<s>
Artificial	B-Application
intelligence	I-Application
(	O
AI	B-Application
)	O
is	O
intelligence	O
—	O
perceiving	O
,	O
synthesizing	O
,	O
and	O
inferring	O
information	O
—	O
demonstrated	O
by	O
machines	O
,	O
as	O
opposed	O
to	O
intelligence	O
displayed	O
by	O
non-human	O
animals	O
or	O
by	O
humans	O
.	O
</s>
<s>
Example	O
tasks	O
in	O
which	O
this	O
is	O
done	O
include	O
speech	B-Application
recognition	I-Application
,	O
computer	B-Application
vision	I-Application
,	O
translation	O
between	O
(	O
natural	O
)	O
languages	O
,	O
as	O
well	O
as	O
other	O
mappings	O
of	O
inputs	O
.	O
</s>
<s>
AI	B-Application
applications	I-Application
include	O
advanced	O
web	B-Application
search	I-Application
engines	I-Application
(	O
e.g.	O
,	O
Google	B-Application
Search	I-Application
)	O
,	O
recommendation	B-Application
systems	I-Application
(	O
used	O
by	O
YouTube	B-General_Concept
,	O
Amazon	B-Application
,	O
and	O
Netflix	O
)	O
,	O
understanding	B-General_Concept
human	I-General_Concept
speech	I-General_Concept
(	O
such	O
as	O
Siri	B-Application
and	O
Alexa	B-Application
)	O
,	O
self-driving	O
cars	O
(	O
e.g.	O
,	O
Waymo	O
)	O
,	O
generative	O
or	O
creative	O
tools	O
(	O
ChatGPT	B-General_Concept
and	O
AI	B-Application
art	O
)	O
,	O
automated	B-General_Concept
decision-making	I-General_Concept
,	O
and	O
competing	O
at	O
the	O
highest	O
level	O
in	O
strategic	O
game	O
systems	O
(	O
such	O
as	O
chess	B-Application
and	O
Go	O
)	O
.	O
</s>
<s>
As	O
machines	O
become	O
increasingly	O
capable	O
,	O
tasks	O
considered	O
to	O
require	O
"	O
intelligence	O
"	O
are	O
often	O
removed	O
from	O
the	O
definition	O
of	O
AI	B-Application
,	O
a	O
phenomenon	O
known	O
as	O
the	O
AI	B-Application
effect	O
.	O
</s>
<s>
For	O
instance	O
,	O
optical	B-Application
character	I-Application
recognition	I-Application
is	O
frequently	O
excluded	O
from	O
things	O
considered	O
to	O
be	O
AI	B-Application
,	O
having	O
become	O
a	O
routine	O
technology	O
.	O
</s>
<s>
Artificial	B-Application
intelligence	I-Application
was	O
founded	O
as	O
an	O
academic	O
discipline	O
in	O
1956	O
,	O
and	O
in	O
the	O
years	O
since	O
it	O
has	O
experienced	O
several	O
waves	O
of	O
optimism	O
,	O
followed	O
by	O
disappointment	O
and	O
the	O
loss	O
of	O
funding	O
(	O
known	O
as	O
an	O
"	O
AI	B-Application
winter	O
"	O
)	O
,	O
followed	O
by	O
new	O
approaches	O
,	O
success	O
,	O
and	O
renewed	O
funding	O
.	O
</s>
<s>
AI	B-Application
research	I-Application
has	O
tried	O
and	O
discarded	O
many	O
different	O
approaches	O
,	O
including	O
simulating	O
the	O
brain	O
,	O
modeling	O
human	O
problem	O
solving	O
,	O
formal	O
logic	O
,	O
large	O
databases	O
of	O
knowledge	O
,	O
and	O
imitating	O
animal	O
behavior	O
.	O
</s>
<s>
The	O
various	O
sub-fields	O
of	O
AI	B-Application
research	I-Application
are	O
centered	O
around	O
particular	O
goals	O
and	O
the	O
use	O
of	O
particular	O
tools	O
.	O
</s>
<s>
The	O
traditional	O
goals	O
of	O
AI	B-Application
research	I-Application
include	O
reasoning	O
,	O
knowledge	O
representation	O
,	O
planning	B-Application
,	O
learning	O
,	O
natural	B-Language
language	I-Language
processing	I-Language
,	O
perception	B-General_Concept
,	O
and	O
the	O
ability	O
to	O
move	O
and	O
manipulate	O
objects	O
.	O
</s>
<s>
To	O
solve	O
these	O
problems	O
,	O
AI	B-Application
researchers	O
have	O
adapted	O
and	O
integrated	O
a	O
wide	O
range	O
of	O
problem-solving	O
techniques	O
,	O
including	O
search	O
and	O
mathematical	O
optimization	O
,	O
formal	O
logic	O
,	O
artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
,	O
and	O
methods	O
based	O
on	O
statistics	O
,	O
probability	O
,	O
and	O
economics	O
.	O
</s>
<s>
AI	B-Application
also	O
draws	O
upon	O
computer	B-General_Concept
science	I-General_Concept
,	O
psychology	O
,	O
linguistics	O
,	O
philosophy	O
,	O
and	O
many	O
other	O
fields	O
.	O
</s>
<s>
Computer	B-General_Concept
scientists	I-General_Concept
and	O
philosophers	O
have	O
since	O
suggested	O
that	O
AI	B-Application
may	O
become	O
an	O
existential	O
risk	O
to	O
humanity	O
if	O
its	O
rational	O
capacities	O
are	O
not	O
steered	O
towards	O
beneficial	O
goals	O
.	O
</s>
<s>
The	O
term	O
artificial	B-Application
intelligence	I-Application
has	O
also	O
been	O
criticized	O
for	O
overhyping	O
AI	B-Application
's	O
true	O
technological	O
capabilities	O
.	O
</s>
<s>
Artificial	O
beings	O
with	O
intelligence	O
appeared	O
as	O
storytelling	O
devices	O
in	O
antiquity	O
,	O
and	O
have	O
been	O
common	O
in	O
fiction	O
,	O
as	O
in	O
Mary	O
Shelley	O
's	O
Frankenstein	B-Application
or	O
Karel	O
Čapek	O
's	O
R.U.R.	O
</s>
<s>
These	O
characters	O
and	O
their	O
fates	O
raised	O
many	O
of	O
the	O
same	O
issues	O
now	O
discussed	O
in	O
the	O
ethics	O
of	O
artificial	B-Application
intelligence	I-Application
.	I-Application
</s>
<s>
The	O
first	O
work	O
that	O
is	O
now	O
generally	O
recognized	O
as	O
AI	B-Application
was	O
McCullouch	O
and	O
Pitts	O
 '	O
1943	O
formal	O
design	O
for	O
Turing-complete	O
"	O
artificial	O
neurons	O
"	O
.	O
</s>
<s>
By	O
the	O
1950s	O
,	O
two	O
visions	O
for	O
how	O
to	O
achieve	O
machine	B-Application
intelligence	I-Application
emerged	O
.	O
</s>
<s>
One	O
vision	O
,	O
known	O
as	O
Symbolic	B-General_Concept
AI	I-General_Concept
or	O
GOFAI	B-General_Concept
,	O
was	O
to	O
use	O
computers	O
to	O
create	O
a	O
symbolic	O
representation	O
of	O
the	O
world	O
and	O
systems	O
that	O
could	O
reason	O
about	O
the	O
world	O
.	O
</s>
<s>
Closely	O
associated	O
with	O
this	O
approach	O
was	O
the	O
"	B-Algorithm
heuristic	I-Algorithm
search	I-Algorithm
"	I-Algorithm
approach	O
,	O
which	O
likened	O
intelligence	O
to	O
a	O
problem	O
of	O
exploring	O
a	O
space	O
of	O
possibilities	O
for	O
answers	O
.	O
</s>
<s>
Proponents	O
of	O
this	O
approach	O
,	O
most	O
prominently	O
Frank	O
Rosenblatt	O
,	O
sought	O
to	O
connect	O
Perceptron	B-Algorithm
in	O
ways	O
inspired	O
by	O
connections	O
of	O
neurons	O
.	O
</s>
<s>
James	O
Manyika	O
and	O
others	O
have	O
compared	O
the	O
two	O
approaches	O
to	O
the	O
mind	O
(	O
Symbolic	B-General_Concept
AI	I-General_Concept
)	O
and	O
the	O
brain	O
(	O
connectionist	O
)	O
.	O
</s>
<s>
Manyika	O
argues	O
that	O
symbolic	O
approaches	O
dominated	O
the	O
push	O
for	O
artificial	B-Application
intelligence	I-Application
in	O
this	O
period	O
,	O
due	O
in	O
part	O
to	O
its	O
connection	O
to	O
intellectual	O
traditions	O
of	O
Descartes	O
,	O
Boole	O
,	O
Gottlob	O
Frege	O
,	O
Bertrand	O
Russell	O
,	O
and	O
others	O
.	O
</s>
<s>
Connectionist	O
approaches	O
based	O
on	O
cybernetics	O
or	O
artificial	B-Architecture
neural	I-Architecture
networks	I-Architecture
were	O
pushed	O
to	O
the	O
background	O
but	O
have	O
gained	O
new	O
prominence	O
in	O
recent	O
decades	O
.	O
</s>
<s>
The	O
field	O
of	O
AI	B-Application
research	I-Application
was	O
born	O
at	O
a	O
workshop	O
at	O
Dartmouth	O
College	O
in	O
1956	O
.	O
</s>
<s>
The	O
attendees	O
became	O
the	O
founders	O
and	O
leaders	O
of	O
AI	B-Application
research	I-Application
.	O
</s>
<s>
Marvin	O
Minsky	O
agreed	O
,	O
writing	O
,	O
"	O
within	O
a	O
