
the
Musikalisches
Wirfelspiel,
but
not
a
great
source
of
worry.
It
does
not
appear
that
anyone
fretted
over
whether
the
game's
designer
or
players
had
a
better
claim
to
copyright
its
minuets.
Two
things
changed
with
the
advent
of
computer-assisted
creativity.
First,
cases
of
algorithmic
creation
went
from
being
literally
parlor
games
to
being
all
around
us.
Everyone
who
applies
the
Add
Noise
filter
in
Photoshop
or
records
a
speed
run
in
a
video
game
with
procedurally
generated
levels
creates
using
a
nondeterministic
algorithm
written
by
other
people,
an
algorithm
whose
inner
workings
and
detailed
effects
the
user
may
not
even
remotely
comprehend.
Second,
the
fact
that
a
computer
was
involved
seemed
to
offer
a
way
out
of
the
fact-bound
swamp
of
deciding
whether
programmer,
user,
both,
or
neither
qualified
as
an
author.
Perhaps
in
cases
involving
pseudorandom
creation
and
limited
user
input,
the
computer
itself
might
be
regarded
as
the
author?
This
suggestion
is
essentially
fallacious,
the
solution
a
mirage.
No
one,
to
my
knowledge,
has
ever
seriously
entertained
the
idea
that
the
Musikalisches
Wirfelspiel
dice
or
the
box
they
came
in
was
the
author
of
the
resulting
minuets.
Why
should
it
matter
that
in
"Mozart's
Dice
Game"
the
die-rolling
algorithm
is
implemented
on
a
computer?
Dice
are
not
authors,
and
neither
are
computer
programs.
It is
only
the
novelty
and
seeming
strangeness
of
computers
that
have
encouraged
people
to
think
otherwise.
It
is
possible
that
some
future
computer
programs
could
qualify
as
authors.
We
could
well
have
artificial
intelligences
that
are
responsive
to
incentives,
unpredictable
enough
that
we
can't
simply
tell
them
what
to
do,
and
that
have
attributes
of
personality
that
make
us
willing
to
regard
them
as
copyright
owners.
28
But
if
that
day
ever
comes,
it
will
because
we
have
already
made
a
decision
in
other
areas
of
life
and
law
to
treat
them
as
persons,
and
copyright
law
will
fall
in
line.
But
unless
those
mechanical
minds
also
invent
workable
time
travel,
their
future
existence
is
of
no
bearing
now.
The
copyright
issues
we
would
face
on
that
far
off
day
are
fundamentally
different
in
kind
from
those
we
face
today.
28.
See
James
Grimnelmann,
Copyright
for
Literate
Robots,
101
IOWA
L.
REv.
657
(2016)
(discussing
changes
that
would
make
it
reasonable
to
treat
artificial
intelligences
as
authors
and
readers
in
a
copyright
sense).
414
[39:3
2016]
THERE'S
No
SUCH
THING
AS
A
COMPUTER-AUTHORED
WORK
CONCLUSION
Computer
authorship
is
a law
of
the
horse.
29
It
is
a
special
case
of
a
more
general
problem,
one
that
presents
few
new
twists
on
familiar
issues.
This
is
a
little
surprising.
Over
on
the
infringement
side
of
the
aisle,
computer
copies
are
easy,
cheap,
instantaneous,
worldwide,
and
inescapable.
Infringement
doctrine
is
in
a
state
of
complete
upheaval,
one
that
may
bring
to
an
end
the
three-hundred-year
run
of
the
"copy"
as
the
basic
unit
of
infringement
analysis.
3
0
On
one
hand,
digital
rights
management
technologies
offer
copyright
owners
the
option
of
regulating
uses
rather
than
copies;
on
the
other,
open
access
and
peer
production
ask
whether
such
regulation
is
necessary
at
all.
Either
way,
big
changes
are
afoot.
Even
back
on
the
copyrightability
side
of
the
aisle,
computer
programs
themselves
pose
fundamental
questions
about
the
nature
of
the
authorship
that
copyright
protects.
