What are LLM Agents?

We learn the history of AI agents, define agent and LLM agents, and compare popular LLM agent frameworks.

For Access: https://cbarkinozer.medium.com/what-are-llm-agents-724141b134c4?sk=63eeae77d6b4c69d27c1486900ff223f

Abstract

The word “Agent” comes from the Latin word “agere,” which means “to do” or “to act.” In general, an agent is something that acts or performs tasks on behalf of another entity. Most likely, the term “agent” transitioned from philosophy to the context of artificial intelligence at MIT in the late 1960s. Humans, software agents, and machine learning algorithms are not artificial intelligence agents. Machine learning algorithms apply the formulas and algorithms they are coded with and cannot respond dynamically to situations. Artificial or deep neural networks, architectures such as LSTM and Transformers, are also merely algorithms that take input and give output; they are not directly considered artificial intelligence agents but can be part of an AI agent. In this article, we define an AI agent as follows: “AI agents are autonomous computational entities that can interact with their environment and learn from these interactions to achieve specific goals.” Therefore, it is suggested to use the term “LLM Agents” for AI agents that use LLM as the organ (Latin organum meaning tool) to enable decision-making capabilities. A comparison…