Doug Lenat, the Cyc Project, and Permion Inc.

John F Sowa

unread,

Mar 12, 2026, 5:56:07 PMMar 12

to ontolog-forum, CG, Arun Majumdar

Doug Lenat had presented a talk at the 2022 Ontology Summit before he died in 2023, Over the years, he and I had interacted in many ways. We had different methods and different goals, but we agreed on the fundamental issues.

Since 2000, Arun Majumdar and I have been working on AI systems that put more emphasis on deriving knowledge from various sources rather than hand-coding the knowledge as in the Cyc project. From 2000 to 2010, we developed symbolic methods for automatically deriving, encoding, and reasoning about knowledge in conceptual graphs. For the current Permion Inc. company, the addition of LLM technology supports neurosymbolic methods for improved integration of natural language processing with the symbolic reasoning methods.

Since Lenat can't speak at the current summit, I recommend his Wikipedia page for a summary of his work and references to his publications: https://en.wikipedia.org/wiki/Douglas_Lenat

For a brief comparison about our different approaches, I asked Google: How did Doug Lenat and John Sowa use ontology for artificial intelligence?

Reply by Google AI:

Lenat’s primary contribution was the Cyc Project, started in 1984. His goal was to build a massive ontology of "common sense" knowledge—things a human knows but rarely says out loud, like "you can't be in two places at once".

Knowledge Representation: He developed CycL, a formal language based on predicate calculus, to hand-code millions of assertions and rules.
The "Knowledge Pump": Lenat believed that once the system reached a "crossover point" of basic common sense, it could begin learning on its own by reading textbooks and reasoning through new information.
Addressing Brittleness: By building a deep upper ontology (high-level concepts like space, time, and causality), he aimed to prevent AI from failing when faced with novel situations that lacked specific pre-programmed rules.

John Sowa focused on the formal structures and mathematical logic required to represent natural language semantics.

Conceptual Graphs (CGs): Sowa introduced Conceptual Graphs as a graphic notation for logic. These were designed to be both human-readable and computationally precise, serving as an intermediate language between natural language and machine-readable data.
Interoperability: While Lenat focused on building one massive, unified knowledge base, Sowa championed a "lattice of theories"—a framework that could reconcile and map different, even inconsistent, ontologies to one another.
Natural Language Processing (NLP): Sowa’s work emphasized using ontologies to capture the "implicit ontology" in human speech, allowing machines to map NL words and phrases directly to logical structures.

Key Comparison

Feature	Doug Lenat (Cyc)	John Sowa (Conceptual Graphs)
Primary Goal	Encoding "consensus reality" (common sense)	Representing natural language semantics
Method	Hand-coding millions of axioms in CycL	Using graphic notations for logic
Ontology Scope	Single, massive, unified knowledge base	A flexible net/lattice of multiple theories

Both researchers aimed to move AI from mere data processing to systems capable of reasoning and understanding, but they approached the problem from different, albeit complementary, angles within the symbolic AI paradigm.

Google's answer above emphasizes the relationships within the symbolic paradigm. Since then the work at Permion combines the symbolic methods with LLMs to support neuro-symbolic methods. I'll discuss the issues involved in that transition, and Arun will present examples that illustrate the methods and results.

In effect, the goals for Permion are similar to the goals for Cyc. But instead of hand-coding the knowledge, we are using LLMs and various resources derived from the WWW to provide the massive amount of knowledge that Lenat had hoped to encode in Cyc. For reasoning, Permion software uses neurosymbolic methods to derive conceptual graphs for symbolic reasoning with guaranteed precision -- no hallucinations or approximations.

John

Dan Brickley

unread,

Mar 13, 2026, 5:24:04 AMMar 13

to ontolo...@googlegroups.com, CG, Arun Majumdar

Hi John,

Thanks for this. One point I am curious about,

Cyc and “Single, massive, unified knowledge base”

Whatever became of “microtheories”?

https://web.archive.org/web/19970630114849/http://www.cyc.com/tech-reports/cyc-050-92/cyc-050-92.html

That 1992 document doesn’t have a particularly compelling answer to the question of how to deal with conflicting answers from different levels of abstraction, but at least it engaged with the issue, unlike

https://cyc.com/archives/glossary/microtheory/

The older material doesn’t sound so far from your “A flexible net/lattice of multiple theories”…

Best,

Dan

--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/19dadb6457fb40c8845297843abc985f%40d1d89cccad814b11b1deb40ea943d42b.

Alex Shkotin

unread,

Mar 13, 2026, 8:14:15 AMMar 13

to ontolo...@googlegroups.com, CG, Arun Majumdar

Hi Dan,

My question is: do they take into account existing theories in science and technology?

