What are Agents?

Oct 20, 2024·
Mathusuthan N Kannan
Mathusuthan N Kannan
· 2 min read
Software Agents - generated by Gemini Pro

Agents are the next big thing today. Even though Agents have been around for a long time, the recent advancements in AI have made it possible to create Agents that can perform a wide range of tasks. In this article, we will discuss what Agents are and how LLMs are being used to enhance Agents.

What are Agents?

An Agent (so called because they have agency i.e., providing service) is a software program that performs tasks on behalf of a user. Agents can be used to automate tasks, perform complex calculations, and interact with other software programs.

Agents can interact with their environment, process information, and make decisions based on that information. LLMs can be used to enhance Agents by providing them with the ability to reason, plan, observe and learn from their environment and take actions based on that information.

Characteristics of LLM Agents

Interaction with External Environments

graph TD A[LLM Agent] -->|Access| B(Web) A -->|Interact| C(APIs) A -->|Control| D(Robots) A -->|Process| E(Multimodal Data)

Tool and Knowledge Utilisation

--- config: theme: neo-dark --- graph TD F[LLM Agent] -->|Retrieve| G(Information) F -->|Utilize| H(Tools) F -->|Integrate| I(Knowledge Bases)

Dynamic Workflow

graph TD J[LLM Agent] -->|Decompose| K(Subtasks) J -->|Delegate| L(Specialized Modules) J -->|Collaborate| M(Other Agents)

Memory and Learning

graph TD N[LLM Agent] -->|Update| O(Memory) N -->|Learn| P(New Tools) N -->|Gain| Q(Understanding)
Agents LLM Generative AI
Mathusuthan N Kannan
Authors
Mathusuthan N Kannan
AI / ML Architect