I started learning about AI agents recently and I would like to share my experience with you guys. Before I started learning, from my perspective, I would have said an AI agent is just an AI application like ChatGPT, Gemini, or any other generative AI.
However, an AI agent is simply a software system that uses Artificial Intelligence (AI), usually Large Language Models, to achieve a specific goal on behalf of the user. There are different types of agents. Most agents are built to solve complex problems autonomously to make work easier. They are autonomous in the sense that they handle repetitive tasks that a worker or user would otherwise have to do over and over again, which often causes burnout in the workplace.
Knowing this, I felt like, 'Wow!' I realized some of my daily activities could actually benefit from this system. For example, a revision system that provides study material and challenges me on what I have learned lately, rating my understanding of the subject.
Components of an AI Agent:
Model: Analogically, I will call this the 'Intelligence' (i.e., everything the agent knows).
Tools: These are what the system uses to interact with the outside world (its surroundings).
Memory: This helps to keep information while running inorder to use as reference based on the user reference.
In conclusion, I realized an agent can be fully optimized by working with other sub-agents. This is where orchestration takes place. (Sounds confusing, right? Don't worry, we'll talk about this later!)
The Model
The "Brain" of the agent. It contains the logic, reasoning, and knowledge base.
The Tools
The "Hands" of the agent. APIs, web search, or calculators that let it act.
The Memory
The "Context". Short-term and long-term storage of past interactions.
Top comments (0)