<s>
MANIC	B-Architecture
,	O
formerly	O
known	O
as	O
PMML.1	B-Architecture
,	O
is	O
a	O
cognitive	B-Architecture
architecture	I-Architecture
developed	O
by	O
the	O
predictive	B-General_Concept
modeling	I-General_Concept
and	O
machine	O
learning	O
laboratory	O
at	O
University	O
of	O
Arkansas	O
.	O
</s>
<s>
It	O
differs	O
from	O
other	O
cognitive	B-Architecture
architectures	I-Architecture
in	O
that	O
it	O
tries	O
to	O
"	O
minimize	O
novelty	O
"	O
.	O
</s>
<s>
While	O
most	O
other	O
cognitive	B-Architecture
architectures	I-Architecture
are	O
inspired	O
by	O
some	O
neurological	O
observation	O
,	O
and	O
are	O
subsequently	O
developed	O
in	O
a	O
top-down	O
manner	O
to	O
behave	O
in	O
some	O
manner	O
like	O
a	O
brain	O
,	O
MANIC	B-Architecture
is	O
inspired	O
only	O
by	O
common	O
practices	O
in	O
computer	O
science	O
,	O
and	O
was	O
developed	O
in	O
a	O
bottom-up	O
manner	O
for	O
the	O
purpose	O
of	O
unifying	O
various	O
methods	O
in	O
machine	O
learning	O
and	O
artificial	B-Application
intelligence	I-Application
.	I-Application
</s>
<s>
At	O
the	O
highest	O
level	O
,	O
MANIC	B-Architecture
describes	O
a	O
software	O
agent	O
that	O
,	O
supposedly	O
,	O
will	O
exhibit	O
cognitive	O
intelligence	O
.	O
</s>
<s>
The	O
observation	O
function	O
is	O
implemented	O
with	O
a	O
generative	O
deep	B-Algorithm
learning	I-Algorithm
architecture	I-Algorithm
.	O
</s>
<s>
It	O
is	O
trained	O
in	O
an	O
unsupervised	B-General_Concept
manner	O
from	O
the	O
observations	O
that	O
the	O
agent	O
makes	O
.	O
</s>
<s>
The	O
transition	O
function	O
is	O
trained	O
in	O
a	O
supervised	B-General_Concept
manner	O
,	O
to	O
predict	O
the	O
next	O
beliefs	O
from	O
the	O
current	O
ones	O
.	O
</s>
<s>
The	O
planning	O
module	O
uses	O
an	O
evolutionary	B-Algorithm
algorithm	I-Algorithm
to	O
evolve	O
a	O
satisficing	O
plan	O
.	O
</s>
<s>
In	O
order	O
to	O
facilitate	O
this	O
reinforcement	O
learning	O
,	O
MANIC	B-Architecture
provides	O
a	O
mechanism	O
for	O
the	O
agent	O
to	O
generate	O
"	O
fantasy	O
videos	O
"	O
that	O
show	O
anticipated	O
observations	O
if	O
a	O
candidate	O
plan	O
were	O
to	O
be	O
executed	O
.	O
</s>
<s>
MANIC	B-Architecture
proposes	O
that	O
the	O
learning	O
system	O
gives	O
the	O
agent	O
awareness	O
of	O
its	O
environment	O
by	O
modeling	O
it	O
,	O
and	O
using	O
that	O
model	O
to	O
anticipate	O
future	O
beliefs	O
.	O
</s>
<s>
To	O
some	O
extent	O
,	O
MANIC	B-Architecture
suggests	O
that	O
existing	O
methods	O
already	O
in	O
use	O
in	O
artificial	B-Application
intelligence	I-Application
are	O
unintentionally	O
creating	O
subjective	O
experiences	O
like	O
those	O
typically	O
associated	O
with	O
conscious	O
beings	O
.	O
</s>
