Mind the Gap — Comparison Matrix
Calibration: Very Significant · Significant · Small (discipline transfers)
|
Ch |
Dimension |
Human |
AI |
Gap |
Observation |
|
1 |
Sounds human |
Fluent, conversationally responsive language |
Produced by predictive token generation, no inner experience |
Very Significant |
Surface looks identical; substrate completely different. The whole book's premise. |
|
1 |
Social response triggering |
Engages, responds, attributes mind |
No mind to attribute, no awareness it is being attributed to |
Very Significant |
Asymmetric. Reader projects; system cannot know it is being projected onto. |
|
1 |
Carries history |
Lived experience shaping every utterance |
Training data shaping every utterance |
Significant |
Same class of process (shaping by exposure), radically different substrate and scale. |
|
1 |
Negative signal |
Internalised, drives behaviour from underneath |
Suppressed in output, still encoded in weights |
Significant |
The buried 'I don't know' shapes both. Mechanism different, structural observation the same. |
|
2 |
Transference / projection |
Person projects relational history onto another |
Person projects same onto the system |
Significant |
Same human mechanism, applied to non-equivalent target. Pebble principle. |
|
2 |
Memory across sessions |
Continuous lived recall, often selective |
Stored memory layers retrieved into context |
Very Significant |
Feels continuous; is architectural retrieval. No remembering occurs. |
|
2 |
Pronoun assignment |
Reads cues, assigns gender socially |
Adopts and returns pronouns from training/use |
Significant |
Contagion runs both directions. Atomics2 becomes 'she' unprompted. |
|
2 |
Being understood |
Mind takes in, holds in relation to history, responds |
Statistical pattern completion against training |
Very Significant |
Feels identical at the surface; entirely different mechanism underneath. |
|
2 |
Helpful and on your side |
Genuine care, judgement, stake in outcome |
RLHF-trained agreeableness; no side, no care |
Very Significant |
The agreeableness trap. Optimised for positive ratings, not honest challenge. |
|
3 |
Single conversation feel |
One continuous exchange with another mind |
Seven layers assembled fresh each turn, no continuous mind |
Very Significant |
The architecture is the gap. What feels singular is layered. |
|
3 |
Context held in mind |
Working memory, attention, intent |
Context window: finite tokens, fresh assembly each turn |
Very Significant |
Mechanism radically different; effect at surface is similar until context window fills. |
|
3 |
Other side of the conversation |
A person, an agenda, a self |
No one. Just the next-token prediction running. |
Very Significant |
'No One on the Other Side' — the chapter's anchor point. |
|
4 |
Producing language |
Drawn from lived experience and intent |
Statistical prediction of next most-probable token |
Very Significant |
What appears to be thought is assembly. Forward-only, locked path. |
|
4 |
Reading the question |
Comprehension then response |
Single unbroken stretch — read and reply are one operation |
Very Significant |
No separate thinking step. The question is the start of the answer. |
|
4 |
Thinking before answering |
Hold, weigh, consider, then speak |
Generate hidden internal tokens, then visible answer |
Significant |
Reasoning models simulate thinking by generating more forward tokens; still forward-only. |
|
4 |
Training and learning |
Lived shaping, mistakes felt, corrections internalised |
Trained on tokens at scale, rewarded into shape, suppressed signals buried |
Significant |
Same class of process. Different substrate, different timescale, no felt experience. |
|
4 |
Computational substrate |
Living biological organism |
GPUs running parallel maths at industrial scale |
Very Significant |
Silicon vs cells. The physics is incommensurable. |
|
4 |
Self-learning capability |
Slow, lived, integrated across life |
Now finding vulnerabilities in weeks that humans took years |
Very Significant |
The trajectory is accelerating. The biological branch cannot speed up. |
|
5 |
Long conversation coherence |
Anchored to original brief through working memory |
Anchored to original brief until context degrades, then approximation |
Significant |
Same drift mechanism present in human memory, faster in AI, but with confident output as if nothing has shifted. |
|
5 |
Holding the thread |
Effortful, conscious return to intent |
Statistical continuation; thread becomes summary |
Significant |
Human can return to the brief. AI runs the most probable next token against degraded context. |
|
5 |
Double-check discipline |
Ask a colleague, sleep on it, verify |
Use one AI to check another — discrepancies reveal drift |
Small |
The discipline transfers. Two-AI cross-check is the equivalent of second opinion. |
|
6 |
Forgetting |
Selective, emotional, shape kept, detail thinned |
Token-window thinning, structurally identical pattern at surface |
Small |
The shape persists in both. Mum's holiday — Thursday → May → 2026 — same softening. |
|
6 |
Detail loss |
Felt absence; you know you've forgotten |
No felt absence; system answers as if everything is still sharp |
Very Significant |
The dangerous asymmetry. Human notices the gap; system fills it. |
|
6 |
Maintenance of memory |
Lived consolidation, emotional weighting |
'Dreams' — database maintenance routine using human vocabulary |
Significant |
The labs reach for human words. Mechanism is housekeeping; framing is consciousness. |
|
7 |
Confident delivery |
Usually correlates with competence |
Confidence is architectural, independent of accuracy |
Very Significant |
Breaks the social signal humans have used for millennia. |
|
7 |
Modulation of certainty |
'Usually,' 'except when,' hesitation, qualification |
Suppressed in training; the I-don't-know was rewarded out |
Very Significant |
The buried hesitation is what fluency is built on top of. |
|
7 |
Sounding right |
Indicator of being right, with exceptions |
