Read-only scholarly HTML edition
Infinitesimal Creativity
Learning by Double Involution
Contents
Navigate by part, chapter, or section.
Front matter
- Prefaceifc-0001
- A Trail Map of Infinitesimal Creativityifc-0002
Foundations of Infinitesimal Creativity
0 A Working Language for Creative Theory Change15 sections
- 0.1 Progressive abstraction as creative reconstructionifc-0004
- 0.2 Categories: typed compositionifc-0005
- 0.3 Diagrams and compositional promisesifc-0006
- 0.4 Functors and natural transformationsifc-0007
- 0.5 Universal constructions: defining by relationsifc-0008
- 0.6 Algebraic theories, sketches, and modelsifc-0009
- 0.7 Lexical creativity: extending a languageifc-0010
- 0.8 Presheaves, sheaves, and contextual agreementifc-0011
- 0.9 Tangent categories and local possibilityifc-0012
- 0.10 Weil algebras as probe shapesifc-0013
- 0.11 Classical double Lie algebroids: a concrete realizationifc-0014
- 0.12 Two axes: Boden and Piagetifc-0015
- 0.13 The geometry example as a creative traceifc-0016
- 0.14 How to read the rest of the bookifc-0017
- Further Readingifc-0018
1 Initial Sketch Acquisition13 sections
- 1.1 The initialization problemifc-0019
- 1.2 Four routes into a domainifc-0020
- 1.3 Explanation-based learning as a precedentifc-0021
- 1.4 Learning apprentices and continuous knowledge acquisitionifc-0022
- 1.5 What must be acquired from a demonstration?ifc-0023
- 1.6 Coaching as typed counter-witness selectionifc-0024
- 1.7 Conservation as developmental sketch acquisitionifc-0025
- 1.8 From acquisition to double involutionifc-0026
- 1.8.1 The readiness boundaryifc-0027
- 1.9 An initial acquisition contractifc-0028
- 1.10 A controlled acquisition benchmarkifc-0029
- 1.11 What this chapter does not assumeifc-0030
- Further Readingifc-0031
2 Creativity as Theory Construction11 sections
- 2.1 Three different achievementsifc-0032
- 2.2 Assimilation, accommodation, and cognitive developmentifc-0033
- 2.3 Two interacting classes of repairifc-0034
- 2.4 Boden’s three modesifc-0035
- 2.5 Four evaluation media, five admission regimesifc-0036
- 2.6 From LINCS repair to creative transformationifc-0037
- 2.7 The status-bearing extension dossierifc-0038
- 2.8 What counts as success?ifc-0039
- 2.9 Personal and historical creativityifc-0040
- 2.10 From a target to a mechanismifc-0041
- Further Readingifc-0042
3 The Differential Geometry of Creativity18 sections
- 3.1 The ambient differential universeifc-0043
- 3.2 Weil algebras as probe shapesifc-0044
- 3.3 The missing bridge: from tangent semantics to double algebroidsifc-0045
- 3.4 Tensor products of sketches as a semantics for doublingifc-0046
- 3.4.1 The proposed internal DIAL targetifc-0047
- 3.4.2 Formal status and empirical burdenifc-0048
- 3.5 A Weil profile of a theory statesec-weil-profile
- 3.6 Double-infinitesimal repair geometryifc-0049
- 3.7 Curve objects and the dynamics of theory revisionifc-0050
- 3.7.1 Two flows rather than oneifc-0051
- 3.7.2 From flow to a change of theoryifc-0052
- 3.8 Assimilation and accommodation in differential formifc-0053
- 3.9 The three modes in differential formifc-0054
- 3.10 Differential theory extensionsifc-0055
- 3.11 A differential creativity processifc-0056
- 3.12 Beyond two directions: the higher-DIAL conjectureifc-0057
- 3.13 What the differential account does not claimifc-0058
- Further Readingifc-0059
4 Skill Optimization for Creative Exploration14 sections
- 4.1 From a script to a creative skillifc-0060
- 4.2 The Lie algebroid of creative movesifc-0061
