LLM MCP and Agents Oh My.
LLMs, ML, NLP, MCPs, RAG, Agents, and MCP Servers . . . the list goes on and on and how the heck is anyone supposed to keep up? Every day a new model, tool or term is introduced it's daunting trying to understand how much you need to learn about them.
Don't worry it's not nearly as scary as it sounds! Once you grasp the fundamental concepts, you will start to see the patterns and how most of the new terms fall into general categories. Then you can pick and choose what works best in your workflow.
So take a deep breath, and lets put this in very basic terms to help you grasp the foundational picture.
Keep in mind that each of these terms and topics can go very deep, but for most of us, you really only need a general understanding of what these things do in order to become more proficient in AI.
LLM - Large Language Model: This is where the magic happens - the brains of the operation. They can be generalized knowledge (Chat GPT) or very specifically trained (BloombergGPT)
What are LLMs? These are "models" or "AI Brains" (Claude, Bart, PaLM , GPT-4 . . .)
In general the average person doesn't need to know much about LLMs. Most of the time you will interact with these via a drop down menu and you try out different models based on what they excel at - images, text, code generation etc
They are built using:
ML - Machine Learning: This is the overall field that allows computers to "learn" from data (based on very complex algorithms, trained on a vast array of data.)NLP - Natural Language Processing: This is foundation for how you "chat" with an LLM (Prompt Engineering or using regular language)
Who needs to know about these in depth? Data Scientists & ML Experts
Model data is limited to the timeframe that it is trained. So if you need the most recent, up-to-date information, you'll need to extend its knowledge by allowing it access to extra tools and data. This is where new tooling and protocols come into play.
AI Agent: These give LLMs (or Models) the ability to "do things" beyond just chatting (like gathering up-to-date documentation, running code, or interacting with other software). AI agents are created by allowing LLMs to access external tools in order to make them "smarter." Many products allow you to do this fairly easily, and just watching a 5-minute tutorial is often enough to get one set up.
RAG (Retrieval Augmented Generation): This is a framework or technique (aka Design Pattern) to allow LLMs to access and use fresh, external data (for example, your own documents or the latest news) to give you more accurate and current answers. These commonly and primarily use vector databases for storing the external "library" of information. So this isn't a service or a tool itself; it's a way of implementing technology to allow models access to more information.
Vector Database: A database designed for (and excelling at) turning unstructured data into something intelligent and searchable for AI applications.
MCP Servers (Model Context Protocol): These are like specialized digital connectors that allow an AI (like Copilot) to use external tools or access specific data sources. For example, when you "add an MCP" to VS Code, you're essentially connecting it to one of these servers, giving your AI (Copilot) new abilities within your workspace.
I hope this basic knowledge will not only help to remove some of the obscurity and overwhelming feeling, but also show you that you are more than capable of learning AI and taking the baby steps needed to level yourself up.
