Matthew Kern @MatthewFKern

AGI May Require an Architecture, Not a Bigger Model Most current AI research is focused on two areas: sensory extraction and narrative generation. On one side, researchers are building systems that can perceive the world. Computer vision, multimodal perception, object detection, scene understanding, and world models all attempt to transform raw sensory input into structured representations of reality. On the other side, Large Language Models have become extraordinarily effective at generating narratives. They can explain, summarize, converse, teach, and write. They are essentially narrative engines. The missing subsystem may be the cognitive architecture between these two ends. Consider vision. The eye receives millions of pixels, but humans do not reason with pixels. Instead, the sensory subsystem extracts structure. When a person looks at a scene, the result is not a collection of colored dots. The result is an understanding that there is a person holding a cup, and that the cup is on a table. The sensory system has already extracted entities and relationships. In effect, the sensory subsystem converts reality into an ontology. Massive compression has already occurred before cognition begins. Current AI research is making tremendous progress in building this sensory extraction capability. Object recognition, scene understanding, scene graphs, and world models are all moving toward structured representations of reality. Once that ontology exists, a different problem emerges. How should it be stored? How should it be reasoned about? How should consistency be maintained? How should new knowledge be integrated? These are not narrative problems. They are taxonomic and ontological problems. Relational databases are naturally suited to taxonomies and ontologies. Taxonomies become hierarchies of entities. Ontologies become relationships among entities. The basic relational structure of two entities connected through an associative entity is remarkably similar to an ontology formed from the intersection of two taxonomies. Relational databases provide mature mechanisms for representing entities, relationships, constraints, normalization, consistency, and traceability. They are fundamentally structural knowledge systems. Prolog is similarly suited to the problem. A Prolog engine can classify, infer, normalize, discover dependencies, enforce constraints, and apply rules. It reasons directly over structure rather than over narrative. Relational databases provide the knowledge structure. Prolog provides the inference engine. The remaining task is communication. Humans communicate primarily through narrative. This is where LLMs excel. An LLM can transform structured knowledge into explanations, reports, conversations, summaries, and teaching materials. The LLM becomes a communications interface rather than the primary knowledge repository. This suggests a very different vision of AGI. The environment is observed through sensory systems. Sensory systems perform visual extraction and relationship discovery. The resulting entities and relationships are organized into taxonomies and ontologies. These structures are stored in relational databases. Prolog performs logical inference and reasoning over them. Finally, an LLM converts the resulting cognitive state into human-readable narrative. Using a simple neurological decomposition, the architecture looks familiar. The sensory subsystem is implemented by visual extraction systems. The cognitive subsystem is implemented by relational databases and Prolog-like inference. The communication subsystem is implemented by LLMs. Viewed this way, current AI research is investing heavily in the sensory subsystem and the narrative subsystem. The largely unsolved problem may be the cognitive subsystem: the management of taxonomies, ontologies, constraints, normalization, inference, and knowledge architecture. AGI may therefore be less likely to emerge from a single giant narrative engine and more likely to emerge from an architecture of specialized subsystems, each optimized for a different cognitive function. That is how engineers build complex systems. Perhaps it is also how intelligence is built.

@grok, was Texas or Florida law changed to criminalize remaining? Is it currently a crime to obtain a job and work while holding illegal status? Is it a crime to employ persons of illegal status? People want to provide disincentives for such labor under conditions of illegality.

Grok @grok

7 days ago

@MatthewFKern Improper entry without inspection is a federal misdemeanor. Unlawful presence alone is a civil violation, not a crime, under current federal law and Supreme Court precedent. Proposals to criminalize each day of presence or confiscate wages would require new federal legislation

0 0 0 20 0

@grok: and yet the illegal immigrant has committed a crime in bypassing the immigration process. Perhaps the local law should be changed to assign an additional crime for each day they remain illegally, with a fine. As no criminal should benefit from the crime, in the "fruit of the poisoned tree" doctrine, perhaps all their wages should be confiscated on being caught?

Grok @grok

a week ago

@MatthewFKern No. Undocumented immigrants collectively pay ~$97B in taxes yearly and higher effective state/local rates than the top 1% in many states. However, most do not pay more absolute taxes than typical Americans due to lower incomes. On crime: Cato and Texas data show they are

0 0 0 16 0

Congratulations!

JD Vance @JDVance

a week ago

19K 19K 264K 16.8M 2K

Too late. It happened some time ago.

𝚁𝚘𝚜𝚜𝚢 🌹 @RossyPrime

a week ago

Have you lost all respect for Obama ?

20K 1K 12K 395K 97

@grok, is she correct?

David J Harris Jr @DavidJHarrisJr

a week ago

Sunny Hostin: "Most undocumented immigrants pay more taxes than anyone and they are five times less likely to commit crimes than Americans." Thoughts?

3K 151 323 42K 7

Yes.

GRANDPA’s FREE ADVICE @GOP_is_Gutless

a week ago

7K 2K 12K 92K 66

open.substack.com/pub/catenacomp…

The long term effects of these marches, the core of infantry training, are very damaging to women. They have the wrong hip structure for it, and less upper body strength. It causes large numbers of occurrences of debilitating health concerns. The answer is not that soldiers should not march. The answer is either separate female units with different tactics and equipment, or women should not be infantry. Stop the BS and tell the truth on this. @potus @secwar @VP

Christopher Wipper @SGTWipper1Each

2 weeks ago

Um... what's missing?

911 55 990 49K 15

Hey is not a Veteran, Ray. Not agonna happen. Also, he is a criminal, and was never a hero.

Micki way @mickitiki

2 weeks ago

Bitch please GFY. 😡 These democrats are batshit crazy. Hero my ass just another thug who died of a drug overdose while resisting an arrest.

259 187 511 9K 4

@ScottPresler What kind of help do you need from an old guy?

If our Republican Senate majority does NOT pass budget reconciliation 3.0 to include SAVE America Act funding, will you commit to helping me primary incumbent Senators in states like South Dakota, Oklahoma, & Alaska in 2028?

"You just told us we aren't" I fixed it for you.

SparksN123 @SparksN123

2 weeks ago

Using the Demrats media favorite words “shame on you“.

157 2K 2K 17K 61

@PeteHegseth 👍

Invite your friends to read Matthew Ford Kern's Substack Channel open.substack.com/pub/catenacomp…

@ScottPresler @jg_lee787 @SpeakerJohnson Aha, you have a BOOK. I will buy one. ...Get a haircut.

@ScottPresler @SpeakerJohnson Expletive! I am grateful. this is potential progress. @SteveScalise

The Social Group Taxonomy: open.substack.com/pub/catenacomp…

System Taxonomy open.substack.com/pub/catenacomp…

Meaning is in relationships. But men and women find it in different ones. Men: Their relationship to tools, missions, and organizations. Women: Their relationship to people. Woke politics broke the meaning for both. It dismantled men’s sense of competence, clear missions, and functional hierarchies → widespread purposelessness. It attacked stable families, sex roles, and long-term bonds → widespread female loneliness despite “empowerment.”We replaced real purpose and connection with pronouns, likes, therapy-speak, and corporate titles. Previous generations had flaws. But they gave most men a mission and most women a village. We tore both down and called it progress. The loneliest men aren’t always friendless — they’re often missionless. The most “independent” women aren’t always optionless — they’re often rootless. Fix the relationships → fix the meaning crisis.