- 4.3 Brackets expose interactionifc-0062
- 4.4 Assimilation and accommodation in skill geometryifc-0063
- 4.5 Latent skill variationifc-0064
- 4.6 Two coupled optimization problemsifc-0065
- 4.7 DIAL-SkillOpt: training a discovery skillsec-dial-skillopt
- 4.7.1 Typed state and actionsifc-0066
- 4.7.2 One discovery episodeifc-0067
- 4.7.3 The training loopifc-0068
- 4.7.4 A preregistered evaluation designifc-0069
- 4.8 Boden’s modes as changes of control geometryifc-0070
- 4.9 Scope and experimental consequencesifc-0071
- Further Readingifc-0072
5 Algebraic Theories as Creative Targets14 sections
- 5.1 Lawvere’s functorial semanticsifc-0073
- 5.2 Sketches present theoriesifc-0074
- 5.3 From products to finite limitsifc-0075
- 5.4 PROPs and monoidal theoriesifc-0076
- 5.5 From presentations to Grothendieck toposesifc-0077
- 5.6 The creative target packagesec-creative-target-package
- 5.7 From a double repair theory to a finite theory changeifc-0078
- 5.8 Kan extensions as the transport semantics of DIALifc-0079
- 5.8.1 Transport does not choose the creative extensionifc-0080
- 5.8.2 A Beck–Chevalley interpretation of double interactionifc-0081
- 5.8.3 Left–right asymmetry and admissionifc-0082
- 5.9 Developmental towers of theoryifc-0083
- 5.10 Boden’s modes revisitedifc-0084
- Further Readingifc-0085
6 Landscapes of Computational Creativity20 sections
- 6.1 A map by creative objectifc-0086
- 6.2 Conceptual spaces and transformationsifc-0087
- 6.3 Creativity as a neurocognitive processifc-0088
- 6.4 Assimilation and accommodation across the landscapeifc-0089
- 6.5 Prediction, explanation, and structural possibilityifc-0090
- 6.5.1 Chomskyan structure and Piagetian developmentifc-0091
- 6.5.2 An empirical target, not a settled verdictifc-0092
- 6.6 Novelty, information, and surpriseifc-0093
- 6.7 Discovery as heuristic searchifc-0094
- 6.7.1 BACON: discovery from numerical regularityifc-0095
- 6.8 Local novelty, integrability, and persistenceifc-0096
- 6.9 Artifacts, process, and evaluationifc-0097
- 6.10 Imagination and counterfactual worldsifc-0098
- 6.11 Cellular automata and emergent conceptual structureifc-0099
- 6.12 Evolution as a creative constructorifc-0100
- 6.13 AI Feynman as a controlled testbedifc-0101
- 6.14 A four-level symbolic-discovery challengeifc-0102
- 6.15 Large language models in scientific discoveryifc-0103
- 6.15.1 Biological foundation models, world models, and co-scientistsifc-0104
- Further Readingifc-0105
7 Combinational Creativity9 sections
- 7.1 From juxtaposition to compositionifc-0106
- 7.2 Interface discoveryifc-0107
- 7.3 Assimilation and accommodationifc-0108
- 7.4 Differential tests for interactionifc-0109
- 7.5 Combination inside a tensor-product theoryifc-0110
- 7.6 Combinations across the AGENTIC testbedsifc-0111
- 7.7 A combinational experiment protocolifc-0112
- 7.8 Failure modes and admissionifc-0113
- Further Readingifc-0114
8 Exploratory Creativity10 sections
- 8.1 The boundary of a conceptual spaceifc-0115
- 8.2 Assimilation as explorationifc-0116
- 8.3 Local exploration with Weil probesifc-0117
- 8.4 Exploration as an integrable pathifc-0118
- 8.5 Active probe selectionifc-0119
- 8.6 Novelty, surprise, and interestingnessifc-0120
- 8.7 From exploration to obstructionifc-0121
- 8.8 Exploration across the testbedsifc-0122
- 8.9 The exploratory audit recordifc-0123
- Further Readingifc-0124
9 Transformational Creativity11 sections
- 9.1 What can be transformed?ifc-0125
- 9.2 Declaration interventionsifc-0126