generation	O
...	O
the	O
problem	O
of	O
creating	O
'	O
artificial	B-Application
intelligence	I-Application
 '	O
will	O
substantially	O
be	O
solved	O
"	O
.	O
</s>
<s>
Progress	O
slowed	O
and	O
in	O
1974	O
,	O
in	O
response	O
to	O
the	O
criticism	O
of	O
Sir	O
James	O
Lighthill	O
and	O
ongoing	O
pressure	O
from	O
the	O
US	O
Congress	O
to	O
fund	O
more	O
productive	O
projects	O
,	O
both	O
the	O
U.S.	O
and	O
British	O
governments	O
cut	O
off	O
exploratory	O
research	O
in	O
AI	B-Application
.	O
</s>
<s>
The	O
next	O
few	O
years	O
would	O
later	O
be	O
called	O
an	O
"	O
AI	B-Application
winter	O
"	O
,	O
a	O
period	O
when	O
obtaining	O
funding	O
for	O
AI	B-Application
projects	O
was	O
difficult	O
.	O
</s>
<s>
In	O
the	O
early	O
1980s	O
,	O
AI	B-Application
research	I-Application
was	O
revived	O
by	O
the	O
commercial	O
success	O
of	O
expert	B-Application
systems	I-Application
,	O
a	O
form	O
of	O
AI	B-Application
program	O
that	O
simulated	O
the	O
knowledge	O
and	O
analytical	O
skills	O
of	O
human	O
experts	O
.	O
</s>
<s>
By	O
1985	O
,	O
the	O
market	O
for	O
AI	B-Application
had	O
reached	O
over	O
a	O
billion	O
dollars	O
.	O
</s>
<s>
At	O
the	O
same	O
time	O
,	O
Japan	O
's	O
fifth	B-Device
generation	I-Device
computer	I-Device
project	O
inspired	O
the	O
U.S.	O
and	O
British	O
governments	O
to	O
restore	O
funding	O
for	O
academic	O
research	O
.	O
</s>
<s>
However	O
,	O
beginning	O
with	O
the	O
collapse	O
of	O
the	O
Lisp	B-Operating_System
Machine	I-Operating_System
market	O
in	O
1987	O
,	O
AI	B-Application
once	O
again	O
fell	O
into	O
disrepute	O
,	O
and	O
a	O
second	O
,	O
longer-lasting	O
winter	O
began	O
.	O
</s>
<s>
Many	O
researchers	O
began	O
to	O
doubt	O
that	O
the	O
symbolic	B-General_Concept
approach	I-General_Concept
would	O
be	O
able	O
to	O
imitate	O
all	O
the	O
processes	O
of	O
human	O
cognition	O
,	O
especially	O
perception	B-General_Concept
,	O
robotics	O
,	O
learning	O
and	O
pattern	O
recognition	O
.	O
</s>
<s>
A	O
number	O
of	O
researchers	O
began	O
to	O
look	O
into	O
"	O
sub-symbolic	B-General_Concept
"	O
approaches	O
to	O
specific	O
AI	B-Application
problems	O
.	O
</s>
<s>
Robotics	O
researchers	O
,	O
such	O
as	O
Rodney	O
Brooks	O
,	O
rejected	O
symbolic	B-General_Concept
AI	I-General_Concept
and	O
focused	O
on	O
the	O
basic	O
engineering	O
problems	O
that	O
would	O
allow	O
robots	O
to	O
move	O
,	O
survive	O
,	O
and	O
learn	O
their	O
environment	O
.	O
</s>
<s>
Interest	O
in	O
neural	B-Architecture
networks	I-Architecture
and	O
"	O
connectionism	O
"	O
was	O
revived	O
by	O
Geoffrey	O
Hinton	O
,	O
David	O
Rumelhart	O
and	O
others	O
in	O
the	O
middle	O
of	O
the	O
1980s	O
.	O
</s>
<s>
Soft	B-General_Concept
computing	I-General_Concept
tools	O
were	O
developed	O
in	O
the	O
1980s	O
,	O
such	O
as	O
neural	B-Architecture
networks	I-Architecture
,	O
fuzzy	B-General_Concept
systems	I-General_Concept
,	O
Grey	O
system	O
theory	O
,	O
evolutionary	O
computation	O
and	O
many	O
tools	O
drawn	O
from	O
statistics	O
or	O
mathematical	O
optimization	O
.	O
</s>
<s>
AI	B-Application
gradually	O
restored	O
its	O
reputation	O
in	O
the	O
late	O
1990s	O
and	O
early	O
21st	O
century	O
by	O
finding	O
specific	O
solutions	O
to	O
specific	O
problems	O
.	O
</s>
<s>
By	O
2000	O
,	O
solutions	O
developed	O
by	O
AI	B-Application
researchers	O
were	O
being	O
widely	O
used	O
,	O
although	O
in	O
the	O
1990s	O
they	O
were	O
rarely	O
described	O
as	O
"	O
artificial	B-Application
intelligence	I-Application
"	O
.	O
</s>
<s>
Faster	O
computers	O
,	O
algorithmic	O
improvements	O
,	O
and	O
access	O
to	O
large	B-Application
amounts	I-Application
of	I-Application
data	I-Application
enabled	O
advances	O
in	O
machine	O
learning	O
and	O
perception	B-General_Concept
;	O
data-hungry	O
deep	B-Algorithm
learning	I-Algorithm
methods	O
started	O
to	O
dominate	O
accuracy	O
benchmarks	O
around	O
2012	O
.	O
</s>
<s>
According	O
to	O
Bloomberg	O
's	O
Jack	O
Clark	O
,	O
2015	O
was	O
a	O
landmark	O
year	O
for	O
artificial	B-Application
intelligence	I-Application
,	O
with	O
the	O
number	O
of	O
software	O
projects	O
that	O
use	O
AI	B-Application
within	O
Google	B-Application
increased	O
from	O
a	O
"	O
sporadic	O
usage	O
"	O
in	O
2012	O
to	O
more	O
than	O
2,700	O
projects	O
.	O
</s>
<s>
He	O
attributed	O
this	O
to	O
an	O
increase	O
in	O
affordable	O
neural	B-Architecture
networks	I-Architecture
,	O
due	O
to	O
a	O
rise	O
in	O
cloud	O
computing	O
infrastructure	O
and	O
to	O
an	O
increase	O
in	O
research	O
tools	O
and	O
datasets	O
.	O
</s>
<s>
In	O
a	O
2017	O
survey	O
,	O
one	O
in	O
five	O
companies	O
reported	O
they	O
had	O
"	O
incorporated	O
AI	B-Application
in	O
some	O
offerings	O
or	O
processes	O
"	O
.	O
</s>
<s>
The	O
amount	O
of	O
research	O
into	O
AI	B-Application
(	O
measured	O
by	O
total	O
publications	O
)	O
increased	O
by	O
50%	O
in	O
the	O
years	O
2015	O
–	O
2019	O
.	O
</s>
<s>
Numerous	O
academic	O
researchers	O
became	O
concerned	O
that	O
AI	B-Application
was	O
no	O
longer	O
pursuing	O
the	O
original	O
goal	O
of	O
creating	O
versatile	O
,	O
fully	O
intelligent	B-Application
machines	I-Application
.	O
</s>
<s>
Much	O
of	O
current	O
research	O
involves	O
statistical	O
AI	B-Application
,	O
which	O
is	O
overwhelmingly	O
used	O
to	O
solve	O
specific	O
problems	O
,	O
even	O
highly	O
successful	O
techniques	O
such	O
as	O
deep	B-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
By	O
the	O
late	O
1980s	O
and	O
1990s	O
,	O
AI	B-Application
research	I-Application
had	O
developed	O
methods	O
for	O
dealing	O
with	O
uncertain	O
or	O
incomplete	O
information	O
,	O
employing	O
concepts	O
from	O
probability	O
and	O
economics	O
.	O
</s>
<s>
Even	O
humans	O
rarely	O
use	O
the	O
step-by-step	O
deduction	O
that	O
early	O
AI	B-Application
research	I-Application
could	O
model	O
.	O
</s>
<s>
Knowledge	O
representation	O
and	O
knowledge	O
engineering	O
allow	O
AI	B-Application
programs	O
to	O
answer	O
questions	O
intelligently	O
and	O
make	O
deductions	O
about	O
real-world	O
facts	O
.	O
</s>
<s>
The	O
semantics	B-Application
of	O
an	O
ontology	O
is	O
typically	O
represented	O
in	O
description	B-Language
logic	I-Language
,	O
such	O
as	O
the	O
Web	B-Language
Ontology	I-Language
Language	I-Language
.	O
</s>
<s>
AI	B-Application
research	I-Application
has	O
developed	O
tools	O
to	O
represent	O
specific	O
domains	O
,	O
such	O
as	O
objects	O
,	O
properties	O
,	O
categories	O
and	O
relations	O
between	O
objects	O
;	O
situations	O
,	O
events	O
,	O
states	O
and	O
time	O
;	O
causes	O
and	O
effects	O
;	O
knowledge	O
about	O
knowledge	O
(	O
what	O
we	O
know	O
about	O
what	O
other	O
people	O
know	O
)	O
;	O
.	O
</s>
<s>
default	B-Application
reasoning	I-Application
(	O
things	O
that	O
humans	O
assume	O
are	O
true	O
until	O
they	O
are	O
told	O
differently	O
and	O
will	O
remain	O
true	O
even	O
when	O
other	O
facts	O
are	O
changing	O
)	O
;	O
as	O
well	O
as	O
other	O
domains	O
.	O
</s>
<s>
Among	O
the	O
most	O
difficult	O
problems	O
in	O
AI	B-Application
are	O
:	O
the	O
breadth	O
of	O
commonsense	O
knowledge	O
(	O
the	O
number	O
of	O
atomic	O
facts	O
that	O