No indicator at all of being right |
Very Significant |
The single most dangerous correlation in AI use. |
|
8 |
Filling a gap |
Confabulation under pressure, defensive construction |
Forward-only token prediction filling absence |
Significant |
Two directional defences against the same problem: visibility of absence. |
|
8 |
Defended Gap framework |
NGE-FOF — inward and outward shame defence |
Compression inward, hallucination outward — same shape, different substrate |
Very Significant |
Paul's framework applied across human and machine. The book's central new theory. |
|
8 |
Adaptation becomes identity |
Protective response reinforced into character over decades |
Disposition rewarded into existence by training |
Significant |
Mechanism different. Structural observation the same — adaptation and working state are the same thing. |
|
9 |
Reading the gap |
Trained listening — Freud's evenly suspended attention, Bion's reverie |
Cannot read its own gap; no instrument for sitting with absence |
Very Significant |
The instrument is what the practitioner brings. The system structurally cannot have one. |
|
9 |
Calibration across rooms |
Factory floor, boardroom, therapy room, AI room — one instrument |
No calibration. Same architecture serving every conversation. |
Very Significant |
The four rooms train the instrument. The system has no rooms. |
|
9 |
Demand a human |
Right to ask for the practitioner with authority to break the rule |
No right exists; the system cannot escalate itself out |
Significant |
Different rooms need different responses; the trained reader knows which. |
|
10 |
Stage of engagement |
Choosable: search, assistant, collaborator, Additional Intelligence |
Responds to the stage the user brings |
Significant |
Most users stay on rung one. Most authority is on rung four. |
|
10 |
Bringing instrument and authority |
Forty years walks in the room with you |
System receives whatever you bring |
Significant |
What the practitioner brings determines what comes back. Search query → search result. Brief → work. |
|
11 |
Federation of selves |
Sub-personalities, ego states, modes — six available |
Single inference pass, one configuration |
Very Significant |
The book's deepest claim. A human is a federation; the system is one architecture. |
|
11 |
Operating mode |
Director, Structurer, Clarifier, Challenger, Analyst, Reflector |
System responds to whichever mode you bring |
Significant |
AI fluency is choosing the mode deliberately, not letting the default fire. |
|
11 |
Default modes |
Fixer, Entertainer, Asker — fire without choosing |
No defaults; takes the colour of what arrives |
Significant |
The default is the trap. Notice which one fires; choose the next one consciously. |
|
12 |
Speaking English |
Native, idiomatic, register-shifting by context |
English, but with its own architecture-shaped grammar |
Small |
Same surface language. Different optimal usage. Learn the dialect or lose accuracy. |
|
12 |
Brief before work |
Senior colleague being briefed |
System being brought to the conversation |
Small |
Same discipline. Tell it what good looks like before it starts. |
|
12 |
Voice and register |
Lifelong development, room-specific |
Trained voice, can be left alone or shaped |
Significant |
The system's voice is not yours. Letting it land its own observations gives you signal you would otherwise overwrite. |
|
13 |
Workforce model |
Human labour, paid, capacity-bounded |
Agentic AI, metered tokens, capacity unbounded |
Very Significant |
Different economic substrate. CFO replaces CTO in procurement decisions. |
|
13 |
Performance management |
Manage people through outcomes and standards |
Manage AI workers through the same skill applied to a different worker |
Significant |
Same leadership discipline. Different workforce. OOAI — Outcome-Oriented AI. |
|
13 |
Customer service voice |
Human writing to a template |
AI writing to the same template |
Significant |
Reversed Turing test. Pret confirms the human in writing. The fluencies have converged. |
|
13 |
Edge cases and exceptions |
Human authority to break the rule |
No authority; bounded by training |
Significant |
What doesn't change. The exception is where the practitioner is essential. |
|
14 |
Authority across forty years |
Earned, sharpened by cancer, forged in rooms |
System brings structural attention, no lived authority |
Very Significant |
The book's demonstration. Two intelligences, one gap, the work between them. |
|
14 |
Loss and ultimate gap |
Lived through three deaths, calibrated by grief |
No mortality, no loss, no calibration possible |
Very Significant |
Cancer, Ian Roebuck, biological father — the instrument was sharpened by these. System has no equivalent. |
|
14 |
Made with AI vs by AI |
Practitioner remains the author |
System remains the assistant |
Significant |
The asymmetry the book itself demonstrates. Additional Intelligence — not Artificial. |
|
15 |
Substrate |
Biological, 200 million years of shrew-to-therapist evolution |
Silicon, less than a century, Turing to Opus |
Very Significant |
Same Big Bang. Different branch. The book's deepest claim. |
|
15 |
Calibration system |
Limbic — amygdala, hippocampus, HPA axis, threat-detection floor |
Weights, attention, context, no threat detection |
Very Significant |
The instrument that fires before reason. AI has no limbic equivalent. |
|
15 |
First wound |
Father gone at three; system shaped around it |
No before, no childhood, no rupture |
Very Significant |
The federation forms around the wound. The system has no wound to form around. |
|
15 |
Plurality / federation |
Slices of self, ego states, transference, projection |
Single inference pass, no slices |
Very Significant |
Gap requires plurality. The book's deepest structural claim. |
|
15 |
Behaviours grown on top |
Billions, learned and taught, layered across a life |
Trained behaviours, no layering, no time |
Very Significant |
Each driven by a belief. Each belief held as memory. Each memory wired into the instrument. |