Then one could say that microtheory is a small part of a substantive theory.

Best,

Alex

пт, 13 мар. 2026 г. в 12:24, 'Dan Brickley' via ontolog-forum <ontolo...@googlegroups.com>:

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAFfrAFr47jxLvCa797-V2JiFM1EKnEwHzY0Mk%2Bq1h%2BV%2BK91kAQ%40mail.gmail.com.

do...@foxvog.org

unread,

Mar 13, 2026, 3:08:22 PMMar 13

to ontolo...@googlegroups.com, CG

I wrote the chapter (#12) on Cyc in Theory and Applications of Ontology:
Computer Applications, Springer, 2010. Section 12.3 discusses Cyc's
contexts, which it calls "microtheories". That would give you a basic
overview of microtheories.

Yes, Cyc does encode existing theories in science & technology in its
microtheories. They inherit more general theories and can be included
(via #$genlMt) in contexts which wish to reason using such theories.
Conflicting theories are modeled in separate microtheories so that a
Problem Solving Context can select which theories to use and not use.

A broader theory can inherit multiple more specific theories. For
example,
(#$genlMt #$NewtonsLawsOfMotionMt #$NewtonsFirstLawOfMotionMt)
(#$genlMt #$NewtonsLawsOfMotionMt #$NewtonsSecondLawOfMotionMt)
(#$genlMt #$NewtonsLawsOfMotionMt #$NewtonsSecondLawOfMotionMt)

One can select reasoning with curved or flat space, using relativistic
physics or not, using various economic or political theories, etc., by
including such theories via #$genllMt to your reasoning
microtheory/context.

Cyc separates theory and data contexts, which means one can include
general scientific contexts in the contexts of fictional works while
adding data that is only true in the context of that work. One can add
separate theory contexts to such works, such as theories of the workings
of magic or supernatural beings and occurances. Religious contexts
include both the data and supernatural theory contexts for the specified
religion.

A user of Cyc will create a context for their project and select the
theory and data contexts they wish to include as general microtheories
(#$genlMt) for their project's main context. The project can have
separate subcontexts, such as DataMicrotheories for different customers,
which have no access to each other.

-- doug foxvog
ex-ontological architect at Cycorp

>>> Lenat’s primary contribution was the Cyc Project [1], started in

>>> 1984. His goal was to build a massive ontology of "common sense"
>>> knowledge—things a human knows but rarely says out loud, like
>>> "you can't be in two places at once".
>>>

>>> * Knowledge Representation: He developed CycL, a formal language

>>> based on predicate calculus, to hand-code millions of assertions
>>> and rules.

>>> * The "Knowledge Pump": Lenat believed that once the system

>>> reached a "crossover point" of basic common sense, it could begin
>>> learning on its own by reading textbooks and reasoning through new
>>> information.

>>> * Addressing Brittleness: By building a deep upper ontology [2]

>>> (high-level concepts like space, time, and causality), he aimed to
>>> prevent AI from failing when faced with novel situations that
>>> lacked specific pre-programmed rules.
>>>
>>> John Sowa focused on the formal structures and mathematical logic
>>> required to represent natural language semantics.
>>>

>>> * Conceptual Graphs (CGs): Sowa introduced Conceptual Graphs [3]

>>> as a graphic notation for logic. These were designed to be both
>>> human-readable and computationally precise, serving as an

>>> intermediate language [4] between natural language and
>>> machine-readable data.
>>> * Interoperability: While Lenat focused on building one massive,

>>> unified knowledge base, Sowa championed a "lattice of
>>> theories"—a framework that could reconcile and map different,
>>> even inconsistent, ontologies to one another.

>>> * Natural Language Processing (NLP): Sowa’s work emphasized

>>> using ontologies to capture the "implicit ontology" in human
>>> speech, allowing machines to map NL words and phrases directly to
>>> logical structures.
>>>
>>> Key Comparison
>>>

>>> FEATURE
>>> DOUG LENAT (CYC)
>>> JOHN SOWA (CONCEPTUAL GRAPHS)

>>>
>>> Primary Goal
>>> Encoding "consensus reality" (common sense)
>>> Representing natural language semantics
>>>
>>> Method
>>> Hand-coding millions of axioms in CycL

>>> Using graphic notations [5] for logic

>>> [6].