the	O
average	O
person	O
knows	O
is	O
enormous	O
)	O
;	O
and	O
the	O
sub-symbolic	B-General_Concept
form	O
of	O
most	O
commonsense	O
knowledge	O
(	O
much	O
of	O
what	O
people	O
know	O
is	O
not	O
represented	O
as	O
"	O
facts	O
"	O
or	O
"	O
statements	O
"	O
that	O
they	O
could	O
express	O
verbally	O
)	O
.	O
</s>
<s>
Machine	O
learning	O
(	O
ML	O
)	O
,	O
a	O
fundamental	O
concept	O
of	O
AI	B-Application
research	I-Application
since	O
the	O
field	O
's	O
inception	O
,	O
is	O
the	O
study	O
of	O
computer	O
algorithms	O
that	O
improve	O
automatically	O
through	O
experience	O
.	O
</s>
<s>
Unsupervised	B-General_Concept
learning	I-General_Concept
finds	O
patterns	O
in	O
a	O
stream	O
of	O
input	O
.	O
</s>
<s>
Supervised	B-General_Concept
learning	I-General_Concept
requires	O
a	O
human	O
to	O
label	O
the	O
input	O
data	O
first	O
,	O
and	O
comes	O
in	O
two	O
main	O
varieties	O
:	O
classification	B-General_Concept
and	O
numerical	O
regression	O
.	O
</s>
<s>
Classification	B-General_Concept
is	O
used	O
to	O
determine	O
what	O
category	O
something	O
belongs	O
in	O
–	O
the	O
program	O
sees	O
a	O
number	O
of	O
examples	O
of	O
things	O
from	O
several	O
categories	O
and	O
will	O
learn	O
to	O
classify	B-General_Concept
new	O
inputs	O
.	O
</s>
<s>
Both	O
classifiers	B-General_Concept
and	O
regression	O
learners	O
can	O
be	O
viewed	O
as	O
"	O
function	O
approximators	O
"	O
trying	O
to	O
learn	O
an	O
unknown	O
(	O
possibly	O
implicit	O
)	O
function	O
;	O
for	O
example	O
,	O
a	O
spam	O
classifier	B-General_Concept
can	O
be	O
viewed	O
as	O
learning	O
a	O
function	O
that	O
maps	O
from	O
the	O
text	O
of	O
an	O
email	O
to	O
one	O
of	O
two	O
categories	O
,	O
"	O
spam	O
"	O
or	O
"	O
not	O
spam	O
"	O
.	O
</s>
<s>
Transfer	B-General_Concept
learning	I-General_Concept
is	O
when	O
the	O
knowledge	O
gained	O
from	O
one	O
problem	O
is	O
applied	O
to	O
a	O
new	O
problem	O
.	O
</s>
<s>
Computational	O
learning	O
theory	O
can	O
assess	O
learners	O
by	O
computational	O
complexity	O
,	O
by	O
sample	B-General_Concept
complexity	I-General_Concept
(	O
how	O
much	O
data	O
is	O
required	O
)	O
,	O
or	O
by	O
other	O
notions	O
of	O
optimization	O
.	O
</s>
<s>
allows	O
machines	O
to	O
read	O
and	O
understand	B-General_Concept
human	O
language	O
.	O
</s>
<s>
A	O
sufficiently	O
powerful	O
natural	B-Language
language	I-Language
processing	I-Language
system	O
would	O
enable	O
natural-language	B-General_Concept
user	I-General_Concept
interfaces	I-General_Concept
and	O
the	O
acquisition	O
of	O
knowledge	O
directly	O
from	O
human-written	O
sources	O
,	O
such	O
as	O
newswire	O
texts	O
.	O
</s>
<s>
Some	O
straightforward	O
applications	O
of	O
NLP	B-Language
include	O
information	B-Library
retrieval	I-Library
,	O
question	B-Algorithm
answering	I-Algorithm
and	O
machine	B-Application
translation	I-Application
.	O
</s>
<s>
Symbolic	B-General_Concept
AI	I-General_Concept
used	O
formal	O
syntax	B-Application
to	O
translate	O
the	O
deep	O
structure	O
of	O
sentences	O
into	O
logic	O
.	O
</s>
<s>
Modern	O
statistical	O
techniques	O
include	O
co-occurrence	O
frequencies	O
(	O
how	O
often	O
one	O
word	O
appears	O
near	O
another	O
)	O
,	O
"	O
Keyword	O
spotting	O
"	O
(	O
searching	O
for	O
a	O
particular	O
word	O
to	O
retrieve	O
information	O
)	O
,	O
transformer-based	O
deep	B-Algorithm
learning	I-Algorithm
(	O
which	O
finds	O
patterns	O
in	O
text	O
)	O
,	O
and	O
others	O
.	O
</s>
<s>
is	O
the	O
ability	O
to	O
use	O
input	O
from	O
sensors	O
(	O
such	O
as	O
cameras	O
,	O
microphones	O
,	O
wireless	O
signals	O
,	O
and	O
active	O
lidar	O
,	O
sonar	O
,	O
radar	B-Application
,	O
and	O
tactile	B-Algorithm
sensors	I-Algorithm
)	O
to	O
deduce	O
aspects	O
of	O
the	O
world	O
.	O
</s>
<s>
Applications	O
include	O
speech	B-Application
recognition	I-Application
,	O
</s>
<s>
Computer	B-Application
vision	I-Application
is	O
the	O
ability	O
to	O
analyze	O
visual	O
input	O
.	O
</s>
<s>
For	O
example	O
,	O
some	O
virtual	B-Protocol
assistants	I-Protocol
are	O
programmed	O
to	O
speak	O
conversationally	O
or	O
even	O
to	O
banter	O
humorously	O
;	O
it	O
makes	O
them	O
appear	O
more	O
sensitive	O
to	O
the	O
emotional	O
dynamics	O
of	O
human	O
interaction	O
,	O
or	O
to	O
otherwise	O
facilitate	O
human	O
–	O
computer	O
interaction	O
.	O
</s>
<s>
Moderate	O
successes	O
related	O
to	O
affective	O
computing	O
include	O
textual	O
sentiment	O
analysis	O
and	O
,	O
more	O
recently	O
,	O
multimodal	B-General_Concept
sentiment	I-General_Concept
analysis	I-General_Concept
)	O
,	O
wherein	O
AI	B-Application
classifies	O
the	O
affects	O
displayed	O
by	O
a	O
videotaped	O
subject	O
.	O
</s>
<s>
Hans	O
Moravec	O
and	O
Marvin	O
Minsky	O
argue	O
that	O
work	O
in	O
different	O
individual	O
domains	O
can	O
be	O
incorporated	O
into	O
an	O
advanced	O
multi-agent	O
system	O
or	O
cognitive	B-Architecture
architecture	I-Architecture
with	O
general	O
intelligence	O
.	O
</s>
<s>
Pedro	O
Domingos	O
hopes	O
that	O
there	O
is	O
a	O
conceptually	O
straightforward	O
,	O
but	O
mathematically	O
difficult	O
,	O
"	O
master	B-Algorithm
algorithm	I-Algorithm
"	O
that	O
could	O
lead	O
to	O
AGI	O
.	O
</s>
<s>
AI	B-Application
can	O
solve	O
many	O
problems	O
by	O
intelligently	O
searching	O
through	O
many	O
possible	O
solutions	O
.	O
</s>
<s>
Planning	B-Application
algorithms	I-Application
search	O
through	O
trees	O
of	O
goals	O
and	O
subgoals	O
,	O
attempting	O
to	O
find	O
a	O
path	O
to	O
a	O
target	O
goal	O
,	O
a	O
process	O
called	O
means-ends	B-General_Concept
analysis	I-General_Concept
.	O
</s>
<s>
Robotics	O
algorithms	O
for	O
moving	O
limbs	O
and	O
grasping	O
objects	O
use	O
local	B-Algorithm
searches	I-Algorithm
in	O
configuration	O
space	O
.	O
</s>
<s>
are	O
rarely	O
sufficient	O
for	O
most	O
real-world	O
problems	O
:	O
the	O
search	B-Application
space	I-Application
(	O
the	O
number	O
of	O
places	O
to	O
search	O
)	O
quickly	O
grows	O
to	O
astronomical	O
numbers	O
.	O
</s>
<s>
The	O
solution	O
,	O
for	O
many	O
problems	O
,	O
is	O
to	O
use	O
"	O
heuristics	B-Algorithm
"	O
or	O
"	O
rules	O
of	O
thumb	O
"	O
that	O
prioritize	O
choices	O
in	O
favor	O
of	O
those	O
more	O
likely	O
to	O
reach	O
a	O
goal	O
and	O
to	O
do	O
so	O
in	O
a	O
shorter	O
number	O
of	O
steps	O
.	O
</s>
<s>
In	O
some	O
search	O
methodologies	O
,	O
heuristics	B-Algorithm
can	O
also	O
serve	O
to	O
eliminate	O
some	O
choices	O
unlikely	O
to	O
lead	O
to	O
a	O
goal	O
(	O
called	O
"	O
pruning	B-Algorithm
the	O
search	B-Data_Structure
tree	I-Data_Structure
"	O
)	O
.	O
</s>
<s>
Heuristics	B-Algorithm
supply	O
the	O
program	O
with	O
a	O
"	O
best	O
guess	O
"	O
for	O
the	O
path	O
on	O
which	O
the	O
solution	O
lies	O
.	O
</s>
<s>
Heuristics	B-Algorithm
limit	O
the	O
search	O
for	O
solutions	O
into	O
a	O
smaller	O
sample	O
size	O
.	O
</s>
<s>
These	O
algorithms	O
can	O
be	O
visualized	O
as	O
blind	O
hill	B-Algorithm
climbing	I-Algorithm
:	O
we	O
begin	O
the	O
search	O
at	O
a	O
random	O
point	O
on	O
the	O
landscape	O
,	O
and	O
then	O
,	O
by	O
jumps	O
or	O
steps	O
,	O
we	O
keep	O
moving	O
our	O
guess	O
uphill	O
,	O
until	O
we	O
reach	O
the	O
