<s>
General	O
game	O
playing	O
(	O
GGP	O
)	O
is	O
the	O
design	O
of	O
artificial	B-Application
intelligence	I-Application
programs	I-Application
to	O
be	O
able	O
to	O
play	O
more	O
than	O
one	O
game	O
successfully	O
.	O
</s>
<s>
For	O
many	O
games	O
like	O
chess	B-Application
,	O
computers	O
are	O
programmed	O
to	O
play	O
these	O
games	O
using	O
a	O
specially	O
designed	O
algorithm	O
,	O
which	O
cannot	O
be	O
transferred	O
to	O
another	O
context	O
.	O
</s>
<s>
For	O
instance	O
,	O
a	O
chess-playing	B-Application
computer	I-Application
program	I-Application
cannot	O
play	O
checkers	O
.	O
</s>
<s>
General	B-General_Concept
video	I-General_Concept
game	I-General_Concept
playing	I-General_Concept
(	O
GVGP	O
)	O
is	O
the	O
concept	O
of	O
GGP	O
adjusted	O
to	O
the	O
purpose	O
of	O
playing	O
video	O
games	O
.	O
</s>
<s>
For	O
video	O
games	O
,	O
game	O
rules	O
have	O
to	O
be	O
either	O
learnt	O
over	O
multiple	O
iterations	O
by	O
artificial	O
players	O
like	O
TD-Gammon	B-Application
,	O
or	O
are	O
predefined	O
manually	O
in	O
a	O
domain-specific	B-Language
language	I-Language
and	O
sent	O
in	O
advance	O
to	O
artificial	O
players	O
like	O
in	O
traditional	O
GGP	O
.	O
</s>
<s>
Starting	O
in	O
2013	O
,	O
significant	O
progress	O
was	O
made	O
following	O
the	O
deep	B-Algorithm
reinforcement	I-Algorithm
learning	I-Algorithm
approach	O
,	O
including	O
the	O
development	O
of	O
programs	O
that	O
can	O
learn	O
to	O
play	O
Atari	B-General_Concept
2600	I-General_Concept
games	O
as	O
well	O
as	O
a	O
program	O
that	O
can	O
learn	O
to	O
play	O
Nintendo	O
Entertainment	O
System	O
games	O
.	O
</s>
<s>
The	O
first	O
commercial	O
usage	O
of	O
general	O
game	O
playing	O
technology	O
was	O
Zillions	B-Application
of	I-Application
Games	I-Application
in	O
1998	O
.	O
</s>
<s>
This	O
was	O
the	O
first	O
program	O
to	O
automatically	O
generate	O
game	O
rules	O
of	O
chess-like	O
games	O
,	O
and	O
one	O
of	O
the	O
earliest	O
programs	O
to	O
use	O
automated	O
game	O
generation	O
.	O
</s>
<s>
This	O
system	O
was	O
able	O
to	O
play	O
a	O
number	O
of	O
chess-like	O
games	O
,	O
given	O
game	O
rules	O
definition	O
in	O
a	O
special	O
language	O
called	O
Game	B-Language
Description	I-Language
Language	I-Language
(	O
GDL	B-Language
)	O
,	O
without	O
any	O
human	O
interaction	O
once	O
the	O
games	O
were	O
generated	O
.	O
</s>
<s>
In	O
1998	O
,	O
the	O
commercial	O
system	O
Zillions	B-Application
of	I-Application
Games	I-Application
was	O
developed	O
by	O
Jeff	O
Mallett	O
and	O
Mark	O
Lefler	O
.	O
</s>
<s>
Zillions	B-Application
of	I-Application
Games	I-Application
derived	O
the	O
evaluation	B-General_Concept
function	I-General_Concept
automatically	O
from	O
the	O
game	O
rules	O
based	O
on	O
piece	O
mobility	O
,	O
board	O
structure	O
and	O
game	O
goals	O
.	O
</s>
<s>
It	O
also	O
employed	O
usual	O
algorithms	O
as	O
found	O
in	O
computer	B-Application
chess	I-Application
systems	O
:	O
alpha-beta	B-Algorithm
pruning	I-Algorithm
with	O
move	O
ordering	O
,	O
transposition	B-General_Concept
tables	I-General_Concept
,	O
etc	O
.	O
</s>
<s>
It	O
also	O
allows	O
definition	O
of	O
computer	B-Application
players	I-Application
,	O
which	O
participate	O
in	O
a	O
play	O
with	O
human	O
subjects	O
.	O
</s>
<s>
It	O
is	O
the	O
most	O
well-known	O
effort	O
at	O
standardizing	O
GGP	O
AI	B-Application
,	O
and	O
generally	O
seen	O
as	O
the	O
standard	O
for	O
GGP	O
systems	O
.	O
</s>
<s>
The	O
games	O
are	O
defined	O
by	O
sets	O
of	O
rules	O
represented	O
in	O
the	O
Game	B-Language
Description	I-Language
Language	I-Language
.	O
</s>
<s>
The	O
competition	O
judges	O
competitor	O
AI	B-Application
's	O
abilities	O
to	O
play	O
a	O
variety	O
of	O
different	O
games	O
,	O
by	O
recording	O
their	O
performance	O
on	O
each	O
individual	O
game	O
.	O
</s>
<s>
In	O
the	O
following	O
runoff	O
round	O
,	O
the	O
AIs	B-Application
face	O
off	O
against	O
each	O
other	O
in	O
increasingly	O
complex	O