>>
>> --
>> All contributions to this forum are covered by an open-source
>> license.
>> For information about the wiki, the license, and how to subscribe or
>>
>> unsubscribe to the forum, see http://ontologforum.org/info
>> ---
>> You received this message because you are subscribed to the Google
>> Groups "ontolog-forum" group.
>> To unsubscribe from this group and stop receiving emails from it,
>> send an email to ontolog-foru...@googlegroups.com.
>> To view this discussion visit
>>

> https://groups.google.com/d/msgid/ontolog-forum/CAFfrAFr47jxLvCa797-V2JiFM1EKnEwHzY0Mk%2Bq1h%2BV%2BK91kAQ%40mail.gmail.com
>> [7].

>
> --
> All contributions to this forum are covered by an open-source license.
> For information about the wiki, the license, and how to subscribe or
> unsubscribe to the forum, see http://ontologforum.org/info
> ---
> You received this message because you are subscribed to the Google
> Groups "ontolog-forum" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to ontolog-foru...@googlegroups.com.
> To view this discussion visit

> https://groups.google.com/d/msgid/ontolog-forum/CAFxxROTPi4CJ8o-L7L8FziSpVDw8YOD04i%2BhG%3D9QXkL0Wo7QKQ%40mail.gmail.com
> [8].
>
>
> Links:
> ------
> [1]
> https://www.google.com/url?sa=i&source=web&rct=j&url=https://en.wikipedia.org/wiki/Cyc&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQIBBAB&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [2]
> https://www.google.com/url?sa=i&source=web&rct=j&url=https://mitosystems.com/an-ontology-of-everything/&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQIBRAD&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [3]
> https://www.google.com/url?sa=i&source=web&rct=j&url=https://en.wikipedia.org/wiki/Conceptual_graph&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQICRAB&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [4]
> https://www.google.com/url?sa=i&source=web&rct=j&url=http://www.jfsowa.com/cg/cgonto.htm&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQICRAC&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [5]
> https://www.google.com/url?sa=i&source=web&rct=j&url=http://www.jfsowa.com/cg/cgif.htm&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQICxAC&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [6]
> https://groups.google.com/d/msgid/ontolog-forum/19dadb6457fb40c8845297843abc985f%40d1d89cccad814b11b1deb40ea943d42b?utm_medium=email&utm_source=footer
> [7]
> https://groups.google.com/d/msgid/ontolog-forum/CAFfrAFr47jxLvCa797-V2JiFM1EKnEwHzY0Mk%2Bq1h%2BV%2BK91kAQ%40mail.gmail.com?utm_medium=email&utm_source=footer
> [8]
> https://groups.google.com/d/msgid/ontolog-forum/CAFxxROTPi4CJ8o-L7L8FziSpVDw8YOD04i%2BhG%3D9QXkL0Wo7QKQ%40mail.gmail.com?utm_medium=email&utm_source=footer

do...@foxvog.org

unread,

Mar 13, 2026, 3:29:25 PMMar 13

to ontolo...@googlegroups.com, CG

Google AI is erroneous. Cyc's microtheories still exist.

Sometime after i left, however, they made what i consider to be a major
mistake in moving a lot of type instance assertions up from data
miccrotheories up to the base knowlege base (#$BaseKB). I am not a
human in all possible worlds, nor is Bilbo Baggins a hobbit in all
possible worlds, so such assertions belong in appropriate data
microtheories, not in the #$BaseKB. In #$BaseKB they should be nothing
more specific than #$Individuals, imho. They don't have mass in all
possible worlds, they don't have a starting time, ....

-- doug foxvog

>> Lenat’s primary contribution was the Cyc Project [1], started in

>> 1984. His goal was to build a massive ontology of "common sense"
>> knowledge—things a human knows but rarely says out loud, like "you
>> can't be in two places at once".
>>

>> * Knowledge Representation: He developed CycL, a formal language

>> based on predicate calculus, to hand-code millions of assertions and
>> rules.

>> * The "Knowledge Pump": Lenat believed that once the system reached

>> a "crossover point" of basic common sense, it could begin learning
>> on its own by reading textbooks and reasoning through new
>> information.

>> * Addressing Brittleness: By building a deep upper ontology [2]

>> (high-level concepts like space, time, and causality), he aimed to
>> prevent AI from failing when faced with novel situations that lacked
>> specific pre-programmed rules.
>>
>> John Sowa focused on the formal structures and mathematical logic
>> required to represent natural language semantics.
>>

>> * Conceptual Graphs (CGs): Sowa introduced Conceptual Graphs [3] as

>> a graphic notation for logic. These were designed to be both
>> human-readable and computationally precise, serving as an

>> intermediate language [4] between natural language and
>> machine-readable data.
>> * Interoperability: While Lenat focused on building one massive,

>> unified knowledge base, Sowa championed a "lattice of theories"—a
>> framework that could reconcile and map different, even inconsistent,
>> ontologies to one another.

