<s>
Cyc	B-Language
(	O
pronounced	O
)	O
is	O
a	O
long-term	O
artificial	B-Application
intelligence	I-Application
project	I-Application
that	O
aims	O
to	O
assemble	O
a	O
comprehensive	O
ontology	O
and	O
knowledge	O
base	O
that	O
spans	O
the	O
basic	O
concepts	O
and	O
rules	O
about	O
how	O
the	O
world	O
works	O
.	O
</s>
<s>
Hoping	O
to	O
capture	O
common	B-General_Concept
sense	I-General_Concept
knowledge	I-General_Concept
,	O
Cyc	B-Language
focuses	O
on	O
implicit	O
knowledge	O
that	O
other	O
AI	B-Application
platforms	O
may	O
take	O
for	O
granted	O
.	O
</s>
<s>
Cyc	B-Language
enables	O
semantic	O
reasoners	O
to	O
perform	O
human-like	O
reasoning	O
and	O
be	O
less	O
"	O
brittle	O
"	O
when	O
confronted	O
with	O
novel	O
situations	O
.	O
</s>
<s>
Douglas	O
Lenat	O
began	O
the	O
project	O
in	O
July	O
1984	O
at	O
MCC	O
,	O
where	O
he	O
was	O
Principal	O
Scientist	O
1984	O
–	O
1994	O
,	O
and	O
then	O
,	O
since	O
January	O
1995	O
,	O
has	O
been	O
under	O
active	O
development	O
by	O
the	O
Cycorp	B-Language
company	O
,	O
where	O
he	O
is	O
the	O
CEO	O
.	O
</s>
<s>
The	O
need	O
for	O
a	O
massive	O
symbolic	O
artificial	B-Application
intelligence	I-Application
project	I-Application
of	O
this	O
kind	O
was	O
born	O
in	O
the	O
early	O
1980s	O
.	O
</s>
<s>
Early	O
AI	B-Application
researchers	O
had	O
ample	O
experience	O
over	O
the	O
previous	O
25	O
years	O
with	O
AI	B-Application
programs	O
that	O
would	O
generate	O
encouraging	O
early	O
results	O
but	O
then	O
fail	O
to	O
"	O
scale	O
up	O
"	O
—	O
move	O
beyond	O
the	O
'	O
training	O
set	O
 '	O
to	O
tackle	O
a	O
broader	O
range	O
of	O
cases	O
.	O
</s>
<s>
The	O
project	O
began	O
in	O
July	O
1984	O
as	O
the	O
flagship	O
project	O
of	O
the	O
400-person	O
Microelectronics	O
and	O
Computer	O
Technology	O
Corporation	O
(	O
MCC	O
)	O
,	O
a	O
research	O
consortium	O
started	O
by	O
two	O
dozen	O
large	O
United	O
States	O
based	O
corporations	O
"	O
to	O
counter	O
a	O
then	O
ominous	O
Japanese	O
effort	O
in	O
AI	B-Application
,	O
the	O
so-called	O
"	O
fifth-generation	B-Device
"	O
project.	O
"	O
</s>
<s>
The	O
objective	O
of	O
the	O
Cyc	B-Language
project	I-Language
was	O
to	O
codify	O
,	O
in	O
machine-usable	O
form	O
,	O
the	O
millions	O
of	O
pieces	O
of	O
knowledge	O
that	O
compose	O
human	O
common	O
sense	O
.	O
</s>
<s>
This	O
entailed	O
,	O
along	O
the	O
way	O
,	O
(	O
1	O
)	O
developing	O
an	O
adequately	O
expressive	O
representation	O
language	O
,	O
CycL	B-Language
,	O
(	O
2	O
)	O
developing	O
an	O
ontology	O
spanning	O
all	O
human	O
concepts	O
down	O
to	O
some	O
appropriate	O
level	O
of	O
detail	O
,	O
(	O
3	O
)	O
developing	O
a	O
knowledge	O
base	O
on	O
that	O
ontological	O
framework	O
,	O
comprising	O
all	O
human	O
knowledge	O
about	O
those	O
concepts	O
down	O
to	O
some	O
appropriate	O
level	O
of	O
detail	O
,	O
and	O
(	O
4	O
)	O
developing	O
an	O
inference	O
engine	O
exponentially	O
faster	O
than	O
those	O
used	O
in	O
then-conventional	O
expert	B-Application
systems	I-Application
,	O
to	O
be	O
able	O
to	O
infer	O
the	O
same	O
types	O
and	O
depth	O
of	O
conclusions	O
that	O
humans	O
are	O
capable	O
of	O
,	O
given	O
their	O