top	O
.	O
</s>
<s>
Other	O
related	O
optimization	O
algorithms	O
include	O
random	B-Algorithm
optimization	I-Algorithm
,	O
beam	B-Algorithm
search	I-Algorithm
and	O
metaheuristics	O
like	O
simulated	B-Algorithm
annealing	I-Algorithm
.	O
</s>
<s>
Classic	O
evolutionary	B-Algorithm
algorithms	I-Algorithm
include	O
genetic	B-Algorithm
algorithms	I-Algorithm
,	O
gene	B-Algorithm
expression	I-Algorithm
programming	I-Algorithm
,	O
and	O
genetic	B-Algorithm
programming	I-Algorithm
.	O
</s>
<s>
Alternatively	O
,	O
distributed	O
search	O
processes	O
can	O
coordinate	O
via	O
swarm	B-Architecture
intelligence	I-Architecture
algorithms	O
.	O
</s>
<s>
Two	O
popular	O
swarm	O
algorithms	O
used	O
in	O
search	O
are	O
particle	B-Algorithm
swarm	I-Algorithm
optimization	I-Algorithm
(	O
inspired	O
by	O
bird	O
flocking	O
)	O
and	O
ant	B-Algorithm
colony	I-Algorithm
optimization	I-Algorithm
(	O
inspired	O
by	O
ant	O
trails	O
)	O
.	O
</s>
<s>
and	O
inductive	B-Application
logic	I-Application
programming	I-Application
is	O
a	O
method	O
for	O
learning	O
.	O
</s>
<s>
Several	O
different	O
forms	O
of	O
logic	O
are	O
used	O
in	O
AI	B-Application
research	I-Application
.	O
</s>
<s>
adds	O
quantifiers	B-Language
and	O
predicates	B-Algorithm
and	O
can	O
express	O
facts	O
about	O
objects	O
,	O
their	O
properties	O
,	O
and	O
their	O
relations	O
with	O
each	O
other	O
.	O
</s>
<s>
Default	B-Application
logics	I-Application
,	O
non-monotonic	O
logics	O
and	O
circumscription	B-Application
are	O
forms	O
of	O
logic	O
designed	O
to	O
help	O
with	O
default	B-Application
reasoning	I-Application
and	O
the	O
qualification	O
problem	O
.	O
</s>
<s>
Several	O
extensions	O
of	O
logic	O
have	O
been	O
designed	O
to	O
handle	O
specific	O
domains	O
of	O
knowledge	O
,	O
such	O
as	O
description	B-Language
logics	I-Language
;	O
</s>
<s>
situation	B-Application
calculus	I-Application
,	O
event	B-Application
calculus	I-Application
and	O
fluent	B-Application
calculus	I-Application
(	O
for	O
representing	O
events	O
and	O
time	O
)	O
;	O
</s>
<s>
Many	O
problems	O
in	O
AI	B-Application
(	O
including	O
in	O
reasoning	O
,	O
planning	B-Application
,	O
learning	O
,	O
perception	B-General_Concept
,	O
and	O
robotics	O
)	O
require	O
the	O
agent	O
to	O
operate	O
with	O
incomplete	O
or	O
uncertain	O
information	O
.	O
</s>
<s>
AI	B-Application
researchers	O
have	O
devised	O
a	O
number	O
of	O
tools	O
to	O
solve	O
these	O
problems	O
using	O
methods	O
from	O
probability	O
theory	O
and	O
economics	O
.	O
</s>
<s>
learning	O
(	O
using	O
the	O
expectation-maximization	B-Algorithm
algorithm	I-Algorithm
)	O
,	O
</s>
<s>
planning	B-Application
(	O
using	O
decision	O
networks	O
)	O
and	O
perception	B-General_Concept
(	O
using	O
dynamic	O
Bayesian	O
networks	O
)	O
.	O
</s>
<s>
Probabilistic	O
algorithms	O
can	O
also	O
be	O
used	O
for	O
filtering	O
,	O
prediction	O
,	O
smoothing	O
and	O
finding	O
explanations	O
for	O
streams	O
of	O
data	O
,	O
helping	O
perception	B-General_Concept
systems	O
to	O
analyze	O
processes	O
that	O
occur	O
over	O
time	O
(	O
e.g.	O
,	O
hidden	O
Markov	O
models	O
or	O
Kalman	O
filters	O
)	O
.	O
</s>
<s>
The	O
simplest	O
AI	B-Application
applications	I-Application
can	O
be	O
divided	O
into	O
two	O
types	O
:	O
classifiers	B-General_Concept
(	O
"	O
if	O
shiny	O
then	O
diamond	O
"	O
)	O
and	O
controllers	O
(	O
"	O
if	O
diamond	O
then	O
pick	O
up	O
"	O
)	O
.	O
</s>
<s>
Controllers	O
do	O
,	O
however	O
,	O
also	O
classify	B-General_Concept
conditions	O
before	O
inferring	O
actions	O
,	O
and	O
therefore	O
classification	B-General_Concept
forms	O
a	O
central	O
part	O
of	O
many	O
AI	B-Application
systems	O
.	O
</s>
<s>
Classifiers	B-General_Concept
are	O
functions	O
that	O
use	O
pattern	B-Language
matching	I-Language
to	O
determine	O
the	O
closest	O
match	O
.	O
</s>
<s>
They	O
can	O
be	O
tuned	O
according	O
to	O
examples	O
,	O
making	O
them	O
very	O
attractive	O
for	O
use	O
in	O
AI	B-Application
.	O
</s>
<s>
In	O
supervised	B-General_Concept
learning	I-General_Concept
,	O
each	O
pattern	O
belongs	O
to	O
a	O
certain	O
predefined	O
class	O
.	O
</s>
<s>
A	O
classifier	B-General_Concept
can	O
be	O
trained	O
in	O
various	O
ways	O
;	O
there	O
are	O
many	O
statistical	O
and	O
machine	O
learning	O
approaches	O
.	O
</s>
<s>
The	O
decision	B-Algorithm
tree	I-Algorithm
is	O
the	O
simplest	O
and	O
most	O
widely	O
used	O
symbolic	O
machine	O
learning	O
algorithm	O
.	O
</s>
<s>
K-nearest	B-General_Concept
neighbor	I-General_Concept
algorithm	I-General_Concept
was	O
the	O
most	O
widely	O
used	O
analogical	O
AI	B-Application
until	O
the	O
mid-1990s	O
.	O
</s>
<s>
Kernel	B-Algorithm
methods	I-Algorithm
such	O
as	O
the	O
support	B-Algorithm
vector	I-Algorithm
machine	I-Algorithm
(	O
SVM	B-Algorithm
)	O
displaced	O
k-nearest	B-General_Concept
neighbor	I-General_Concept
in	O
the	O
1990s	O
.	O
</s>
<s>
The	O
naive	B-General_Concept
Bayes	I-General_Concept
classifier	I-General_Concept
is	O
reportedly	O
the	O
"	O
most	O
widely	O
used	O
learner	O
"	O
at	O
Google	B-Application
,	O
due	O
in	O
part	O
to	O
its	O
scalability	O
.	O
</s>
<s>
Neural	B-Architecture
networks	I-Architecture
are	O
also	O
used	O
for	O
classification	B-General_Concept
.	O
</s>
<s>
Classifier	B-General_Concept
performance	O
depends	O
greatly	O
on	O
the	O
characteristics	O
of	O
the	O
data	O
to	O
be	O
classified	O
,	O
such	O
as	O
the	O
dataset	O
size	O
,	O
distribution	O
of	O
samples	O
across	O
classes	O
,	O
dimensionality	O
,	O
and	O
the	O
level	O
of	O
noise	O
.	O
</s>
<s>
Model-based	O
classifiers	B-General_Concept
perform	O
well	O
if	O
the	O
assumed	O
model	O
is	O
an	O
extremely	O
good	O
fit	O
for	O
the	O
actual	O
data	O
.	O
</s>
<s>
Otherwise	O
,	O
if	O
no	O
matching	O
model	O
is	O
available	O
,	O
and	O
if	O
accuracy	O
(	O
rather	O
than	O
speed	O
or	O
scalability	O
)	O
is	O
the	O
sole	O
concern	O
,	O
conventional	O
wisdom	O
is	O
that	O
discriminative	O
classifiers	B-General_Concept
(	O
especially	O
SVM	B-Algorithm
)	O
tend	O
to	O
be	O
more	O
accurate	O
than	O
model-based	O
classifiers	B-General_Concept
such	O
as	O
"	O
naive	B-General_Concept
Bayes	I-General_Concept
"	O
on	O
most	O
practical	O
data	O
sets	O
.	O
</s>
<s>
Modern	O
neural	B-Architecture
networks	I-Architecture
model	O
complex	O
relationships	O
between	O
inputs	O
and	O
outputs	O
and	O
find	O
patterns	O
in	O
data	O
.	O
</s>
<s>
Neural	B-Architecture
networks	I-Architecture
can	O
be	O
viewed	O
as	O
a	O
type	O
of	O
mathematical	O
optimization	O
–	O
they	O
perform	O
gradient	B-Algorithm
descent	I-Algorithm
on	O
a	O
multi-dimensional	O
topology	O
that	O
was	O
created	O
by	O
training	O
the	O
network	O
.	O
</s>
<s>
The	O
most	O
common	O
training	O
technique	O
is	O
the	O
backpropagation	B-Algorithm
algorithm	O
.	O
</s>
<s>
Other	O
learning	O
techniques	O
for	O
neural	B-Architecture
networks	I-Architecture
are	O
Hebbian	O
learning	O
(	O
"	O
fire	O
together	O
,	O
wire	O
together	O
"	O
)	O
,	O
GMDH	B-Algorithm
or	O
competitive	B-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
The	O
main	O
categories	O
of	O
networks	O
are	O
acyclic	O