>> * Natural Language Processing (NLP): Sowa’s work emphasized using

>> ontologies to capture the "implicit ontology" in human speech,
>> allowing machines to map NL words and phrases directly to logical
>> structures.
>>
>> Key Comparison
>>

>> FEATURE
>> DOUG LENAT (CYC)
>> JOHN SOWA (CONCEPTUAL GRAPHS)
>>

>> Primary Goal
>> Encoding "consensus reality" (common sense)
>> Representing natural language semantics
>>
>> Method
>> Hand-coding millions of axioms in CycL

>> Using graphic notations [5] for logic

>> [6].

>
> --
> All contributions to this forum are covered by an open-source license.
> For information about the wiki, the license, and how to subscribe or
> unsubscribe to the forum, see http://ontologforum.org/info
> ---
> You received this message because you are subscribed to the Google
> Groups "ontolog-forum" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to ontolog-foru...@googlegroups.com.
> To view this discussion visit

> https://groups.google.com/d/msgid/ontolog-forum/CAFfrAFr47jxLvCa797-V2JiFM1EKnEwHzY0Mk%2Bq1h%2BV%2BK91kAQ%40mail.gmail.com
> [7].

>
>
> Links:
> ------
> [1]
> https://www.google.com/url?sa=i&source=web&rct=j&url=https://en.wikipedia.org/wiki/Cyc&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQIBBAB&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [2]
> https://www.google.com/url?sa=i&source=web&rct=j&url=https://mitosystems.com/an-ontology-of-everything/&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQIBRAD&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [3]
> https://www.google.com/url?sa=i&source=web&rct=j&url=https://en.wikipedia.org/wiki/Conceptual_graph&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQICRAB&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [4]
> https://www.google.com/url?sa=i&source=web&rct=j&url=http://www.jfsowa.com/cg/cgonto.htm&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQICRAC&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [5]
> https://www.google.com/url?sa=i&source=web&rct=j&url=http://www.jfsowa.com/cg/cgif.htm&ved=2ahUKEwjkyq_GwrSSAxWbM1kFHdLJDm8Qy_kOegQICxAC&opi=89978449&cd&psig=AOvVaw14x8hF2Pe6GlFR3YCotsiR&ust=1769905668179000
> [6]
> https://groups.google.com/d/msgid/ontolog-forum/19dadb6457fb40c8845297843abc985f%40d1d89cccad814b11b1deb40ea943d42b?utm_medium=email&utm_source=footer
> [7]

> https://groups.google.com/d/msgid/ontolog-forum/CAFfrAFr47jxLvCa797-V2JiFM1EKnEwHzY0Mk%2Bq1h%2BV%2BK91kAQ%40mail.gmail.com?utm_medium=email&utm_source=footer

Alex Shkotin

unread,

Mar 14, 2026, 6:06:24 AMMar 14

to ontolo...@googlegroups.com, CG

Doug,

Thanks for the comprehensive answer: "Cyc does encode existing theories in science & technology in its microtheories."

Is there any chance of getting a specific microtheory? For example, #$NewtonsLawsOfMotionMt or #$GeometryGMt.

Perhaps there's a publicly available list of all existing microtheories? I read somewhere that there are 6,000 of them.

To be completely happy, I'd like to think of a microtheory as a formal theory, that is, a formalization of some scientific or technological theory.

6,000 formalized theories—that's impressive.

Meanwhile, I'm looking at systematization, concentration, and formalization, Hilbert Geometry (to compare with Geocoq), and Newtonian Mechanics (to compare with Cyc?).

I am ready to add a line for CycL in the Geometry framework. Like this

I21 axiom

rus	Пусть A, B - точки. Если A и B различны то существует единственная прямая которая соединяет A с B.
eng	Let A and B be points. If A and B are distinct, then there exists a unique straight line that connects A to B.
yfl	∀p1,p2:Po p1≠p2 → (∃1l:SL connects(l p1 p2)).
coq	line_existence : ∀ A B, A ≠ B → ∃ l, Incid A l ∧ Incid B l; line_uniqueness : ∀ A B l m, A ≠ B → Incid A l → Incid B l → Incid A m → Incid B m → EqL l m;
cyc	???

Alex

пт, 13 мар. 2026 г. в 22:08, <do...@foxvog.org>:

To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/e4a78cde65f6cf7f97fe9471f9be1b72%40foxvog.org.

Reply all

Reply to author

Forward