- 9.3 Accommodation and equilibrationifc-0127
- 9.4 From infinitesimal evidence to a finite jumpifc-0128
- 9.4.1 Transforming a factor, interchange, or domain theoryifc-0129
- 9.5 Transport, forgetting, and conceptual lossifc-0130
- 9.6 An obstruction-to-extension workflowifc-0131
- 9.7 Transformations across the testbedsifc-0132
- 9.8 Evaluating transformational creativityifc-0133
- 9.9 The boundary of the claimifc-0134
- Further Readingifc-0135
Algorithmic Realizations
10 DIAL as an Algorithmic Family19 sections
- 10.1 The missing algorithmic layerifc-0137
- 10.2 The four-level implementation invariantifc-0138
- 10.3 Where creativity entersifc-0139
- 10.3.1 The creative deltaifc-0140
- 10.4 Three targets for every DIAL-X methodifc-0141
- 10.4.1 The artifact dossiersec-dial-artifact-dossier
- 10.4.2 Five evidential statusessec-dial-evidential-statuses
- 10.5 The DIAL-X contractifc-0142
- 10.5.1 The target cardifc-0143
- 10.6 The generic DIAL-X algorithmifc-0144
- 10.6.1 Four control actionsifc-0145
- 10.7 DIAL-GIRL: the target is an admitted decision stateifc-0146
- 10.8 DIAL-ALLORA: the target is an admitted adapter compositionifc-0147
- 10.9 CLIC: creative causal learningifc-0148
- 10.10 OPTIC: creative skill optimizationifc-0149
- 10.11 Further members of the DIAL-X familyifc-0150
- 10.12 How DIAL-X sharpens an experimentifc-0151
- 10.13 Boundary of the algorithmic claimifc-0152
- Further Readingifc-0153
12 Skill Optimization with Infinitesimal Creativity12 sections
- 12.1 The OPTIC target cardifc-0161
- 12.2 From vector-field brackets to involution algebroidsifc-0162
- 12.3 The OPTIC contractifc-0163
- 12.4 An executable OPTIC episodeifc-0164
- 12.5 Evidence for skill factorizationifc-0165
- 12.5.1 Exact mixed-witness calibrationifc-0166
- 12.5.2 The compound-skill failureifc-0167
- 12.5.3 Factored subskillsifc-0168
- 12.6 What has been learned—and what has notifc-0169
- 12.7 The open construction problemifc-0170
- 12.8 CLIC, OPTIC, and RELICifc-0171
- Further Readingifc-0172
13 Reinforcement Learning with Infinitesimal Creativity13 sections
- 13.1 What is creative about reinforcement learning?ifc-0173
- 13.2 The RELIC target cardifc-0174
- 13.3 From GIRL to RELICifc-0175
- 13.4 Schema learning in an embodied text worldifc-0176
- 13.5 The ALFWorld benchmark contractifc-0177
- 13.6 Competence of persistent structural artifactsifc-0178
- 13.7 Why novelty and one-step guidance are insufficientifc-0179
- 13.8 Admitting a supplied temporal composition (B3)sec-relic-b3
- 13.9 A DIAL reading of the resultifc-0180
- 13.10 Constructing the temporal composition (B4)sec-relic-b4
- 13.11 Residual failure and declaration scopeifc-0181
- 13.12 Constructing a missing stageifc-0182
- Further Readingifc-0183
14 Composing Creative Workflows13 sections
- 14.1 Three maintained theoriesifc-0184
- 14.2 The exchange objectifc-0185
- 14.3 A small algebra of workflow compositionifc-0186
- 14.4 Three modes of meta-creativityifc-0187
- 14.5 Cross-workflow obstructionsifc-0188
- 14.6 Canonical compositionsifc-0189
- 14.7 An integrated episodeifc-0190
- 14.8 An integrated testbedifc-0191
- 14.9 Ablations and interaction effectsifc-0192
- 14.10 A controlled demonstration: active evidence across the workflowsec-agentic-controlled-demonstration
- 14.11 What surprising outcomes would meanifc-0193
- 14.12 From orchestration sketch to workflow theoryifc-0194
- Further Readingifc-0195
AGENTIC Testbeds and Open Domains
15 Mathematical Theory Construction26 sections
- 15.1 The AGENTIC mathematical contractifc-0197