knowledge	O
of	O
the	O
world	O
.	O
</s>
<s>
and	O
then	O
even	O
higher-level	O
orders	O
of	O
logic	O
including	O
modal	O
logic	O
,	O
reflection	O
(	O
enabling	O
the	O
system	O
to	O
reason	O
about	O
its	O
progress	O
so	O
far	O
,	O
on	O
a	O
problem	O
on	O
which	O
it	O
's	O
working	O
)	O
,	O
and	O
context	O
logic	O
(	O
enabling	O
the	O
system	O
to	O
reason	O
explicitly	O
about	O
the	O
contexts	O
in	O
which	O
its	O
various	O
premises	O
and	O
conclusions	O
might	O
hold	O
)	O
,	O
non-monotonic	O
logic	O
,	O
and	O
circumscription	B-Application
.	O
</s>
<s>
By	O
1989	O
,	O
CycL	B-Language
had	O
expanded	O
in	O
expressive	O
power	O
to	O
higher-order	B-Algorithm
logic	I-Algorithm
(	O
HOL	O
)	O
.	O
</s>
<s>
Triplestore	O
representations	O
(	O
which	O
are	O
akin	O
to	O
the	O
Frame	O
-and-slot	O
representation	O
languages	O
of	O
the	O
1970s	O
from	O
which	O
RLL	O
sprang	O
)	O
are	O
widespread	O
today	O
in	O
AI	B-Application
.	O
</s>
<s>
It	O
may	O
be	O
useful	O
to	O
cite	O
a	O
few	O
examples	O
that	O
stress	O
or	O
break	O
that	O
type	O
of	O
representation	O
,	O
typical	O
of	O
the	O
examples	O
that	O
forced	O
the	O
Cyc	B-Language
project	I-Language
to	O
move	O
from	O
a	O
triplestore	O
representation	O
to	O
a	O
much	O
more	O
expressive	O
one	O
during	O
the	O
period	O
1984	O
–	O
1989	O
:	O
English	O
sentences	O
including	O
negations	O
(	O
"	O
Fred	O
does	O
not	O
own	O
a	O
dog	O
"	O
)	O
,	O
nested	O
quantifiers	O
(	O
"	O
Every	O
American	O
has	O
a	O
mother	O
"	O
means	O
for-all	O
x	O
there-exists	O
y	O
...	O
but	O
"	O
Every	O
American	O
has	O
a	O
President	O
"	O
means	O
there-exists	O
y	O
such	O
that	O
for-all	O
x	O
...	O
)	O
,	O
nested	O
modals	O
such	O
as	O
"	O
The	O
United	O
States	O
believes	O
that	O
Germany	O
wants	O
NATO	O
to	O
avoid	O
pursuing	O
...	O
"	O
and	O
it	O
's	O
even	O
awkward	O
to	O
represent	O
,	O
in	O
a	O
Triplestore	O
,	O
relationships	O
of	O
arity	O
higher	O
than	O
2	O
,	O
such	O
as	O
"	O
Los	O
Angeles	O
is	O
between	O
San	O
Diego	O
and	O
San	O
Francisco	O
along	O
US101.	O
"	O
</s>
<s>
Cyc	B-Language
's	O
ontology	O
grew	O
to	O
about	O
100,000	O
terms	O
during	O
the	O
first	O
decade	O
of	O
the	O
project	O
,	O
to	O
1994	O
,	O
and	O
as	O
of	O
2017	O
contained	O
about	O
1,500,000	O
terms	O
.	O
</s>
<s>
42,500	O
predicates	B-Algorithm
(	O
relations	O
,	O
attributes	O
,	O
fields	O
,	O
properties	O
,	O
functions	O
)	O
,	O
</s>
<s>
An	O
arbitrarily	O
large	O
number	O
of	O
additional	O
terms	O
are	O
also	O
implicitly	O
present	O
in	O
the	O
Cyc	B-Language
ontology	O
,	O
in	O
the	O
sense	O
that	O
there	O
are	O
term-denoting	O
functions	O
such	O
as	O
CalendarYearFn	O
(	O
when	O
given	O
the	O
argument	O
2016	O
,	O
it	O
denotes	O
the	O
calendar	O
year	O
2016	O
)	O
,	O
GovernmentFn	O
(	O
when	O
given	O
the	O
argument	O
France	O
it	O
denotes	O
the	O
government	O
of	O
France	O
)	O
,	O
Meter	O
(	O
when	O
given	O
the	O
argument	O
2016	O
,	O
it	O
denotes	O
a	O
distance	O
of	O
2.016	O
kilometers	O
)	O
,	O
and	O
nestings	O
and	O
compositions	O
of	O
such	O
function-denoting	O
terms	O
.	O
</s>
<s>
The	O
Cyc	B-Language
knowledge	O
base	O
of	O
general	O
common-sense	O
rules	O
and	O
