Autonomous Agents is the concept of artificial intelligence (AI) systems that can perform a series of complex tasks entirely undirected to achieve a goal. For this an autonomous agent needs to reason, plan and interact with the necessary systems & real world to solve the problem.
Sparked by the rapid advance of Large Language Models (LLMs), there is an increased amount of exploration towards a world of autonomous agents by building more purpose-built "agents" that can perform a set of tasks. To build these agents, researchers and companies are using different techniques to leverage LLM capabilities and work around limitations. These techniques include different prompting strategies such as "Chain of Thought" prompts where an LLM is asked to think through a problem step by step and outline what actions it would like to perform.
Some of the most known research projects that have been moving the exploration of agents forward are the Generative Agents paper that used GPT-3.5 to create a small town simulation of different agents interacting with each other, as well as the Voyager project an LLM-powered agent that was left alone to play and control Minecraft by being able to generate code and learn new skills.
As AI companies develop increasingly-capable agents, more companies are looking to leverage agents for parts of their customer engagement, whether for customer support, initial sales engagements or to provide customers with a type of "copilot" that can help the customer navigate and set up their product.