- 15.2 Historical and generated worldsifc-0198
- 15.3 A hierarchy of mathematical constructionifc-0199
- 15.4 An AGENTIC mathematical episodeifc-0200
- 15.5 Admission records and formal certificatesifc-0201
- 15.6 Persistenceifc-0202
- 15.7 Early calibration: theory extension with Leaifc-0203
- 15.8 A worked episode: inventing an intrinsic propertyifc-0204
- 15.9 From interaction laws to latent presentationstab-ipc-math-ladder
- 15.10 Symbolic laws and active mathematical experimentstab-af-math-ladder
- 15.11 Cellular-automaton theory inductionifc-0205
- 15.12 An integrated benchmarkifc-0206
- 15.13 A grand challenge: constructing a coalgebraic theory of learningifc-0207
- 15.13.1 Assimilation and accommodation over behavior functorsifc-0208
- 15.13.2 The admitted artifactifc-0209
- 15.13.3 The DIAL-URL evidence sequenceifc-0210
- 15.13.4 The synthetic laboratory behind the laddertab-dial-url-environments
- 15.13.5 The concrete sequential worldsifc-0211
- 15.13.6 What the grammar constructors mean experimentallyifc-0212
- 15.13.7 In what sense is this theory construction?ifc-0213
- 15.13.8 Finite-trajectory accommodationtab-dial-url1
- 15.13.9 Active coalgebraic discriminationtab-dial-url2
- 15.13.10 From named families to functor expressionsifc-0214
- 15.13.11 Discovering a withheld composition lawtab-dial-url4
- 15.14 Pattern repair before causal explanationifc-0215
- Further Readingifc-0216
16 Simulator-Grounded Theory Extension17 sections
- 16.1 The AGENTIC simulator contractifc-0217
- 16.2 Executable semanticsifc-0218
- 16.2.1 From relations to mechanismsifc-0219
- 16.3 Three validity layersifc-0220
- 16.4 Double involution in an executable worldifc-0221
- 16.5 The integrated loopifc-0222
- 16.6 Three Prometheus executable substratesifc-0223
- 16.6.1 Neural projection expansion in singing miceifc-0224
- 16.6.2 Hydroclimate and the Harappan transformationifc-0225
- 16.6.3 Colored airborne plastics and radiative forcingifc-0226
- 16.7 A worked theory-repair traceifc-0227
- 16.8 Ontology extensionifc-0228
- 16.9 An evidence hierarchyifc-0229
- 16.10 AGENTIC-SIM: the integrated benchmarkifc-0230
- 16.11 What would count as scientific creativity?ifc-0231
- 16.12 Open problemsifc-0232
- Further Readingifc-0233
17 From Scientific Documents to Testable Theories33 sections
- 17.1 The transition from simulation to literatureifc-0234
- 17.2 Two directions between theory and evidenceifc-0235
- 17.3 Documents as implicit sketch presentationsifc-0236
- 17.4 The AGENTIC corpus contractifc-0237
- 17.4.1 How to read the evidence sequenceifc-0238
- 17.5 From reports to local modelsifc-0239
- 17.6 Why a corpus is not a simulatorifc-0240
- 17.7 The Theory Extension Packageifc-0241
- 17.8 Controlled evidence for corpus-to-theory constructionifc-0242
- 17.8.1 Source discipline (CTTE–0)ifc-0243
- 17.8.2 Entity registration (CTTE–1)ifc-0244
- 17.8.3 Context-cover construction (CTTE–2)ifc-0245
- 17.8.4 Active overlap acquisition (CTTE–3)ifc-0246
- 17.8.5 Missing-presentation construction (CTTE–4)ifc-0247
- 17.8.6 Cross-domain transport (CTTE–5)ifc-0248
- 17.9 A corpus-to-theory evidence hierarchyifc-0249
- 17.10 The GLP-1 challengeifc-0250
- 17.10.1 Treatment status as a missing context (GLP1–PILOT–0)ifc-0251
- 17.10.2 Discovering the missing context (GLP1–PILOT–1)ifc-0252
- 17.11 A grant proposal as a capstone artifactifc-0253
- 17.11.1 Structural compilation without semantic admissionifc-0254
- 17.11.2 Semantic admission as a type systemifc-0255
