<s>
In	O
artificial	B-Application
intelligence	I-Application
,	O
apprenticeship	B-General_Concept
learning	I-General_Concept
(	O
or	O
learning	O
from	O
demonstration	O
)	O
is	O
the	O
process	O
of	O
learning	O
by	O
observing	O
an	O
expert	O
.	O
</s>
<s>
It	O
can	O
be	O
viewed	O
as	O
a	O
form	O
of	O
supervised	B-General_Concept
learning	I-General_Concept
,	O
where	O
the	O
training	O
dataset	O
consists	O
of	O
task	O
executions	O
by	O
a	O
demonstration	O
teacher	O
.	O
</s>
<s>
In	O
2017	O
,	O
OpenAI	O
and	O
DeepMind	B-Application
applied	O
deep	B-Algorithm
learning	I-Algorithm
to	O
the	O
cooperative	O
inverse	O
reinforcement	O
learning	O
in	O
simple	O
domains	O
such	O
as	O
Atari	O
games	O
and	O
straightforward	O
robot	O
tasks	O
such	O
as	O
backflips	O
.	O
</s>
<s>
Apprenticeship	O
via	O
inverse	O
reinforcement	O
learning	O
(	O
AIRP	O
)	O
was	O
developed	O
by	O
in	O
2004	O
Pieter	O
Abbeel	O
,	O
Professor	O
in	O
Berkeley	O
's	O
EECS	O
department	O
,	O
and	O
Andrew	O
Ng	O
,	O
Associate	O
Professor	O
in	O
Stanford	O
University	O
's	O
Computer	B-General_Concept
Science	I-General_Concept
Department	O
.	O
</s>
<s>
Learning	O
from	O
demonstration	O
is	O
often	O
explained	O
from	O
a	O
perspective	O
that	O
the	O
working	O
Robot-control-system	B-Application
is	O
available	O
and	O
the	O
human-demonstrator	O
is	O
using	O
it	O
.	O
</s>
<s>
The	O
idea	O
from	O
Schaal	O
was	O
,	O
not	O
to	O
use	O
a	O
Brute-force	B-Algorithm
solver	I-Algorithm
but	O
record	O
the	O
movements	O
of	O
a	O
human-demonstration	O
.	O
</s>
<s>
In	O
computer	O
animation	O
,	O
the	O
principle	O
is	O
called	O
spline	B-Algorithm
animation	I-Algorithm
.	O
</s>
