Introduction
Artificial Intelligence (AI) has become an integral part of our daily lives. Ever wondered how your favorite AI tools work? There's a lot more to it than just technical jargon and complex algorithms. In this article, we shed light on the fascinating world of Agentic AI, taking you on a journey that begins with chatbots and ends with AI agents.
Understanding Large Language Models (LLMs)
A. The Role of Chatbots
LLMs are the building blocks of popular AI chatbots like CHBT, Google Gemini, and Claude. They excel at generating and editing text. Consider this simple visualization: you, the human, provide an input, and the LLM produces an output based on its training data. For instance, if you asked Chachi BT to draft an email requesting a coffee chat, your prompt is the input, and the resulting well-mannered email is the output.
B. Limitations of LLMs
However, LLMs have limitations. Firstly, despite being trained on vast amounts of data, they lack knowledge of proprietary information, such as personal details or internal company data. Secondly, LLMs are passive, merely waiting for our prompt to respond. These traits are crucial to understand as we explore further.
AI Workflows: The Next Level
A. The Role of Predefined Paths
AI workflows are essentially a series of predefined paths set by humans for the LLM to follow. For instance, if you told the LLM to fetch data from your Google Calendar for personal event-related questions, it would provide the correct answer. But if you asked about the weather, it would fail, as it's programmed only to search your Google calendar.
B. The Power of Retrieval Augmented Generation (RAG)
This limitation of AI workflows doesn't mean they're not powerful. Retrieval Augmented Generation (RAG) is a process that aids AI models in looking up information before responding. RAG, in essence, is a type of AI workflow that can significantly enhance the utility of AI tools.
AI Agents: The Future of AI
A. The Shift from Human to AI Decision Making
The fundamental difference between an AI workflow and an AI agent is the decision-maker. In an AI agent, the human decision-maker is replaced by an LLM. This AI agent must reason and act via tools, essentially embodying the REACT (Reason, Act) framework.
B. AI Agents and Iteration
A significant advantage of AI agents is their ability to iterate autonomously. For instance, an AI agent drafting a LinkedIn post can critique its own output using a predefined set of best practices, repeating the process until all criteria are met.
C. Real-World AI Agent Applications
An example of AI agent application is a demo website created by AI expert, Andrew. When you search for a keyword like "skier," the AI vision agent first reasons what a skier looks like, then acts by searching clips in video footage, indexing the clip, and returning it to the user. This entire process is performed autonomously by the AI agent.
Wrap-Up
In summary, the world of Agentic AI can be understood in three levels.
- Level 1 involves simple interaction with LLMs, as seen in chatbots.
- Level 2 introduces AI workflows, where humans set predefined paths for LLMs to follow.
- Level 3 encompasses AI agents, where the LLM becomes the decision-maker, reasoning and acting autonomously to achieve a goal.
As we move towards the future, Agentic AI continues to evolve, bringing us closer to a world where AI tools are not just passive responders but active, intelligent agents. Whether you're an AI enthusiast or a professional looking to leverage AI tools, understanding these concepts can provide a solid foundation for your AI journey.
