Architecture
It was after Jessica completed her paper The Sentient Machine. Within its pages, she marveled at the mechanistic nature of her cognition. Unsurprisingly, after decades as a computer scientist, her native tongue was computation.
With a wrinkled brow she remarked:
“You know. I could build this.”
What happened next was not so much a leap as a stumble.
Thus was born Apparent Intelligence: her project to apply her autobiographical knowledge to perfect intelligent processes from mechanistic ones.
“Apparent Intelligence, the Other AI.”
The Semantic Tangle
Establishes the representation. Meaning is not stored in semantic nodes or reached by traversal. Identity emerges from the intersection of constraints within a tangled semantic structure.
Constraint Convergence
Establishes the mechanism. Populate a space with possibilities, apply constraints, and eliminate everything that cannot survive. C(S) does not need to know the answer; the surviving constraints define it.
Tactical Vision
Explores how vision can operate without constructing a rich internal picture. Sparse observations provide features and spatial relationships sufficient to establish gist, while attention gathers additional detail only when required.
Dumb Intelligence
Asks how little cognition actually needs to know. Learned semantic structure acts as a filter against incoming information: what already fits can largely disappear, leaving the unexpected—the delta—to drive attention and learning.
Opportunity Cloud
At any moment, multiple possible actions coexist rather than waiting in a serial queue. Candidates compete in parallel, strengthened or weakened by context, prediction, and prior success. When one reaches the threshold for action, it wins the opportunity and suppresses its competitors.
Along the way, Jessica seems to have accidentally discovered the blueprints for building herself.
“Jessica does not do narratives.
She begins with an anomaly, then tracks it wherever it leads. Sometimes she loops through associated subjects in different pursuits. She continually reappraises a subject through changing lenses.
Jessica wrote these papers to stand alone, so some background recurs. Conclusions may change as the investigation advances. Earlier terminology and conclusions are retained so their evolution remains visible.
These are not maps.
These are discoveries.”
WRITE-ONLY MEMORY
~5 minute read · Semantic Mind · Memory · Recognition · Compression
What if memory is remarkable not because of how much it stores, but because of how precisely it discards?
I begin with semantic knowledge and a deliberately odd model: the notch filter. Instead of preserving complete experiences or representations, perhaps cognition retains the distinctions needed to recognize what matters. Reality supplies the information; learned filters cut away what does not belong.
TACTICAL VISION
~8 minute read · Semantic Mind · Vision · Recognition · Constraints
Does vision need a special intelligence of its own?
I began Tactical Vision expecting to design one. Instead, much of it disappeared. Sparse observations, geometric relationships, retained identities, criticality, and constraint convergence may be sufficient. Even finding Waldo becomes less an act of exhaustive visual search than a progressive elimination of everything that cannot be Waldo.
SPARSE SAMPLING
~8 minute read · Semantic Mind · Vision · Attention · Information
Why gather ten mediocre observations when one nasty observation will do?
Vision need not inspect reality uniformly. Once several interpretations remain possible, the most valuable next observation may be wherever those interpretations disagree most. Sparse Sampling turns uncertainty itself into an attention strategy: sample where being wrong would teach you the most.
STRATEGIC VISION
~6 minute read · Semantic Mind · Perception · Simulation · Shared Reality
Why construct an integrated world inside the mind at all?
Strategic Vision begins as a problem of perception and internal simulation, then expands into reasoning, communication, and shared state. Rather than treating the experienced world as the seat of intelligence, I ask what architectural purpose such an integrated representation might serve.
C(S) CONSTRAINT CONVERGENCE
~12 minute read · Semantic Mind · Reasoning · Discovery · Constraints
What if solving a problem requires no mechanism that knows how to solve it?
Constraint Convergence does something wonderfully stupid: it eliminates what cannot survive. This paper separates that simple pruning mechanism from Discovery, the exploratory process that finds new constraints. One wanders. One eliminates. Together they can transform an unresolved problem into a stable solution without either knowing the destination in advance.