or	O
feedforward	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
(	O
where	O
the	O
signal	O
passes	O
in	O
only	O
one	O
direction	O
)	O
and	O
recurrent	B-Algorithm
neural	I-Algorithm
networks	I-Algorithm
(	O
which	O
allow	O
feedback	O
and	O
short-term	O
memories	O
of	O
previous	O
input	O
events	O
)	O
.	O
</s>
<s>
Among	O
the	O
most	O
popular	O
feedforward	O
networks	O
are	O
perceptrons	B-Algorithm
,	O
multi-layer	B-Algorithm
perceptrons	I-Algorithm
and	O
radial	B-Algorithm
basis	I-Algorithm
networks	I-Algorithm
.	O
</s>
<s>
For	O
example	O
,	O
in	O
image	B-Algorithm
processing	I-Algorithm
,	O
lower	O
layers	O
may	O
identify	O
edges	O
,	O
while	O
higher	O
layers	O
may	O
identify	O
the	O
concepts	O
relevant	O
to	O
a	O
human	O
such	O
as	O
digits	O
or	O
letters	O
or	O
faces	O
.	O
</s>
<s>
Deep	B-Algorithm
learning	I-Algorithm
has	O
drastically	O
improved	O
the	O
performance	O
of	O
programs	O
in	O
many	O
important	O
subfields	O
of	O
artificial	B-Application
intelligence	I-Application
,	O
including	O
computer	B-Application
vision	I-Application
,	O
speech	B-Application
recognition	I-Application
,	O
image	O
classification	B-General_Concept
and	O
others	O
.	O
</s>
<s>
Deep	B-Algorithm
learning	I-Algorithm
often	O
uses	O
convolutional	B-Architecture
neural	I-Architecture
networks	I-Architecture
for	O
many	O
or	O
all	O
of	O
its	O
layers	O
.	O
</s>
<s>
In	O
a	O
recurrent	B-Algorithm
neural	I-Algorithm
network	I-Algorithm
(	O
RNN	O
)	O
the	O
signal	O
will	O
propagate	O
through	O
a	O
layer	O
more	O
than	O
once	O
;	O
</s>
<s>
thus	O
,	O
an	O
RNN	O
is	O
an	O
example	O
of	O
deep	B-Algorithm
learning	I-Algorithm
.	O
</s>
<s>
RNNs	O
can	O
be	O
trained	O
by	O
gradient	B-Algorithm
descent	I-Algorithm
,	O
</s>
<s>
however	O
long-term	O
gradients	O
which	O
are	O
back-propagated	O
can	O
"	O
vanish	O
"	O
(	O
that	O
is	O
,	O
they	O
can	O
tend	O
to	O
zero	O
)	O
or	O
"	O
explode	O
"	O
(	O
that	O
is	O
,	O
they	O
can	O
tend	O
to	O
infinity	O
)	O
,	O
known	O
as	O
the	O
vanishing	B-Algorithm
gradient	I-Algorithm
problem	I-Algorithm
.	O
</s>
<s>
The	O
long	B-Algorithm
short	I-Algorithm
term	I-Algorithm
memory	I-Algorithm
(	O
LSTM	B-Algorithm
)	O
technique	O
can	O
prevent	O
this	O
in	O
most	O
cases	O
.	O
</s>
<s>
Specialized	O
languages	O
for	O
artificial	B-Application
intelligence	I-Application
have	O
been	O
developed	O
,	O
such	O
as	O
Lisp	B-Language
,	O
Prolog	B-Language
,	O
TensorFlow	B-Language
and	O
many	O
others	O
.	O
</s>
<s>
Hardware	O
developed	O
for	O
AI	B-Application
includes	O
AI	B-General_Concept
accelerators	I-General_Concept
and	O
neuromorphic	O
computing	O
.	O
</s>
<s>
AI	B-Application
is	O
relevant	O
to	O
any	O
intellectual	O
task	O
.	O
</s>
<s>
Modern	O
artificial	B-Application
intelligence	I-Application
techniques	O
are	O
pervasive	O
and	O
are	O
too	O
numerous	O
to	O
list	O
here	O
.	O
</s>
<s>
Frequently	O
,	O
when	O
a	O
technique	O
reaches	O
mainstream	O
use	O
,	O
it	O
is	O
no	O
longer	O
considered	O
artificial	B-Application
intelligence	I-Application
;	O
this	O
phenomenon	O
is	O
described	O
as	O
the	O
AI	B-Application
effect	O
.	O
</s>
<s>
In	O
the	O
2010s	O
,	O
AI	B-Application
applications	I-Application
were	O
at	O
the	O
heart	O
of	O
the	O
most	O
commercially	O
successful	O
areas	O
of	O
computing	O
,	O
and	O
have	O
become	O
a	O
ubiquitous	O
feature	O
of	O
daily	O
life	O
.	O
</s>
<s>
AI	B-Application
is	O
used	O
in	O
search	B-Application
engines	I-Application
(	O
such	O
as	O
Google	B-Application
Search	I-Application
)	O
,	O
</s>
<s>
targeting	O
online	O
advertisements	O
,	O
recommendation	B-Application
systems	I-Application
(	O
offered	O
by	O
Netflix	O
,	O
YouTube	B-General_Concept
or	O
Amazon	B-Application
)	O
,	O
</s>
<s>
driving	O
internet	O
traffic	O
,	O
targeted	O
advertising	O
(	O
AdSense	O
,	O
Facebook	B-Application
)	O
,	O
</s>
<s>
virtual	B-Protocol
assistants	I-Protocol
(	O
such	O
as	O
Siri	B-Application
or	O
Alexa	B-Application
)	O
,	O
autonomous	O
vehicles	O
(	O
including	O
drones	O
,	O
ADAS	O
and	O
self-driving	O
cars	O
)	O
,	O
</s>
<s>
automatic	B-Application
language	I-Application
translation	I-Application
(	O
Microsoft	B-Application
Translator	I-Application
,	O
Google	B-Application
Translate	I-Application
)	O
,	O
</s>
<s>
,	O
spam	O
filtering	O
and	O
chatbots	B-Application
(	O
such	O
as	O
Chat	B-General_Concept
GPT	I-General_Concept
)	O
.	O
</s>
<s>
There	O
are	O
also	O
thousands	O
of	O
successful	O
AI	B-Application
applications	I-Application
used	O
to	O
solve	O
problems	O
for	O
specific	O
industries	O
or	O
institutions	O
.	O
</s>
<s>
A	O
few	O
examples	O
are	O
energy	O
storage	O
,	O
deepfakes	B-Application
,	O
medical	O
diagnosis	O
,	O
military	O
logistics	O
,	O
foreign	O
policy	O
,	O
or	O
supply	O
chain	O
management	O
.	O
</s>
<s>
Game	B-Application
playing	I-Application
has	O
been	O
a	O
test	O
of	O
AI	B-Application
's	O
strength	O
since	O
the	O
1950s	O
.	O
</s>
<s>
Deep	B-General_Concept
Blue	I-General_Concept
became	O
the	O
first	O
computer	O
chess-playing	O
system	O
to	O
beat	O
a	O
reigning	O
world	O
chess	B-Application
champion	O
,	O
Garry	O
Kasparov	O
,	O
on	O
11	O
May	O
1997	O
.	O
</s>
<s>
quiz	B-Application
show	I-Application
exhibition	O
match	O
,	O
IBM	O
's	O
question	B-Algorithm
answering	I-Algorithm
system	I-Algorithm
,	O
Watson	B-Application
,	O
defeated	O
the	O
two	O
greatest	O
Jeopardy	O
!	O
</s>
<s>
In	O
March	O
2016	O
,	O
AlphaGo	B-Application
won	O
4	O
out	O
of	O
5	O
games	O
of	O
Go	O
in	O
a	O
match	O
with	O
Go	O
champion	O
Lee	O
Sedol	O
,	O
becoming	O
the	O
first	O
computer	O
Go-playing	O
system	O
to	O
beat	O
a	O
professional	O
Go	O
player	O
without	O
handicaps	O
.	O
</s>
<s>
DeepMind	B-Application
in	O
the	O
2010s	O
developed	O
a	O
"	O
generalized	O
artificial	B-Application
intelligence	I-Application
"	O
that	O
could	O
learn	O
many	O
diverse	O
Atari	B-General_Concept
games	O
on	O
its	O
own	O
.	O
</s>
<s>
By	O
2020	O
,	O
Natural	B-Language
Language	I-Language
Processing	I-Language
systems	O
such	O
as	O
the	O
enormous	O
GPT-3	B-General_Concept
(	O
then	O
by	O
far	O
the	O
largest	O
artificial	B-Architecture
neural	I-Architecture
network	I-Architecture
)	O
were	O
matching	O
human	O
performance	O
on	O
pre-existing	O
benchmarks	O
,	O
albeit	O
without	O
the	O
system	O
attaining	O
a	O
commonsense	O
understanding	O
of	O
the	O
contents	O
of	O
the	O
benchmarks	O
.	O
</s>
<s>
DeepMind	B-Application
's	O
AlphaFold	B-Application
2	I-Application
(	O
2020	O
)	O
demonstrated	O
the	O
ability	O
to	O
approximate	O
,	O
in	O
hours	O
rather	O
than	O
months	O
,	O
the	O
3D	O
structure	O
of	O
a	O
protein	O
.	O
</s>
<s>
Other	O
applications	O
predict	O
the	O
result	O
of	O
judicial	O
decisions	O
,	O
create	O
art	O
(	O
such	O
as	O
poetry	O
or	O
painting	O
)	O
and	O
prove	B-Application
mathematical	I-Application
theorems	I-Application
.	O
</s>
<s>
AI	B-Application
content	O
detector	O
tools	O
are	O
software	O
applications	O
that	O
use	O
artificial	B-Application
intelligence	I-Application
(	O
AI	B-Application
)	O
algorithms	O
to	O
analyze	O
and	O
detect	O
specific	O
