Chatbots — Rule-Based to Neural to LLM Agents
ELIZA replied with pattern matches. DialogFlow mapped intents. GPT answered from weights. Claude runs tools and verifies. Each era solved the previous one's worst failure. A user says "I want to change my flight." The system has to figure out what they want, what information is missing, how to get it, and how to complete the action. Then the user says "wait, what if I cancel instead?" and the system has to remember the context, switch tasks, and preserve state. Conversation is hard for an ML system. The input is open-ended. The output has to be coherent over many turns. The system may need to act on the world (change a flight, charge a card). Every wrong step is visible to the user. Chatbot architectures have cycled through four paradigms, each introduced because the previous one failed too visibly. This lesson walks them in order. The 2026 production landscape is a hybrid of the last two. Chatbot evolution: rule-based → retrieval → neural → agent The first paradigm did not last five years. It lasted fifty. Knowing its arc matters because every system in it is the same machine — match input, emit a canned response, update a little state — and fifty years of adding rules to that machine never produced the general case. That ceiling is why paradigms…
Chatbots — Rule-Based to Neural to LLM Agents: ELIZA replied with pattern matches. DialogFlow mapped intents. GPT answered from weights. Claude runs tools and…
This free lesson is part of the AI Engineering from Scratch curriculum. Read the full explanation, run the lesson code, and verify the result in the interactive reader or from the repository source.
Browse the complete course catalog or open this lesson on GitHub.