From LLMs to AI Systems

Описание: In this lecture, we explore the evolution of artificial intelligence from standalone Large Language Models to complete AI systems.

We begin by explaining what a Large Language Model really is from a technical perspective: a probabilistic model that predicts the next token. From there, we examine why powerful models alone are not enough for real-world applications. Issues such as hallucination, lack of memory, lack of access to real-time or private data, and inability to perform external actions make it clear that modern AI requires much more than a single model.

The lecture then traces the progression from traditional model-based AI systems to LLM-based applications, Retrieval-Augmented Generation, and finally agentic AI systems. We also introduce the modern AI system stack, including models, tools, workflows, memory, orchestration, and infrastructure.

By the end of the session, students will have a clear architectural view of how modern AI applications are designed and why the future of AI is not just about better models, but about building better systems.

This lecture is ideal for students, engineers, and anyone interested in AI architecture, agentic systems, RAG, and production-ready intelligent applications.

Topics covered:

What a Large Language Model really is

Limitations of pure LLM systems

Why AI applications need retrieval, memory, and tools

The evolution from chatbots to AI systems

RAG and agentic AI

The modern AI systems stack

Course roadmap for AI systems architecture

Course: AI Systems Architecture
Instructor: Dr. Mohamad Aoude