types	O
of	O
content	O
in	O
digital	O
media	O
,	O
such	O
as	O
text	O
,	O
images	O
,	O
and	O
videos	O
.	O
</s>
<s>
Some	O
benefits	O
of	O
using	O
AI	B-Application
content	O
detector	O
tools	O
include	O
improved	O
efficiency	O
and	O
accuracy	O
in	O
detecting	O
inappropriate	O
content	O
,	O
increased	O
safety	O
and	O
security	O
for	O
users	O
,	O
and	O
reduced	O
legal	O
and	O
reputational	O
risks	O
for	O
websites	O
and	O
platforms	O
.	O
</s>
<s>
Smart	B-Device
traffic	I-Device
lights	I-Device
have	O
been	O
developed	O
at	O
Carnegie	O
Mellon	O
since	O
2009	O
.	O
</s>
<s>
In	O
2019	O
,	O
WIPO	O
reported	O
that	O
AI	B-Application
was	O
the	O
most	O
prolific	O
emerging	O
technology	O
in	O
terms	O
of	O
the	O
number	O
of	O
patent	O
applications	O
and	O
granted	O
patents	O
,	O
the	B-Operating_System
Internet	I-Operating_System
of	I-Operating_System
things	I-Operating_System
was	O
estimated	O
to	O
be	O
the	O
largest	O
in	O
terms	O
of	O
market	O
size	O
.	O
</s>
<s>
It	O
was	O
followed	O
,	O
again	O
in	O
market	O
size	O
,	O
by	O
big	B-Application
data	I-Application
technologies	O
,	O
robotics	O
,	O
AI	B-Application
,	O
3D	O
printing	O
and	O
the	O
fifth	O
generation	O
of	O
mobile	O
services	O
(	O
5G	O
)	O
.	O
</s>
<s>
Since	O
AI	B-Application
emerged	O
in	O
the	O
1950s	O
,	O
340,000	O
AI-related	O
patent	O
applications	O
were	O
filed	O
by	O
innovators	O
and	O
1.6	O
million	O
scientific	O
papers	O
have	O
been	O
published	O
by	O
researchers	O
,	O
with	O
the	O
majority	O
of	O
all	O
AI-related	O
patent	O
filings	O
published	O
since	O
2013	O
.	O
</s>
<s>
Companies	O
represent	O
26	O
out	O
of	O
the	O
top	O
30	O
AI	B-Application
patent	O
applicants	O
,	O
with	O
universities	O
or	O
public	O
research	O
organizations	O
accounting	O
for	O
the	O
remaining	O
four	O
.	O
</s>
<s>
The	O
ratio	O
of	O
scientific	O
papers	O
to	O
inventions	O
has	O
significantly	O
decreased	O
from	O
8:1	O
in	O
2010	O
to	O
3:1	O
in	O
2016	O
,	O
which	O
is	O
attributed	O
to	O
be	O
indicative	O
of	O
a	O
shift	O
from	O
theoretical	O
research	O
to	O
the	O
use	O
of	O
AI	B-Application
technologies	O
in	O
commercial	O
products	O
and	O
services	O
.	O
</s>
<s>
Machine	O
learning	O
is	O
the	O
dominant	O
AI	B-Application
technique	O
disclosed	O
in	O
patents	O
and	O
is	O
included	O
in	O
more	O
than	O
one-third	O
of	O
all	O
identified	O
inventions	O
(	O
134,777	O
machine	O
learning	O
patents	O
filed	O
for	O
a	O
total	O
of	O
167,038	O
AI	B-Application
patents	O
filed	O
in	O
2016	O
)	O
,	O
with	O
computer	B-Application
vision	I-Application
being	O
the	O
most	O
popular	O
functional	O
application	O
.	O
</s>
<s>
AI-related	O
patents	O
not	O
only	O
disclose	O
AI	B-Application
techniques	O
and	O
applications	O
,	O
they	O
often	O
also	O
refer	O
to	O
an	O
application	O
field	O
or	O
industry	O
.	O
</s>
<s>
Other	O
sectors	O
included	O
banking	O
,	O
entertainment	O
,	O
security	O
,	O
industry	O
and	O
manufacturing	O
,	O
agriculture	O
,	O
and	O
networks	O
(	O
including	O
social	O
networks	O
,	O
smart	O
cities	O
and	O
the	B-Operating_System
Internet	I-Operating_System
of	I-Operating_System
things	I-Operating_System
)	O
.	O
</s>
<s>
IBM	O
has	O
the	O
largest	O
portfolio	O
of	O
AI	B-Application
patents	O
with	O
8,290	O
patent	O
applications	O
,	O
followed	O
by	O
Microsoft	O
with	O
5,930	O
patent	O
applications	O
.	O
</s>
<s>
Russell	O
and	O
Norvig	O
agree	O
with	O
Turing	O
that	O
AI	B-Application
must	O
be	O
defined	O
in	O
terms	O
of	O
"	O
acting	O
"	O
and	O
not	O
"	O
thinking	O
"	O
.	O
</s>
<s>
AI	B-Application
founder	O
John	O
McCarthy	O
agreed	O
,	O
writing	O
that	O
"	O
Artificial	B-Application
intelligence	I-Application
is	O
not	O
,	O
by	O
definition	O
,	O
simulation	O
of	O
human	O
intelligence	O
"	O
.	O
</s>
<s>
Another	O
AI	B-Application
founder	O
,	O
Marvin	O
Minsky	O
similarly	O
defines	O
it	O
as	O
"	O
the	O
ability	O
to	O
solve	O
hard	O
problems	O
"	O
.	O
</s>
<s>
A	O
definition	O
that	O
has	O
also	O
been	O
adopted	O
by	O
Google	B-Application
-	O
major	O
practitionary	O
in	O
the	O
field	O
of	O
AI	B-Application
.	O
</s>
<s>
No	O
established	O
unifying	O
theory	O
or	O
paradigm	O
has	O
guided	O
AI	B-Application
research	I-Application
for	O
most	O
of	O
its	O
history	O
.	O
</s>
<s>
The	O
unprecedented	O
success	O
of	O
statistical	O
machine	O
learning	O
in	O
the	O
2010s	O
eclipsed	O
all	O
other	O
approaches	O
(	O
so	O
much	O
so	O
that	O
some	O
sources	O
,	O
especially	O
in	O
the	O
business	O
world	O
,	O
use	O
the	O
term	O
"	O
artificial	B-Application
intelligence	I-Application
"	O
to	O
mean	O
"	O
machine	O
learning	O
with	O
neural	B-Architecture
networks	I-Architecture
"	O
)	O
.	O
</s>
<s>
This	O
approach	O
is	O
mostly	O
sub-symbolic	B-General_Concept
,	O
neat	O
,	O
soft	B-General_Concept
and	O
narrow	O
(	O
see	O
below	O
)	O
.	O
</s>
<s>
Critics	O
argue	O
that	O
these	O
questions	O
may	O
have	O
to	O
be	O
revisited	O
by	O
future	O
generations	O
of	O
AI	B-Application
researchers	O
.	O
</s>
<s>
Symbolic	B-General_Concept
AI	I-General_Concept
(	O
or	O
"	O
GOFAI	B-General_Concept
"	O
)	O
simulated	O
the	O
high-level	O
conscious	O
reasoning	O
that	O
people	O
use	O
when	O
they	O
solve	O
puzzles	O
,	O
express	O
legal	O
reasoning	O
and	O
do	O
mathematics	O
.	O
</s>
<s>
However	O
,	O
the	O
symbolic	B-General_Concept
approach	I-General_Concept
failed	O
on	O
many	O
tasks	O
that	O
humans	O
solve	O
easily	O
,	O
such	O
as	O
learning	O
,	O
recognizing	O
an	O
object	O
or	O
commonsense	O
reasoning	O
.	O
</s>
<s>
Moravec	O
's	O
paradox	O
is	O
the	O
discovery	O
that	O
high-level	O
"	O
intelligent	O
"	O
tasks	O
were	O
easy	O
for	O
AI	B-Application
,	O
but	O
low	O
level	O
"	O
instinctive	O
"	O
tasks	O
were	O
extremely	O
difficult	O
.	O
</s>
<s>
Although	O
his	O
arguments	O
had	O
been	O
ridiculed	O
and	O
ignored	O
when	O
they	O
were	O
first	O
presented	O
,	O
eventually	O
,	O
AI	B-Application
research	I-Application
came	O
to	O
agree	O
.	O
</s>
<s>
The	O
issue	O
is	O
not	O
resolved	O
:	O
sub-symbolic	B-General_Concept
reasoning	O
can	O
make	O
many	O
of	O
the	O
same	O
inscrutable	O
mistakes	O
that	O
human	O
intuition	O
does	O
,	O
such	O
as	O
algorithmic	O
bias	O
.	O
</s>
<s>
Critics	O
such	O
as	O
Noam	O
Chomsky	O
argue	O
continuing	O
research	O
into	O
symbolic	B-General_Concept
AI	I-General_Concept
will	O
still	O
be	O
necessary	O
to	O
attain	O
general	O
intelligence	O
,	O
in	O
part	O
because	O
sub-symbolic	B-General_Concept
AI	I-General_Concept
is	O
a	O
move	O
away	O
from	O
explainable	O
AI	B-Application
:	O
it	O
can	O
be	O
difficult	O
or	O
impossible	O
to	O
understand	B-General_Concept
why	O
a	O
modern	O
statistical	O
AI	B-Application
program	O
made	O
a	O
particular	O
decision	O
.	O
</s>
<s>
The	O
emerging	O
field	O
of	O
neuro-symbolic	B-General_Concept
artificial	I-General_Concept
intelligence	I-General_Concept
attempts	O
to	O
bridge	O
the	O
two	O
approaches	O
.	O
</s>
<s>
"	O
Neats	O
"	O
hope	O
that	O
intelligent	O
behavior	O
is	O
described	O
using	O
simple	O
,	O
elegant	O
principles	O
(	O
such	O
as	O
logic	O
,	O
optimization	O
,	O
or	O
neural	B-Architecture