assertions	O
involving	O
those	O
ontological	O
terms	O
was	O
largely	O
created	O
by	O
hand	O
axiom-writing	O
;	O
it	O
grew	O
to	O
about	O
1	O
million	O
in	O
1994	O
,	O
and	O
as	O
of	O
2017	O
is	O
about	O
24.5	O
million	O
and	O
has	O
taken	O
well	O
over	O
1,000	O
person-years	O
of	O
effort	O
to	O
construct	O
.	O
</s>
<s>
It	O
is	O
important	O
to	O
understand	O
that	O
the	O
Cyc	B-Language
ontological	O
engineers	O
strive	O
to	O
keep	O
those	O
numbers	O
as	O
small	O
as	O
possible	O
,	O
not	O
inflate	O
them	O
,	O
so	O
long	O
as	O
the	O
deductive	O
closure	O
of	O
the	O
knowledge	O
base	O
is	O
n't	O
reduced	O
.	O
</s>
<s>
Suppose	O
Cyc	B-Language
is	O
told	O
about	O
one	O
billion	O
individual	O
people	O
,	O
animals	O
,	O
etc	O
.	O
</s>
<s>
But	O
instead	O
of	O
that	O
,	O
one	O
could	O
tell	O
Cyc	B-Language
10,000	O
Linnaean	O
taxonomy	O
rules	O
followed	O
by	O
just	O
108	O
rules	O
of	O
the	O
form	O
"	O
No	O
mouse	O
is	O
a	O
moose	O
"	O
.	O
</s>
<s>
And	O
even	O
more	O
compactly	O
,	O
Cyc	B-Language
could	O
instead	O
just	O
be	O
given	O
those	O
10,000	O
Linnaean	O
taxonomy	O
rules	O
followed	O
by	O
just	O
one	O
rule	O
of	O
the	O
form	O
"	O
For	O
any	O
two	O
Linnaean	O
taxons	O
,	O
if	O
neither	O
is	O
explicitly	O
known	O
to	O
be	O
a	O
supertaxon	O
of	O
the	O
other	O
,	O
then	O
they	O
are	O
disjoint	O
"	O
.	O
</s>
<s>
The	O
Cyc	B-Language
inference	O
engine	O
design	O
separates	O
the	O
epistemological	O
problem	O
(	O
what	O
content	O
should	O
be	O
in	O
the	O
Cyc	B-Language
KB	O
)	O
from	O
the	O
heuristic	O
problem	O
(	O
how	O
Cyc	B-Language
could	O
efficiently	O
infer	O
arguments	O
hundreds	O
of	O
steps	O
deep	O
,	O
in	O
a	O
sea	O
of	O
tens	O
of	O
millions	O
of	O
axioms	O
)	O
.	O
</s>
<s>
To	O
do	O
the	O
former	O
,	O
the	O
CycL	B-Language
language	I-Language
and	O
well-understood	O
logical	O
inference	O
might	O
suffice	O
.	O
</s>
<s>
For	O
the	O
latter	O
,	O
Cyc	B-Language
used	O
a	O
community-of-agents	O
architecture	O
,	O
where	O
specialized	O
reasoning	O
modules	O
,	O
each	O
with	O
its	O
own	O
data	O
structure	O
and	O
algorithm	O
,	O
"	O
raised	O
their	O
hand	O
"	O
if	O
they	O
could	O
efficiently	O
make	O
progress	O
on	O
any	O
of	O
the	O
currently	O
open	O
sub-problems	O
.	O
</s>
<s>
Some	O
of	O
these	O
HL	O
modules	O
are	O
very	O
general	O
,	O
such	O
as	O
a	O
module	O
that	O
caches	O
the	O
Kleene	O
Star	O
(	O
transitive	O
closure	O
)	O
of	O
all	O
the	O
commonly-used	O
transitive	O
relations	O
in	O
Cyc	B-Language
's	O
ontology	O
.	O
</s>
<s>
CycL	B-Language
has	O
a	O
publicly	O
released	O
specification	O
and	O
dozens	O
of	O
HL	O
modules	O
were	O
described	O
in	O
Lenat	O
and	O
Guha	O
's	O
textbook	O
,	O
but	O
the	O
actual	O
Cyc	B-Language
inference	O
engine	O
code	O
,	O
and	O
the	O
full	O
list	O
of	O
1000+	O
HL	O
modules	O
,	O
is	O
Cycorp-proprietary	O
.	O
</s>
<s>
The	O
name	O
"	O
Cyc	B-Language
"	O
(	O
from	O
"	O
encyclopedia	O
"	O
,	O
pronounced	O
,	O
like	O
"	O
syke	O
"	O
)	O
is	O
a	O
registered	O
trademark	O
owned	O
by	O
Cycorp	B-Language
.	O
</s>
<s>
Access	O
to	O
Cyc	B-Language
is	O
through	O
paid	O
licenses	O
,	O
but	O
bona	O
fide	O
AI	B-Application
