ARTIFICIAL INTELLIGENCE The Complete Guide (SECOND EDITION): From neural networks to agents: understanding, choosing and governing AI

What changes when artificial intelligence stops simply answering questions and begins to act?
Artificial Intelligence: The Complete Guide explores the foundations, applications and decisions behind that transition. In this updated second edition, Achille De Tommaso connects the principles of machine learning with the realities of generative AI, large language models and AI agents.
The book begins with data, logic, probability, knowledge representation and linear algebra, explaining the role these concepts play in artificial intelligence. It then develops machine learning, neural networks, deep learning, computer vision and natural language processing. Mathematical ideas are introduced to clarify how systems work, without requiring the reader to be a specialist.
From there, the focus moves to foundation models and the choices behind their behaviour. Discover how attention, context windows, mixture of experts, post-training, quantisation and distillation affect capability and efficiency. Explore retrieval-augmented generation, multimodal systems and models that allocate more computation before answering. Learn why fluent language is not proof of accuracy, and why a convincing demonstration is only the beginning of evaluation.
AI agents receive sustained attention. The book examines memory, state, tool calling, orchestration and interoperability, including MCP and agent-to-agent communication. It explains how approvals, sandboxes, least privilege and audit trails help govern systems that can modify information or trigger actions. Practical discussions of prompt injection, poisoned tools and compromised memory show why agent security must be designed into the workflow.
Private AI and data sovereignty are treated as architectural and organisational choices. Readers follow the path from document archives to searchable services, considering permissions, retrieval, evidence, deletion, telemetry, backups and migration. European and Italian models and computing initiatives provide a further perspective on capability, ownership and operational independence.
The economics of adoption receives the same scrutiny as the technology. Compare API-based services, internal infrastructure and dedicated deployments. Understand total cost of ownership, review time, infrastructure expenditure, return on investment and the cost of error. Worked teaching examples reveal how assumptions about volume, quality and human checking can change a project's financial case.
To move beyond experimentation, the book explains how to design test sets, choose metrics, combine automatic evaluation with human judgement and detect regressions after updates. Red teaming, acceptance tests, purchasing criteria and staged adoption help readers translate an attractive idea into a service with measurable performance and defined responsibilities.
Applications in healthcare, financial services, manufacturing, mobility, defence and scientific discovery connect the foundations with real fields of use. Chapters on employment, privacy, regulation, ethics, sustainability and consciousness examine the consequences of AI without confusing established capabilities with speculation.
Written for professionals, managers, students and intellectually curious readers, this book supports both systematic study and targeted consultation. Its two-part structure lets readers build their understanding from the beginning or turn directly to questions about models, agents, deployment and governance.
Whether you are assessing an AI investment, planning an organisational project or seeking a clearer understanding of the systems reshaping everyday life, this edition offers concepts, examples and decision criteria for thinking clearly. Understand what AI can do, test what it actually delivers and retain responsibility for how it is used. Build your own informed view.