- 17.11.3 Compiling prose from a typed intermediate programifc-0256
- 17.11.4 Scientific invariants across a research programifc-0257
- 17.11.5 Auditable rhetorical realizationifc-0258
- 17.11.6 Sanitized realization and surface separationifc-0259
- 17.11.7 Structure-first rhetorical realizationifc-0260
- 17.11.8 Local metric and semantic boundsifc-0261
- 17.11.9 Controlled paraphrase and evaluator failureifc-0262
- 17.11.10 From an AI-authored study to an AI-constructed programifc-0263
- 17.12 Admission beyond the corpusifc-0264
- 17.13 Toward an integrated open-domain benchmarkifc-0265
- Further Readingifc-0266
18 Generative Visual Languages27 sections
- 18.1 Diffusion is one infinitesimal directionifc-0267
- 18.2 The second direction: proto-accommodation and semantic extensionifc-0268
- 18.3 A candidate double geometryifc-0269
- 18.4 From a new token to a new languageifc-0270
- 18.5 ARTISTIC as an AGENTIC systemifc-0271
- 18.6 The ARTISTIC procedureifc-0272
- 18.7 Textual inversion as staged calibrationifc-0273
- 18.8 A visual-language benchmarkifc-0274
- 18.9 An evidence architecture for visual assuranceifc-0275
- 18.10 Lexical calibration and typed realizationifc-0276
- 18.11 Learning the declarationifc-0277
- 18.12 Transfer to natural photographsifc-0278
- 18.13 Intervals, abstention, and active acquisitionifc-0279
- 18.14 From image repair to structural assuranceifc-0280
- 18.14.1 View-conditioned contractsifc-0281
- 18.14.2 A sealed Firefly cricket challengeifc-0282
- 18.14.3 Three kinds of Nano Banana repairifc-0283
- 18.14.4 Strong generators change the opportunityifc-0284
- 18.14.5 The generator interfaceifc-0285
- 18.14.6 Movie Mode as a semantic audit traceifc-0286
- 18.14.7 Three practical axesifc-0287
- 18.15 Visual theory invention in Bongard problemsifc-0288
- 18.15.1 Five evidence regimes for concept inductionifc-0289
- 18.16 What the experiments establishifc-0290
- 18.17 Measurements and failure modesifc-0291
- 18.18 Open mathematicsifc-0292
- Further Readingifc-0293
19 Artistic Theory Extension18 sections
- 19.1 What CAN makes creativeifc-0294
- 19.2 The state space is a space of artistic theoriesifc-0295
- 19.3 Two local artistic directionsifc-0296
- 19.4 From noncommutativity to structural obstructionifc-0297
- 19.5 The Weil profile of an obstructionifc-0298
- 19.6 Constructing a candidate techniqueifc-0299
- 19.7 Transporting the old artistic theoryifc-0300
- 19.8 The admission firewallifc-0301
- 19.9 The hierarchy from CAN to DILATEifc-0302
- 19.10 Calibration of the mixed-order diagnosticifc-0303
- 19.11 A paintbrush over the space of generative dynamicsifc-0304
- 19.12 Observer-induced metric and discrete transportifc-0305
- 19.13 Registered paintbrush evidenceifc-0306
- 19.14 Theory routing across natural videosifc-0307
- 19.15 The withheld-generator experimentifc-0308
- 19.16 Baselines and ablationsifc-0309
- 19.17 Open mathematical obligationsifc-0310
- Further Readingifc-0311
20 Creativity as Compositional Theory Change14 sections
- 20.1 The argument in one viewifc-0312
- 20.2 What the present evidence supportsifc-0313
- 20.3 Why two interacting directions matterifc-0314
- 20.4 What DIAL does not explainifc-0315
- 20.5 Music as an open compositional domainifc-0316
- 20.5.1 A controlled benchmark proposalifc-0317
- 20.6 Other domains not explored hereifc-0318
- 20.7 Evaluation beyond noveltyifc-0319
- 20.8 Open obligationsifc-0320
- 20.9 Conclusionifc-0321
- Further Readingifc-0322
- Referencesreferences
- 22 Glossary of Notationifc-0323
- AA