research	I-Application
groups	O
are	O
given	O
research-only	O
no-cost	O
licenses	O
(	O
cf	O
.	O
</s>
<s>
Typical	O
pieces	O
of	O
knowledge	O
represented	O
in	O
the	O
Cyc	B-Language
knowledge	O
base	O
are	O
"	O
Every	O
tree	O
is	O
a	O
plant	O
"	O
and	O
"	O
Plants	O
die	O
eventually	O
"	O
.	O
</s>
<s>
Most	O
of	O
Cyc	B-Language
's	O
knowledge	O
,	O
outside	O
math	O
,	O
is	O
only	O
true	O
by	O
default	O
.	O
</s>
<s>
For	O
example	O
,	O
Cyc	B-Language
knows	O
that	O
as	O
a	O
default	O
parents	O
love	O
their	O
children	O
,	O
when	O
you	O
're	O
made	O
happy	O
you	O
smile	O
,	O
taking	O
your	O
first	O
step	O
is	O
a	O
big	O
accomplishment	O
,	O
when	O
someone	O
you	O
love	O
has	O
a	O
big	O
accomplishment	O
that	O
makes	O
you	O
happy	O
,	O
and	O
only	O
adults	O
have	O
children	O
.	O
</s>
<s>
When	O
asked	O
whether	O
a	O
picture	O
captioned	O
"	O
Someone	O
watching	O
his	O
daughter	O
take	O
her	O
first	O
step	O
"	O
contains	O
a	O
smiling	O
adult	O
person	O
,	O
Cyc	B-Language
can	O
logically	O
infer	O
that	O
the	O
answer	O
is	O
Yes	O
,	O
and	O
"	O
show	O
its	O
work	O
"	O
by	O
presenting	O
the	O
step	O
by	O
step	O
logical	O
argument	O
using	O
those	O
five	O
pieces	O
of	O
knowledge	O
from	O
its	O
knowledge	O
base	O
.	O
</s>
<s>
These	O
are	O
formulated	O
in	O
the	O
language	O
CycL	B-Language
,	O
which	O
is	O
based	O
on	O
predicate	B-Application
calculus	I-Application
and	O
has	O
a	O
syntax	B-Application
similar	O
to	O
that	O
of	O
the	O
Lisp	B-Language
programming	I-Language
language	I-Language
.	O
</s>
<s>
In	O
2008	O
,	O
Cyc	B-Language
resources	O
were	O
mapped	O
to	O
many	O
Wikipedia	O
articles	O
.	O
</s>
<s>
Cyc	B-Language
is	O
presently	O
connected	O
to	O
Wikidata	O
.	O
</s>
<s>
Future	O
plans	O
may	O
connect	O
Cyc	B-Language
to	O
both	O
DBpedia	B-Application
and	O
Freebase	O
.	O
</s>
<s>
Much	O
of	O
the	O
current	O
work	O
Cyc	B-Language
continues	O
to	O
be	O
knowledge	O
engineering	O
,	O
representing	O
facts	O
about	O
the	O
world	O
by	O
hand	O
,	O
and	O
implementing	O
efficient	O
inference	O
mechanisms	O
on	O
that	O
knowledge	O
.	O
</s>
<s>
Increasingly	O
,	O
however	O
,	O
work	O
at	O
Cycorp	B-Language
involves	O
giving	O
the	O
Cyc	B-Language
system	O
the	O
ability	O
to	O
communicate	O
with	O
end	O
users	O
in	O
natural	O
language	O
,	O
and	O
to	O
assist	O
with	O
the	O
ongoing	O
knowledge	O
formation	O
process	O
via	O
machine	O
learning	O
and	O
natural-language	B-General_Concept
understanding	I-General_Concept
.	O
</s>
<s>
Another	O
large	O
effort	O
at	O
Cycorp	B-Language
is	O
building	O
a	O
suite	O
of	O
Cyc-powered	O
ontological	O
engineering	O
tools	O
to	O
lower	O
the	O
bar	O
to	O
entry	O
for	O
individuals	O
to	O
contribute	O
to	O
,	O
edit	O
,	O
browse	O
,	O
and	O
query	O
Cyc	B-Language
.	O
</s>
<s>
Like	O
many	O
companies	O
,	O
Cycorp	B-Language
has	O
ambitions	O
to	O
use	O
Cyc	B-Language
's	O
natural-language	B-Language
processing	I-Language
to	O
parse	O
the	O
entire	O
internet	O
to	O
extract	O
structured	O
data	O
;	O
unlike	O
all	O
others	O
,	O
it	O
is	O
able	O
to	O
call	O
on	O
the	O
Cyc	B-Language
system	O
itself	O
to	O
act	O
as	O
an	O
inductive	B-General_Concept
