Foundation Models and General-Purpose AI

Definitions of foundation models and general-purpose AI systems, their distinguishing traits such as broad generalization and adaptation to many tasks, their taxonomies, their relation to large language models, and the notion of emergent capabilities. Excludes the chronological account of their emergence, which belongs to 1.8, and the regulatory obligations placed on them, which belong to 14.3.

• 1.8 Foundation and Generative Models
• 14.3 The EU AI Act: Risks, Obligations and Dates

• 5.2 The Transformer and Beyond
• 5.3 Mamba, State-Space Models, MoE and Emerging Architectures
• 5.4 Pre-training and Scaling Laws
• 5.5 Fine-tuning, RLHF, DPO and ORPO
• 5.6 Context Protocols and Tool Use
• 5.7 Multi-Agent Orchestration
• 5.8 Open, Open-Weight and Closed Models
• 5.9 Natively Multimodal Models
• 5.10 Hallucination and Structural Reliability
• 5.11 Sovereign and Localized Models