Research & Knowledge / AI Knowledge Atlas
AI Knowledge Atlas: Foundations, Techniques, Systems and Applications
Nested, expandable research map that separates disciplines, scientific techniques, technical systems, operational environments, governance and application domains. The directory is designed for breadth and progressive source validation, not as a claim that the whole field has already been exhaustively documented.
Research continuity: This atlas organises longstanding study of AI, computing and technical systems; it does not imply these subjects have only recently been undertaken. The 2017 early-years programming and robotics work is separately recorded in the systems atlas.
7 broad domains · 35 initial research topics
1. Foundations and paradigms · 5 topics
Reasoning and problem-solving
Search, deduction, heuristics and resource-bounded inference.
Knowledge representation
Symbols, ontologies, semantic networks and knowledge graphs.
Planning and decision making
Goal selection, utility, partial observability and feedback.
AI history and philosophy
Origins, paradigms, competing definitions and conceptual limitations.
Mathematical foundations
Linear algebra, probability, calculus, information and optimisation.
2. Learning and computation · 6 topics
Supervised and unsupervised learning
Labelled prediction and pattern discovery from unlabelled data.
Reinforcement and transfer learning
Rewards, policies, generalisation and task transfer.
Classifiers and statistical learning
Trees, regression, neighbours, kernels and ensembles.
Artificial neural networks
Feedforward, convolutional, recurrent, graph and attention-based networks.
Deep learning and foundation models
Multilayer representation learning, transformers and generative training.
Learning theory and evaluation
Sample complexity, generalisation, distribution shifts and uncertainty.
3. Languages, perception and social interaction · 5 topics
Natural language processing
Text understanding, generation, translation, extraction and retrieval.
LLMs and GPT systems
Token prediction, context, post-training, inference and hallucination risks.
Vision and multimodal perception
Recognition, detection, tracking, sensor fusion and spatial inference.
Speech and audio
Recognition, synthesis, diarisation, sound and music analysis.
Social intelligence and HCI
Social signals, affective computing, accessibility and human oversight.
4. Search and formal reasoning · 4 topics
State-space and adversarial search
Graph/tree search, A*, heuristics and strategic competition.
Local and evolutionary optimisation
Gradient descent, population search, swarm optimisation and constraints.
Formal logic
Propositional, predicate, fuzzy, modal and non-monotonic inference.
Probabilistic and causal reasoning
Bayesian networks, MDPs, causal graphs and counterfactuals.
5. Agents, robotics and autonomy · 4 topics
AI agents and orchestrators
Tool use, planning loops, memory, permissions, handoffs and multi-agent coordination.
Robotics and physical AI
Sensors, actuators, localisation, SLAM, manipulation and control.
Autonomous systems
Degrees of autonomy, safety interlocks, supervision and recovery.
Artificial general intelligence
Broad competence claims, evaluation limits and unsettled definitions.
6. Infrastructure and engineering · 5 topics
Hardware and software
CPU/GPU, accelerators, operating systems, frameworks and edge computing.
Data engineering and provenance
Pipelines, quality, lineage, storage, metadata and access control.
APIs and integration
REST, event-driven architectures, RAG, agent protocols and messaging.
Cloud and MLOps
Deployment, monitoring, cost, reproducibility, model lifecycle and resilience.
Testing and technical QA
Security testing, evaluation, software regression, model validation and benchmarks.
7. Domains, governance and impact · 6 topics
Healthcare, science and mathematics
Discovery, diagnostics, proof support, experiments and clinical safeguards.
Government, justice and education
Public service use cases, administrative systems and institutional accountability.
Business, labour and finance
Enterprise automation, employment shifts, financial operations and risk.
Creative and cultural applications
Media, recommendation, design, translation and creative production.
AI ethics, safety and law
Fairness, privacy, security, oversight, consent and rights impacts.
Research gaps and comparative systems
Evidence audits, novelty tests, architecture variation and unresolved questions.
Cross-references and research methods
Industry systems atlas (20 industry groups) · Research-agent demonstration · 20-workstream implementation register · 28 future ventures
Sources for continued breadth and validation
UK AI Adoption Survey 2026 · NIST AI RMF · OWASP Agentic AI · W3C PROV · C4 architecture · ACM CS2023
Taxonomies overlap: a technique may serve multiple purposes; classification is not proof of technical or scientific completeness. New sources require version, access date, evidence strength and claim-specific validation.