bias	I-General_Concept
and	O
as	O
an	O
adjudicator	O
of	O
ambiguity	O
,	O
metaphor	O
,	O
and	O
ellipsis	O
.	O
</s>
<s>
There	O
are	O
few	O
,	O
if	O
any	O
,	O
systematic	O
benchmark	O
studies	O
of	O
Cyc	B-Language
's	O
performance	O
.	O
</s>
<s>
The	O
concept	O
names	O
in	O
Cyc	B-Language
are	O
CycL	B-Language
terms	O
or	O
constants	O
.	O
</s>
<s>
and	O
predicates	B-Algorithm
.	O
</s>
<s>
Two	O
important	O
binary	O
predicates	B-Algorithm
are	O
#$isa	O
and	O
#$genls	O
.	O
</s>
<s>
Facts	O
about	O
concepts	O
are	O
asserted	O
using	O
certain	O
CycL	B-Language
sentences	O
.	O
</s>
<s>
Predicates	B-Algorithm
are	O
written	O
before	O
their	O
arguments	O
,	O
in	O
parentheses	O
:	O
</s>
<s>
One	O
important	O
rule	O
asserted	O
about	O
the	O
#$isa	O
predicate	B-Algorithm
reads	O
:	O
</s>
<s>
for	O
every	O
chordate	O
)	O
,	O
there	B-Algorithm
exists	I-Algorithm
a	O
female	O
animal	O
(	O
instance	O
of	O
#$FemaleAnimal	O
)	O
,	O
which	O
is	O
its	O
mother	O
(	O
described	O
by	O
the	O
predicate	B-Algorithm
#$biologicalMother	O
)	O
.	O
</s>
<s>
Each	O
microtheory	O
is	O
a	O
first-class	O
object	O
in	O
the	O
Cyc	B-Language
ontology	O
;	O
it	O
has	O
a	O
name	O
that	O
is	O
a	O
regular	O
constant	O
;	O
microtheory	O
constants	O
contain	O
the	O
string	O
"	O
Mt	O
"	O
by	O
convention	O
.	O
</s>
<s>
The	O
Cyc	B-Language
inference	O
engine	O
performs	O
general	O
logical	O
deduction	O
(	O
including	O
modus	O
ponens	O
,	O
modus	O
tollens	O
,	O
universal	O
quantification	O
and	O
existential	B-Algorithm
quantification	I-Algorithm
)	O
.	O
</s>
<s>
The	O
knowledge	O
base	O
was	O
released	O
under	O
the	O
Apache	B-Application
License	I-Application
.	O
</s>
<s>
Cycorp	B-Language
stated	O
its	O
intention	O
to	O
release	O
OpenCyc	O
under	O
parallel	O
,	O
unrestricted	O
licences	O
to	O
meet	O
the	O
needs	O
of	O
its	O
users	O
.	O
</s>
<s>
The	O
CycL	B-Language
and	O
SubL	B-Language
interpreter	O
(	O
the	O
program	O
that	O
allows	O
users	O
to	O
browse	O
and	O
edit	O
the	O
database	O
as	O
well	O
as	O
to	O
draw	O
inferences	O
)	O
was	O
released	O
free	O
of	O
charge	O
,	O
but	O
only	O
as	O
a	O
binary	O
,	O
without	O
source	O
code	O
.	O
</s>
<s>
It	O
was	O
made	O
available	O
for	O
Linux	B-Application
and	O
Microsoft	B-Application
Windows	I-Application
.	O
</s>
<s>
OpenCyc	O
4.0	O
included	O
much	O
of	O
the	O
Cyc	B-Language
ontology	O
at	O
that	O
time	O
,	O
containing	O
hundreds	O
of	O
thousands	O
of	O
terms	O
,	O
along	O
with	O
millions	O
of	O
assertions	O
relating	O
the	O
terms	O
to	O
each	O
other	O
;	O
however	O
,	O
these	O
are	O
mainly	O
taxonomic	O
assertions	O
,	O
not	O
the	O
complex	O
rules	O
available	O
in	O
Cyc	B-Language
.	O
</s>
<s>
The	O
main	O
point	O
of	O
releasing	O
OpenCyc	O
was	O
to	O
help	O
AI	B-Application
researchers	O
understand	O
what	O
was	O
missing	O
from	O
what	O
they	O
now	O
call	O
ontologies	O
and	O
knowledge	B-Application
graphs	I-Application
.	O
</s>
<s>
It	O
's	O
useful	O
and	O
important	O
to	O
have	O
properly	O
taxonomized	O
concepts	O
like	O
person	O
,	O
night	O
,	O
sleep	O
,	O
lying	O
down	O
,	O
waking	O
,	O
happy	O
,	O
etc.	O
,	O
but	O
what	O
's	O
missing	O
from	O
the	O