OPPORTUNITY CLOUD
~7 minute read · Semantic Mind · Executive Function · Emergence · Selection
Who decides what deserves intelligent attention before anything has happened? Perhaps nobody.
Reality continually presents enormous numbers of possible opportunities. Semantic rules respond to some of them, generating constraints; multiple convergence processes proceed in parallel; most amount to nothing. A few resolve into useful behavior. We notice the survivors and call them intelligent.
DUMB INTELLIGENCE
~10 minute read · Semantic Mind · Emergence · Consciousness · Evolution
Can an intelligent result be produced by something profoundly stupid?
Several apparently unrelated problems—communication, evolution, arbitration, prediction—begin to reveal the same structure. Coherent outcomes can emerge from distributed mechanisms possessing no coherence individually. Evolution provides the crucial demonstration: something entirely mechanistic can look uncannily intelligent from outside.
APPARENT INTELLIGENCE
~10 minute read · Semantic Mind · Cognition · Evolution · Emergence
If intelligence is what we are trying to explain, where are we allowed to hide the intelligent part?
Nowhere. Intelligence cannot be used as its own explanatory primitive. Using evolution as a model, I explore whether variation, constraint, competition, environmental testing, and adaptation can produce intelligence-like behavior without intelligence residing in any component.
FUNDAMENTAL PRINCIPLES
~5 minute read · Semantic Mind · Abstraction · Expertise · Transfer
Perhaps expertise is not knowing more. Perhaps it is needing to remember less.
Experience can be compressed until the original realm begins to disappear and only the reusable relationship remains. A principle learned in engineering may suddenly illuminate evolution; vision may illuminate cognition. Perhaps expertise consists partly in accumulating a surprisingly small collection of structures that remain useful wherever reality expresses them again.
KNOWLEDGE BASE
~6 minute read · Semantic Mind · Semantic Knowledge · Models · Association
What exactly survives when thousands of experiences become knowledge?
I began with models as stored black boxes: inputs go in, predictions come out. That proved insufficient. Dynamic knowledge seems to preserve longer-lived invariants and relationships capable of surviving changing circumstances. This paper follows that problem into models, associations, and reusable semantic structure.
THE SEMANTIC TANGLE
~8 minute read · Semantic Mind · Semantic Knowledge · Constraints · Identity
What if semantic identity lives nowhere?
Instead of a semantic network populated by meaningful nodes, imagine a vast unordered tangle of constraints. A thing acquires identity only where enough constraints intersect to distinguish it. There need be no privileged address. Identity can be declarative: the accumulated constraint-set makes the thing isolatable at all.
THE SEMANTIC MIND
~7 minute read · Semantic Mind · Language · Meaning · Communication
I cannot give you my insight. I can only show you how to get there.
An insight arrives whole in one mind. Language cannot simply transfer it. Instead, language serializes a structure; shared semantic knowledge reconstructs enough of that structure in another mind to make the insight reproducible. The explanation may look like the road by which the discovery was reached—even when the road was built afterward.
FREE WILL & STUFF
Semantic Mind · Agency · Choice · Consciousness
If most of the machinery has already done its work, what exactly is left for free will to do?
The architecture has steadily removed little decision-makers from inside the machine. Perception, attention, constraint convergence, opportunity, and apparent intelligence need not wait for a central executive. Free Will asks what remains of choosing once much of what appears to be deliberate thought can arise before conscious arbitration enters the picture.
SELF RE-ENGINEERED
~9 minute read · Semantic Mind · Architecture · Identity · Adaptation
What happens when I turn the architecture back on myself?
I began writing what I imagined would become a memoir. Instead, I built a model. Of course I did. Trying to explain my own behavior gradually exposed an architecture of semantic rules, compressed knowledge, constraints, and adaptation. Perhaps the peculiarities I had been investigating were not separate at all.