networks	I-Architecture
)	O
.	O
</s>
<s>
Soft	B-General_Concept
computing	I-General_Concept
is	O
a	O
set	O
of	O
techniques	O
,	O
including	O
genetic	B-Algorithm
algorithms	I-Algorithm
,	O
fuzzy	O
logic	O
and	O
neural	B-Architecture
networks	I-Architecture
,	O
that	O
are	O
tolerant	O
of	O
imprecision	O
,	O
uncertainty	O
,	O
partial	O
truth	O
and	O
approximation	O
.	O
</s>
<s>
Soft	B-General_Concept
computing	I-General_Concept
was	O
introduced	O
in	O
the	O
late	O
80s	O
and	O
most	O
successful	O
AI	B-Application
programs	O
in	O
the	O
21st	O
century	O
are	O
examples	O
of	O
soft	B-General_Concept
computing	I-General_Concept
with	O
neural	B-Architecture
networks	I-Architecture
.	O
</s>
<s>
AI	B-Application
researchers	O
are	O
divided	O
as	O
to	O
whether	O
to	O
pursue	O
the	O
goals	O
of	O
artificial	O
general	O
intelligence	O
and	O
superintelligence	O
(	O
general	O
AI	B-Application
)	O
directly	O
or	O
to	O
solve	O
as	O
many	O
specific	O
problems	O
as	O
possible	O
(	O
narrow	O
AI	B-Application
)	O
in	O
hopes	O
these	O
solutions	O
will	O
lead	O
indirectly	O
to	O
the	O
field	O
's	O
long-term	O
goals	O
.	O
</s>
<s>
General	O
intelligence	O
is	O
difficult	O
to	O
define	O
and	O
difficult	O
to	O
measure	O
,	O
and	O
modern	O
AI	B-Application
has	O
had	O
more	O
verifiable	O
successes	O
by	O
focusing	O
on	O
specific	O
problems	O
with	O
specific	O
solutions	O
.	O
</s>
<s>
Mainstream	O
AI	B-Application
research	I-Application
considers	O
this	O
issue	O
irrelevant	O
because	O
it	O
does	O
not	O
affect	O
the	O
goals	O
of	O
the	O
field	O
.	O
</s>
<s>
Stuart	O
Russell	O
and	O
Peter	O
Norvig	O
observe	O
that	O
most	O
AI	B-Application
researchers	O
"	O
do	O
n't	O
care	O
about	O
the	O
[	O
philosophy	O
of	O
AI ]	O
–	O
as	O
long	O
as	O
the	O
program	O
works	O
,	O
they	O
do	O
n't	O
care	O
whether	O
you	O
call	O
it	O
a	O
simulation	O
of	O
intelligence	O
or	O
real	O
intelligence.	O
"	O
</s>
<s>
It	O
is	O
also	O
typically	O
the	O
central	O
question	O
at	O
issue	O
in	O
artificial	B-Application
intelligence	I-Application
in	O
fiction	O
.	O
</s>
<s>
This	O
philosophical	O
position	O
was	O
inspired	O
by	O
the	O
work	O
of	O
AI	B-Application
researchers	O
and	O
cognitive	O
scientists	O
in	O
the	O
1960s	O
and	O
was	O
originally	O
proposed	O
by	O
philosophers	O
Jerry	O
Fodor	O
and	O
Hilary	O
Putnam	O
.	O
</s>
<s>
Philosopher	O
John	O
Searle	O
characterized	O
this	O
position	O
as	O
"	O
strong	O
AI	B-Application
"	O
:	O
"	O
The	O
appropriately	O
programmed	O
computer	O
with	O
the	O
right	O
inputs	O
and	O
outputs	O
would	O
thereby	O
have	O
a	O
mind	O
in	O
exactly	O
the	O
same	O
sense	O
human	O
beings	O
have	O
minds.	O
"	O
</s>
<s>
Robot	O
designer	O
Hans	O
Moravec	O
,	O
cyberneticist	O
Kevin	O
Warwick	O
,	O
and	O
inventor	O
Ray	O
Kurzweil	O
have	O
predicted	O
that	O
humans	O
and	O
machines	O
will	O
merge	O
in	O
the	O
future	O
into	O
cyborgs	B-Application
that	O
are	O
more	O
capable	O
and	O
powerful	O
than	O
either	O
.	O
</s>
<s>
Edward	O
Fredkin	O
argues	O
that	O
"	O
artificial	B-Application
intelligence	I-Application
is	O
the	O
next	O
stage	O
in	O
evolution	O
"	O
,	O
an	O
idea	O
first	O
proposed	O
by	O
Samuel	O
Butler	O
's	O
"	O
Darwin	O
among	O
the	O
Machines	O
"	O
as	O
far	O
back	O
as	O
1863	O
,	O
and	O
expanded	O
upon	O
by	O
George	O
Dyson	O
in	O
his	O
book	O
of	O
the	O
same	O
name	O
in	O
1998	O
.	O
</s>
<s>
In	O
the	O
past	O
,	O
technology	O
has	O
tended	O
to	O
increase	O
rather	O
than	O
reduce	O
total	O
employment	O
,	O
but	O
economists	O
acknowledge	O
that	O
"	O
we	O
're	O
in	O
uncharted	O
territory	O
"	O
with	O
AI	B-Application
.	O
</s>
<s>
A	O
survey	O
of	O
economists	O
showed	O
disagreement	O
about	O
whether	O
the	O
increasing	O
use	O
of	O
robots	O
and	O
AI	B-Application
will	O
cause	O
a	O
substantial	O
increase	O
in	O
long-term	O
unemployment	O
,	O
but	O
they	O
generally	O
agree	O
that	O
it	O
could	O
be	O
a	O
net	O
benefit	O
if	O
productivity	O
gains	O
are	O
redistributed	O
.	O
</s>
<s>
Unlike	O
previous	O
waves	O
of	O
automation	O
,	O
many	O
middle-class	O
jobs	O
may	O
be	O
eliminated	O
by	O
artificial	B-Application
intelligence	I-Application
;	O
The	O
Economist	O
states	O
that	O
"	O
the	O
worry	O
that	O
AI	B-Application
could	O
do	O
to	O
white-collar	O
jobs	O
what	O
steam	O
power	O
did	O
to	O
blue-collar	O
ones	O
during	O
the	O
Industrial	O
Revolution	O
"	O
is	O
"	O
worth	O
taking	O
seriously	O
"	O
.	O
</s>
<s>
AI	B-Application
provides	O
a	O
number	O
of	O
tools	O
that	O
are	O
particularly	O
useful	O
for	O
authoritarian	O
governments	O
:	O
smart	O
spyware	O
,	O
face	O
recognition	O
and	O
voice	B-Application
recognition	I-Application
allow	O
widespread	O
surveillance	O
;	O
such	O
surveillance	O
allows	O
machine	O
learning	O
to	O
classify	B-General_Concept
potential	O
enemies	O
of	O
the	O
state	O
and	O
can	O
prevent	O
them	O
from	O
hiding	O
;	O
recommendation	B-Application
systems	I-Application
can	O
precisely	O
target	O
propaganda	O
and	O
misinformation	O
for	O
maximum	O
effect	O
;	O
deepfakes	B-Application
aid	O
in	O
producing	O
misinformation	O
;	O
advanced	O
AI	B-Application
can	O
make	O
centralized	O
decision	O
making	O
more	O
competitive	O
with	O
liberal	O
and	O
decentralized	O
systems	O
such	O
as	O
markets	O
.	O
</s>
<s>
Terrorists	O
,	O
criminals	O
and	O
rogue	O
states	O
may	O
use	O
other	O
forms	O
of	O
weaponized	O
AI	B-Application
such	O
as	O
advanced	O
digital	O
warfare	O
and	O
lethal	O
autonomous	O
weapons	O
.	O
</s>
<s>
Machine-learning	O
AI	B-Application
is	O
also	O
able	O
to	O
design	O
tens	O
of	O
thousands	O
of	O
toxic	O
molecules	O
in	O
a	O
matter	O
of	O
hours	O
.	O
</s>
<s>
AI	B-Application
programs	O
can	O
become	O
biased	O
after	O
learning	O
from	O
real-world	O
data	O
.	O
</s>
<s>
It	O
can	O
also	O
emerge	O
from	O
correlations	O
:	O
AI	B-Application
is	O
used	O
to	O
classify	B-General_Concept
individuals	O
into	O
groups	O
and	O
then	O
make	O
predictions	O
assuming	O
that	O
the	O
individual	O
will	O
resemble	O
other	O
members	O
of	O
the	O
group	O
.	O
</s>
<s>
An	O
example	O
of	O
this	O
is	O
COMPAS	B-Algorithm
,	O
a	O
commercial	O
program	O
widely	O
used	O
by	O
U.S.	O
courts	O
to	O
assess	O
the	O
likelihood	O
of	O
a	O
defendant	O
becoming	O
a	O
recidivist	O
.	O
</s>
<s>
ProPublica	O
claims	O
that	O
the	O
COMPAS-assigned	O
recidivism	O
risk	O
level	O
of	O
black	O
defendants	O
is	O
far	O
more	O
likely	O
to	O
be	O
overestimated	O
than	O
that	O
of	O
white	O
defendants	O
,	O
despite	O
the	O
fact	O
that	O
the	O
program	O
was	O
not	O
told	O
the	O
races	O
of	O
the	O
defendants	O
.	O
</s>
<s>
Health	O
equity	O
issues	O
may	O
also	O
be	O
exacerbated	O
when	O
many-to-many	B-Application
mapping	O
are	O
done	O
without	O
taking	O
steps	O
to	O
ensure	O
equity	O
for	O
populations	O
at	O
risk	O
for	O
bias	O
.	O
</s>
<s>
Other	O
examples	O
where	O
algorithmic	O
bias	O
can	O
lead	O
to	O
unfair	O
outcomes	O
are	O
when	O
AI	B-Application
is	O
used	O
for	O
credit	O
rating	O
or	O
hiring	O
.	O
</s>
<s>
At	O
its	O
2022	O
Conference	O
on	O
Fairness	O
,	O
Accountability	O
,	O
and	O
Transparency	O
(	O
ACM	O
FAccT	O