OpenCyc	O
content	O
about	O
those	O
terms	O
,	O
but	O
present	O
in	O
the	O
Cyc	B-Language
KB	O
content	O
,	O
are	O
the	O
various	O
rules	O
of	O
thumb	O
that	O
most	O
of	O
us	O
share	O
about	O
those	O
terms	O
:	O
that	O
(	O
as	O
a	O
default	O
,	O
in	O
the	O
ModernWesternHumanCultureMt	O
)	O
each	O
person	O
sleeps	O
at	O
night	O
,	O
sleeps	O
lying	O
down	O
,	O
can	O
be	O
woken	O
up	O
,	O
is	O
not	O
happy	O
about	O
being	O
woken	O
up	O
,	O
and	O
so	O
on	O
.	O
</s>
<s>
In	O
July	O
2006	O
,	O
Cycorp	B-Language
released	O
the	O
executable	B-Application
of	O
ResearchCyc	O
1.0	O
,	O
a	O
version	O
of	O
Cyc	B-Language
aimed	O
at	O
the	O
research	O
community	O
,	O
at	O
no	O
charge	O
.	O
</s>
<s>
In	O
addition	O
to	O
the	O
taxonomic	O
information	O
contained	O
in	O
OpenCyc	O
,	O
ResearchCyc	O
includes	O
significantly	O
more	O
semantic	O
knowledge	O
(	O
i.e.	O
,	O
additional	O
facts	O
and	O
rules	O
of	O
thumb	O
)	O
involving	O
the	O
concepts	O
in	O
its	O
knowledge	O
base	O
;	O
it	O
also	O
includes	O
a	O
large	O
lexicon	O
,	O
English	O
parsing	O
and	O
generation	O
tools	O
,	O
and	O
Java-based	O
interfaces	O
for	O
knowledge	O
editing	O
and	O
querying	O
.	O
</s>
<s>
In	O
addition	O
it	O
contains	O
a	O
system	O
for	O
ontology-based	B-General_Concept
data	I-General_Concept
integration	I-General_Concept
.	O
</s>
<s>
Cycorp	B-Language
expects	O
to	O
improve	O
and	O
overhaul	O
tools	O
for	O
external	O
developers	O
over	O
the	O
coming	O
years	O
.	O
</s>
<s>
There	O
have	O
been	O
over	O
a	O
hundred	O
successful	O
applications	O
of	O
Cyc	B-Language
;	O
listed	O
here	O
are	O
a	O
few	O
mutually	O
dissimilar	O
instances	O
:	O
</s>
<s>
For	O
over	O
a	O
decade	O
,	O
Glaxo	O
has	O
used	O
Cyc	B-Language
to	O
semi-automatically	O
integrate	O
all	O
the	O
large	O
(	O
hundreds	O
of	O
thousands	O
of	O
terms	O
)	O
thesauri	O
of	O
pharmaceutical-industry	O
terms	O
that	O
reflect	O
differing	O
usage	O
across	O
companies	O
,	O
countries	O
,	O
years	O
,	O
and	O
sub-industries	O
.	O
</s>
<s>
This	O
ontology	O
integration	O
task	O
requires	O
domain	O
knowledge	O
,	O
shallow	O
semantic	O
knowledge	O
,	O
but	O
also	O
arbitrarily	O
deep	O
common	B-General_Concept
sense	I-General_Concept
knowledge	I-General_Concept
and	O
reasoning	O
.	O
</s>
<s>
Cyc	B-Language
is	O
used	O
as	O
the	O
universal	O
interlingua	O
capable	O
of	O
representing	O
the	O
union	O
of	O
all	O
the	O
terms’	O
“	O
true	O
meanings	O
”	O
,	O
and	O
capable	O
of	O
representing	O
the	O
300k	O
transformations	O
between	O
each	O
of	O
those	O
controlled	O
vocabularies	O
and	O
Cyc	B-Language
,	O
thereby	O
converting	O
an	O
n²	O
problem	O
into	O
a	O
linear	O
one	O
without	O
introducing	O
the	O
usual	O
sort	O
of	O
“	O
telephone	O
game	O
”	O
attenuation	O
of	O
meaning	O
.	O
</s>
<s>
Furthermore	O
,	O
creating	O
each	O
of	O
those	O
300k	O
mappings	O
for	O
each	O
thesaurus	O
is	O
done	O
in	O
a	O
largely	O
automated	O
fashion	O
,	O
by	O
Cyc	B-Language
.	O
</s>
<s>
The	O
comprehensive	O
Terrorism	O
Knowledge	O
Base	O
was	O
an	O
application	O
of	O
Cyc	B-Language
in	O
development	O
that	O
tried	O
to	O
ultimately	O
contain	O
all	O
relevant	O
knowledge	O