games	O
.	O
</s>
<s>
The	O
AI	B-Application
that	O
wins	O
the	O
most	O
games	O
at	O
this	O
stage	O
wins	O
the	O
competition	O
,	O
and	O
until	O
2013	O
its	O
creator	O
used	O
to	O
win	O
a	O
$	O
10,000	O
prize	O
.	O
</s>
<s>
PolyGamo	O
Player	O
2017	O
Released	O
by	O
David	O
M	O
.	O
Bennett	O
in	O
September	O
2017	O
based	O
on	O
the	O
Unity	B-Application
game	I-Application
engine	I-Application
.	O
</s>
<s>
GVGP	O
could	O
potentially	O
be	O
used	O
to	O
create	O
real	O
video	B-Application
game	I-Application
AI	I-Application
automatically	O
,	O
as	O
well	O
as	O
"	O
to	O
test	O
game	O
environments	O
,	O
including	O
those	O
created	O
automatically	O
using	O
procedural	O
content	O
generation	O
and	O
to	O
find	O
potential	O
loopholes	O
in	O
the	O
gameplay	O
that	O
a	O
human	O
player	O
could	O
exploit	O
"	O
.	O
</s>
<s>
GVGP	O
has	O
also	O
been	O
used	O
to	O
generate	O
game	O
rules	O
,	O
and	O
estimate	O
a	O
game	O
's	O
quality	O
based	O
on	O
Relative	O
Algorithm	O
Performance	O
Profiles	O
(	O
RAPP	O
)	O
,	O
which	O
compare	O
the	O
skill	O
differentiation	O
that	O
a	O
game	O
allows	O
between	O
good	O
AI	B-Application
and	O
bad	O
AI	B-Application
.	O
</s>
<s>
The	O
General	O
Video	B-Application
Game	I-Application
AI	I-Application
Competition	O
(	O
GVGAI	O
)	O
has	O
been	O
running	O
since	O
2014	O
.	O
</s>
<s>
It	O
has	O
offered	O
a	O
way	O
for	O
researchers	O
and	O
practitioners	O
to	O
test	O
and	O
compare	O
their	O
best	O
general	B-General_Concept
video	I-General_Concept
game	I-General_Concept
playing	I-General_Concept
algorithms	O
.	O
</s>
<s>
The	O
competition	O
has	O
an	O
associated	O
software	O
framework	O
including	O
a	O
large	O
number	O
of	O
games	O
written	O
in	O
the	O
Video	O
Game	B-Language
Description	I-Language
Language	I-Language
(	O
VGDL	O
)	O
,	O
which	O
should	O
not	O
be	O
confused	O
with	O
GDL	B-Language
and	O
is	O
a	O
coding	O
language	O
using	O
simple	O
semantics	O
and	O
commands	O
that	O
can	O
easily	O
be	O
parsed	O
.	O
</s>
<s>
To	O
simplify	O
the	O
process	O
of	O
creating	O
an	O
AI	B-Application
that	O
can	O
interpret	O
video	O
games	O
,	O
games	O
for	O
this	O
purpose	O
are	O
written	O
in	O
VGDL	O
manually	O
.	O
</s>
<s>
Since	O
GGP	O
AI	B-Application
must	O
be	O
designed	O
to	O
play	O
multiple	O
games	O
,	O
its	O
design	O
cannot	O
rely	O
on	O
algorithms	O
created	O
specifically	O
for	O
certain	O
games	O
.	O
</s>
<s>
Instead	O
,	O
the	O
AI	B-Application
must	O
be	O
designed	O
using	O
algorithms	O
whose	O
methods	O
can	O
be	O
applied	O
to	O
a	O
wide	O
range	O
of	O
games	O
.	O
</s>
<s>
The	O
AI	B-Application
must	O
also	O
be	O
an	O
ongoing	O
process	O
,	O
that	O
can	O
adapt	O
to	O
its	O
current	O
state	O
rather	O
than	O
the	O
output	O
of	O
previous	O
states	O
.	O
</s>
<s>
A	O
popular	O
method	O
for	O
developing	O
GGP	O
AI	B-Application
is	O
the	O
Monte	B-Application
Carlo	I-Application
tree	I-Application
search	I-Application
(	O
MCTS	O
)	O
algorithm	O
.	O
</s>
<s>
Another	O
variation	O
of	O
tree-search	O
algorithms	O
used	O
is	O
the	O
Directed	B-Algorithm
Breadth-first	I-Algorithm
Search	I-Algorithm
(	O
DBS	O
)	O
,	O
in	O
which	O
a	O
child	O
node	O
to	O
the	O
current	O
state	O
is	O
created	O
for	O
each	O
available	O
action	O
,	O
and	O
visits	O
each	O
child	O
ordered	O
by	O
highest	O
average	O
reward	O
,	O
until	O
either	O
the	O
game	O
ends	O
or	O
runs	O
out	O
of	O
time	O
.	O
</s>
<s>
In	O
each	O
tree-search	O
method	O
,	O
the	O
AI	B-Application
simulates	O
potential	O
actions	O
and	O
ranks	O
each	O
based	O
on	O
the	O
average	O
highest	O
reward	O
of	O
each	O
path	O
,	O
in	O
terms	O
of	O
points	O
earned	O
.	O
</s>
<s>
Using	O
these	O
assumptions	O
,	O
game	O
playing	O
AI	B-Application
can	O
be	O
created	O
by	O
quantifying	O
the	O
player	O
input	O
,	O
the	O
game	O
outcomes	O
,	O
and	O
how	O
the	O
various	O
rules	O
apply	O
,	O
and	O
using	O
algorithms	O
to	O
compute	O
the	O
most	O
favorable	O
path	O
.	O
</s>