31
Programs
are
dual
artifacts-useful
texts-and
they
interleave
creative
choice
and
functional
constraint
in
complex
ways.
Of
course,
copyright
has
always
worried
about
the
line
between
aesthetic
and
utilitarian,
but
computers
raise
the
stakes
because
they
are
such
literal-minded
readers.
So
computers
do
raise
hard
problems
for
copyright-just
not
here.
For
all
present
practical
purposes,
new
copyright
doctrines
for
computer-generated
works
are
a
terrible
idea.
32
The
problem
of
assigning
copyright
in
computer-generated
works
may
be
a
hard
problem,
but
it
is
not
a
new
problem.
It
is
hard
for
the
same
reason
that
copyright
has
always
been
hard-it
requires
us
to
make
objective
legal
judgments
about
the
ultimate
in
human
subjectivity:
aesthetics.
Putting
a
computer
somewhere
along
the
pathway
from
brain
to
brain
brings
some
of
these
difficulties
to
the
fore,
but
it
does
not
change
the
nature
of
the
question.
There
are
no
distinctive
conceptual
problems
with
computer-generated
works,
because
all
works
are
computer-generated-or
at
least,
they
might
have
been,
which
is
close
enough.
Because
computer-generated
works
are
no
different
in
kind
than
other
works,
special-purpose
doctrines
have
little
to
offer.
Indeed,
they
can
make
things
much
worse;
the
danger
of
claiming
that
there
is
"a"
rule
for
computer-generated
works
is
that
it
blinds
us
to
the
immense
diversity
that
category
encompasses.
There
can
be
no
rule
treating
all
computer-generated
works
alike
for
the
same
reason
there
can
be
no
rule
treating
all
fair
use
cases
alike:
relevant
differences
require
different
treatment.
It
is
only
the
apparent
novelty
of
computer-generated
works
that
seems
29.
See
Frank
H.
Easterbrook,
Cyberspace
and
the
Law
of
the
Horse,
1996
U.
CHi.
LEGAL
F.
207
(1996).
30.
See
Peter
Yu,
The
Copy
in
Copyright,
in
INTELLECTUAL
PROPERTY
&
ACCESS
TO
IMMATERIAL
GOODS
(Jessica
C.
Lai
and
Antoinette
Maget
Dominic6
eds.,
Edward
Elgar
Publishing
2016).
31.
See,
e.g.,
Pamela
Samuelson
et
al.,
A
Manifesto
Concerning
the
Legal
Protection
of
Computer
Programs,
94
COLUM.
L.
REv.
2308
(1994).
32.
To
put
it idiomatically,
computers
raise
difficult
§
102(b)
issues,
but
not
difficult
§
102(a)
issues.
For
a
recent
discussion
of
the
§
102(b)
issues,
see
Pamela
Samuelson,
Functionality
and
Expression
in
Computer
Programs:
Refining
the
Tests
for
Software
Copyright
Infringement
(UC
Berkeley
Pub.
L.
Res.
Paper
No.
2667740),
http://papers.ssm.com/sol3/papers.cfmn?abstract_id=
26
67 7
4
0
[https://perma.cc/9XG4-52TB]
(last
visited
Mar.
16,
2016).
415
416
COLUMBIA
JOURNAL
OF
LAW
&
THE
ARTS
[39:3
to
call
out
for
new
bright-line
rules.
But
that
is
the
same
deadly
siren
song
that
new
technologies
always
sing.
There
is
no
new
categorical
rule
of
personal
jurisdiction
on
the
Internet;
there
is
no
new
categorical
rule
for
drone
overflights;
there
is
no
new
categorical
rule
for
privacy
on
social
networks;
there
is
no
new
categorical
rule
for
a
thousand
other
technologies,
nor
should
there
be.
People
are
always
talking
about
technology
law,
when
they
should
be
talking
about
technology
facts.