about	O
"	O
terrorist	O
"	O
groups	O
,	O
their	O
members	O
,	O
leaders	O
,	O
ideology	O
,	O
founders	O
,	O
sponsors	O
,	O
affiliations	O
,	O
facilities	O
,	O
locations	O
,	O
finances	O
,	O
capabilities	O
,	O
intentions	O
,	O
behaviors	O
,	O
tactics	O
,	O
and	O
full	O
descriptions	O
of	O
specific	O
terrorist	O
events	O
.	O
</s>
<s>
The	O
Cleveland	O
Clinic	O
has	O
used	O
Cyc	B-Language
to	O
develop	O
a	O
natural-language	B-General_Concept
query	I-General_Concept
interface	O
of	O
biomedical	O
information	O
,	O
spanning	O
decades	O
of	O
information	O
on	O
cardiothoracic	O
surgeries	O
.	O
</s>
<s>
A	O
query	O
is	O
parsed	O
into	O
a	O
set	O
of	O
CycL	B-Language
(	O
higher-order	B-Algorithm
logic	I-Algorithm
)	O
fragments	O
with	O
open	O
variables	O
(	O
e.g.	O
,	O
"	O
this	O
question	O
is	O
talking	O
about	O
a	O
person	O
who	O
developed	O
an	O
endocarditis	O
infection	O
"	O
,	O
"	O
this	O
question	O
is	O
talking	O
about	O
a	O
subset	O
of	O
Cleveland	O
Clinic	O
patients	O
who	O
underwent	O
surgery	O
there	O
in	O
2009	O
"	O
,	O
etc	O
.	O
</s>
<s>
)	O
;	O
then	O
various	O
constraints	O
are	O
applied	O
(	O
medical	O
domain	O
knowledge	O
,	O
common	O
sense	O
,	O
discourse	O
pragmatics	O
,	O
syntax	B-Application
)	O
to	O
see	O
how	O
those	O
fragments	O
could	O
possibly	O
fit	O
together	O
into	O
one	O
semantically	O
meaningful	O
formal	O
query	O
;	O
significantly	O
,	O
in	O
most	O
cases	O
,	O
there	O
is	O
exactly	O
one	O
and	O
only	O
one	O
such	O
way	O
of	O
incorporating	O
and	O
integrating	O
those	O
fragments	O
.	O
</s>
<s>
That	O
logical	O
(	O
CycL	B-Language
)	O
query	O
is	O
then	O
converted	O
into	O
a	O
SPARQL	B-Language
query	O
that	O
is	O
passed	O
to	O
the	O
CCF	O
SemanticDB	O
that	O
is	O
its	O
data	B-General_Concept
lake	I-General_Concept
.	O
</s>
<s>
One	O
Cyc	B-Language
application	O
aims	O
to	O
help	O
students	O
doing	O
math	O
at	O
a	O
6th	O
grade	O
level	O
,	O
helping	O
them	O
much	O
more	O
deeply	O
understand	O
that	O
subject	O
matter	O
.	O
</s>
<s>
Unlike	O
almost	O
all	O
other	O
educational	O
software	O
,	O
where	O
the	O
computer	O
plays	O
the	O
role	O
of	O
the	O
teacher	O
,	O
this	O
application	O
of	O
Cyc	B-Language
,	O
called	O
MathCraft	O
,	O
has	O
Cyc	B-Language
play	O
the	O
role	O
of	O
a	O
fellow	O
student	O
who	O
is	O
always	O
slightly	O
more	O
confused	O
than	O
you	O
,	O
the	O
user	O
,	O
are	O
about	O
the	O
subject	O
.	O
</s>
<s>
The	O
user	O
's	O
role	O
is	O
to	O
observe	O
the	O
Cyc	B-Language
avatar	O
and	O
give	O
it	O
advice	O
,	O
correct	O
its	O
errors	O
,	O
mentor	O
it	O
,	O
get	O
it	O
to	O
see	O
what	O
it	O
's	O
doing	O
wrong	O
,	O
etc	O
.	O
</s>
<s>
As	O
the	O
user	O
gives	O
good	O
advice	O
,	O
Cyc	B-Language
allows	O
the	O
avatar	O
to	O
make	O
fewer	O
mistakes	O
of	O
that	O
type	O
,	O
hence	O
,	O
from	O
the	O
user	O
's	O
point	O
of	O
view	O
,	O
it	O
seems	O
as	O
though	O
the	O
user	O
has	O
just	O
successfully	O
taught	O
it	O
something	O
.	O
</s>
<s>
The	O
Cyc	B-Language
project	I-Language
has	O
been	O
described	O
as	O
"	O
one	O
of	O
the	O
most	O
controversial	O
endeavors	O
of	O
the	B-Application
artificial	I-Application
intelligence	I-Application