2022	O
)	O
the	O
Association	O
for	O
Computing	O
Machinery	O
,	O
in	O
Seoul	O
,	O
South	O
Korea	O
,	O
presented	O
and	O
published	O
findings	O
recommending	O
that	O
until	O
AI	B-Application
and	O
robotics	O
systems	O
are	O
demonstrated	O
to	O
be	O
free	O
of	O
bias	O
mistakes	O
,	O
they	O
are	O
unsafe	O
and	O
the	O
use	O
of	O
self-learning	O
neural	B-Architecture
networks	I-Architecture
trained	O
on	O
vast	O
,	O
unregulated	O
sources	O
of	O
flawed	O
internet	O
data	O
should	O
be	O
curtailed	O
.	O
</s>
<s>
Superintelligent	O
AI	B-Application
may	O
be	O
able	O
to	O
improve	O
itself	O
to	O
the	O
point	O
that	O
humans	O
could	O
not	O
control	O
it	O
.	O
</s>
<s>
Philosopher	O
Nick	O
Bostrom	O
argues	O
that	O
sufficiently	O
intelligent	O
AI	B-Application
,	O
if	O
it	O
chooses	O
actions	O
based	O
on	O
achieving	O
some	O
goal	O
,	O
will	O
exhibit	O
convergent	O
behavior	O
such	O
as	O
acquiring	O
resources	O
or	O
protecting	O
itself	O
from	O
being	O
shut	O
down	O
.	O
</s>
<s>
If	O
this	O
AI	B-Application
's	O
goals	O
do	O
not	O
fully	O
reflect	O
humanity	O
's	O
,	O
it	O
might	O
need	O
to	O
harm	O
humanity	O
to	O
acquire	O
more	O
resources	O
or	O
prevent	O
itself	O
from	O
being	O
shut	O
down	O
,	O
ultimately	O
to	O
better	O
achieve	O
its	O
goal	O
.	O
</s>
<s>
He	O
concludes	O
that	O
AI	B-Application
poses	O
a	O
risk	O
to	O
mankind	O
,	O
however	O
humble	O
or	O
"	O
friendly	O
"	O
its	O
stated	O
goals	O
might	O
be	O
.	O
</s>
<s>
The	O
opinion	O
of	O
experts	O
and	O
industry	O
insiders	O
is	O
mixed	O
,	O
with	O
sizable	O
fractions	O
both	O
concerned	O
and	O
unconcerned	O
by	O
risk	O
from	O
eventual	O
superhumanly-capable	O
AI	B-Application
.	O
</s>
<s>
Stephen	O
Hawking	O
,	O
Microsoft	O
founder	O
Bill	O
Gates	O
,	O
history	O
professor	O
Yuval	O
Noah	O
Harari	O
,	O
and	O
SpaceX	O
founder	O
Elon	O
Musk	O
have	O
all	O
expressed	O
serious	O
misgivings	O
about	O
the	O
future	O
of	O
AI	B-Application
.	O
</s>
<s>
Prominent	O
tech	O
titans	O
including	O
Peter	O
Thiel	O
(	O
Amazon	B-Application
Web	I-Application
Services	I-Application
)	O
and	O
Musk	O
have	O
committed	O
more	O
than	O
$1billion	O
to	O
nonprofit	O
companies	O
that	O
champion	O
responsible	O
AI	B-Application
development	O
,	O
such	O
as	O
OpenAI	O
and	O
the	O
Future	O
of	O
Life	O
Institute	O
.	O
</s>
<s>
Mark	O
Zuckerberg	O
(	O
CEO	O
,	O
Facebook	B-Application
)	O
has	O
said	O
that	O
artificial	B-Application
intelligence	I-Application
is	O
helpful	O
in	O
its	O
current	O
form	O
and	O
will	O
continue	O
to	O
assist	O
humans	O
.	O
</s>
<s>
Rodney	O
Brooks	O
,	O
in	O
particular	O
,	O
has	O
said	O
that	O
"	O
malevolent	O
"	O
AI	B-Application
is	O
still	O
centuries	O
away	O
.	O
</s>
<s>
AI	B-Application
's	O
decisions	O
making	O
abilities	O
raises	O
the	O
questions	O
of	O
legal	O
responsibility	O
and	O
copyright	O
status	O
of	O
created	O
works	O
.	O
</s>
<s>
Friendly	O
AI	B-Application
are	O
machines	O
that	O
have	O
been	O
designed	O
from	O
the	O
beginning	O
to	O
minimize	O
risks	O
and	O
to	O
make	O
choices	O
that	O
benefit	O
humans	O
.	O
</s>
<s>
Eliezer	O
Yudkowsky	O
,	O
who	O
coined	O
the	O
term	O
,	O
argues	O
that	O
developing	O
friendly	O
AI	B-Application
should	O
be	O
a	O
higher	O
research	O
priority	O
:	O
it	O
may	O
require	O
a	O
large	O
investment	O
and	O
it	O
must	O
be	O
completed	O
before	O
AI	B-Application
becomes	O
an	O
existential	O
risk	O
.	O
</s>
<s>
The	O
regulation	O
of	O
artificial	B-Application
intelligence	I-Application
is	O
the	O
development	O
of	O
public	O
sector	O
policies	O
and	O
laws	O
for	O
promoting	O
and	O
regulating	O
artificial	B-Application
intelligence	I-Application
(	O
AI	B-Application
)	O
;	O
it	O
is	O
therefore	O
related	O
to	O
the	O
broader	O
regulation	O
of	O
algorithms	O
.	O
</s>
<s>
The	O
regulatory	O
and	O
policy	O
landscape	O
for	O
AI	B-Application
is	O
an	O
emerging	O
issue	O
in	O
jurisdictions	O
globally	O
.	O
</s>
<s>
Between	O
2016	O
and	O
2020	O
,	O
more	O
than	O
30	O
countries	O
adopted	O
dedicated	O
strategies	O
for	O
AI	B-Application
.	O
</s>
<s>
Most	O
EU	O
member	O
states	O
had	O
released	O
national	O
AI	B-Application
strategies	O
,	O
as	O
had	O
Canada	O
,	O
China	O
,	O
India	O
,	O
Japan	O
,	O
Mauritius	O
,	O
the	O
Russian	O
Federation	O
,	O
Saudi	O
Arabia	O
,	O
United	O
Arab	O
Emirates	O
,	O
US	O
and	O
Vietnam	O
.	O
</s>
<s>
Others	O
were	O
in	O
the	O
process	O
of	O
elaborating	O
their	O
own	O
AI	B-Application
strategy	O
,	O
including	O
Bangladesh	O
,	O
Malaysia	O
and	O
Tunisia	O
.	O
</s>
<s>
The	O
Global	O
Partnership	O
on	O
Artificial	B-Application
Intelligence	I-Application
was	O
launched	O
in	O
June	O
2020	O
,	O
stating	O
a	O
need	O
for	O
AI	B-Application
to	O
be	O
developed	O
in	O
accordance	O
with	O
human	O
rights	O
and	O
democratic	O
values	O
,	O
to	O
ensure	O
public	O
confidence	O
and	O
trust	O
in	O
the	O
technology	O
.	O
</s>
<s>
Henry	O
Kissinger	O
,	O
Eric	O
Schmidt	O
,	O
and	O
Daniel	O
Huttenlocher	O
published	O
a	O
joint	O
statement	O
in	O
November	O
2021	O
calling	O
for	O
a	O
government	O
commission	O
to	O
regulate	O
AI	B-Application
.	O
</s>
<s>
A	O
common	O
trope	O
in	O
these	O
works	O
began	O
with	O
Mary	O
Shelley	O
's	O
Frankenstein	B-Application
,	O
where	O
a	O
human	O
creation	O
becomes	O
a	O
threat	O
to	O
its	O
masters	O
.	O
</s>
<s>
This	O
includes	O
such	O
works	O
as	O
Arthur	O
C	O
.	O
Clarke	O
's	O
and	O
Stanley	O
Kubrick	O
's	O
2001	O
:	O
A	O
Space	O
Odyssey	O
(	O
both	O
1968	O
)	O
,	O
with	O
HAL	B-General_Concept
9000	I-General_Concept
,	O
the	O
murderous	O
computer	O
in	O
charge	O
of	O
the	O
Discovery	O
One	O
spaceship	O
,	O
as	O
well	O
as	O
The	B-General_Concept
Terminator	I-General_Concept
(	O
1984	O
)	O
and	O
The	O
Matrix	O
(	O
1999	O
)	O
.	O
</s>
<s>
while	O
almost	O
all	O
artificial	B-Application
intelligence	I-Application
researchers	O
are	O
familiar	O
with	O
Asimov	O
's	O
laws	O
through	O
popular	O
culture	O
,	O
they	O
generally	O
consider	O
the	O
laws	O
useless	O
for	O
many	O
reasons	O
,	O
one	O
of	O
which	O
is	O
their	O
ambiguity	O
.	O
</s>
<s>
Transhumanism	O
(	O
the	O
merging	O
of	O
humans	O
and	O
machines	O
)	O
is	O
explored	O
in	O
the	O
manga	O
Ghost	B-Application
in	I-Application
the	I-Application
Shell	I-Application
and	O
the	O
science-fiction	O
series	O
Dune	O
.	O
</s>
<s>
Several	O
works	O
use	O
AI	B-Application
to	O
force	O
us	O
to	O
confront	O
the	O
fundamental	O
question	O
of	O
what	O
makes	O
us	O
human	O
,	O
showing	O
us	O
artificial	O
beings	O
that	O
have	O
the	O
ability	O
to	O
feel	O
,	O
and	O
thus	O
to	O
suffer	O
.	O
</s>
<s>
This	O
appears	O
in	O
Karel	O
Čapek	O
's	O
R.U.R.	O
,	O
the	O
films	O
A.I.	B-Application
</s>
<s>
Artificial	B-Application
Intelligence	I-Application
and	O
Ex	O
Machina	O
,	O
as	O
well	O
as	O
the	O
novel	O
Do	O
Androids	O
Dream	O
of	O
Electric	O
Sheep	O
?,	O
by	O
Philip	O
K	O
.	O
Dick	O
.	O
</s>
<s>
Dick	O
considers	O
the	O
idea	O
that	O
our	O
understanding	O
of	O
human	O
subjectivity	O
is	O
altered	O
by	O
technology	O
created	O
with	O
artificial	B-Application
intelligence	I-Application
.	I-Application
</s>