history	O
"	O
.	O
</s>
<s>
Catherine	O
Havasi	O
,	O
CEO	O
of	O
Luminoso	O
,	O
says	O
that	O
Cyc	B-Language
is	O
the	O
predecessor	O
project	O
to	O
IBM	B-Application
's	I-Application
Watson	I-Application
.	O
</s>
<s>
Machine-learning	O
scientist	O
Pedro	O
Domingos	O
refers	O
to	O
the	O
project	O
as	O
a	O
"	O
catastrophic	O
failure	O
"	O
for	O
several	O
reasons	O
,	O
including	O
the	O
unending	O
amount	O
of	O
data	O
required	O
to	O
produce	O
any	O
viable	O
results	O
and	O
the	O
inability	O
for	O
Cyc	B-Language
to	O
evolve	O
on	O
its	O
own	O
.	O
</s>
<s>
A	O
similar	O
sentiment	O
was	O
expressed	O
by	O
Marvin	O
Minsky	O
:	O
"	O
Unfortunately	O
,	O
the	O
strategies	O
most	O
popular	O
among	O
AI	B-Application
researchers	O
in	O
the	O
1980s	O
have	O
come	O
to	O
a	O
dead	O
end	O
,	O
"	O
said	O
Minsky	O
.	O
</s>
<s>
So-called	O
“	O
expert	B-Application
systems	I-Application
,	O
”	O
which	O
emulated	O
human	O
expertise	O
within	O
tightly	O
defined	O
subject	O
areas	O
like	O
law	O
and	O
medicine	O
,	O
could	O
match	O
users’	O
queries	O
to	O
relevant	O
diagnoses	O
,	O
papers	O
and	O
abstracts	O
,	O
yet	O
they	O
could	O
not	O
learn	O
concepts	O
that	O
most	O
children	O
know	O
by	O
the	O
time	O
they	O
are	O
3	O
years	O
old	O
.	O
</s>
<s>
“	O
For	O
each	O
different	O
kind	O
of	O
problem	O
,	O
”	O
said	O
Minsky	O
,	O
“	O
the	O
construction	O
of	O
expert	B-Application
systems	I-Application
had	O
to	O
start	O
all	O
over	O
again	O
,	O
because	O
they	O
did	O
n’t	O
accumulate	O
common-sense	O
knowledge.	O
”	O
Only	O
one	O
researcher	O
has	O
committed	O
himself	O
to	O
the	O
colossal	O
task	O
of	O
building	O
a	O
comprehensive	O
common-sense	O
reasoning	O
system	O
,	O
according	O
to	O
Minsky	O
.	O
</s>
<s>
Douglas	O
Lenat	O
,	O
through	O
his	O
Cyc	B-Language
project	I-Language
,	O
has	O
directed	O
the	O
line-by-line	O
entry	O
of	O
more	O
than	O
1	O
million	O
rules	O
into	O
a	O
commonsense	B-General_Concept
knowledge	I-General_Concept
base	I-General_Concept
.	O
</s>
<s>
Gary	O
Marcus	O
,	O
a	O
professor	O
of	O
psychology	O
and	O
neural	O
science	O
at	O
New	O
York	O
University	O
and	O
the	O
cofounder	O
of	O
an	O
AI	B-Application
company	O
called	O
Geometric	O
Intelligence	O
,	O
says	O
"	O
it	O
represents	O
an	O
approach	O
that	O
is	O
very	O
different	O
from	O
all	O
the	O
deep-learning	O
stuff	O
that	O
has	O
been	O
in	O
the	O
news.	O
”	O
This	O
is	O
consistent	O
with	O
Doug	O
Lenat	O
's	O
position	O
that	O
"	O
Sometimes	O
the	O
veneer	O
of	O
intelligence	O
is	O
not	O
enough	O
"	O
.	O
</s>
<s>
Stephen	O
Wolfram	B-Application
writes	O
:	O
</s>
<s>
Every	O
few	O
years	O
since	O
it	O
began	O
publishing	O
(	O
1993	O
)	O
,	O
there	O
is	O
a	O
new	O
Wired	O
Magazine	O
article	O
about	O
Cyc	B-Language
,	O
some	O
positive	O
and	O
some	O
negative	O
(	O
including	O
one	O
issue	O
which	O
contained	O
one	O
of	O
each	O
)	O
.	O
</s>
<s>
This	O
is	O
a	O
list	O
of	O
some	O
of	O
the	O
notable	O
people	O
who	O
work	O
or	O
have	O
worked	O
on	O
Cyc	B-Language
either	O
while	O
it	O
was	O
a	O
project	O
at	O
MCC	O
(	O
where	O
Cyc	B-Language
was	O
first	O
started	O
)	O
or	O
Cycorp	B-Language
.	O
</s>
