Original subtitles

- [Instructor] Agent AI automates AI

using the reasoning power of large language models

to run complex tasks by breaking them into simple tasks.

Agent AI has some special characteristics

that makes it different from intelligent chatbots.

First, agent AI uses the reasoning power of LLM

to make decisions.

These are decisions the AI will make on our behalf

in our absence,

and so reasoning is very important for our safety.

We will talk more about this throughout the course.

This will be helpful for us to design

and build agent AI products and systems in this course.

Second, agent AI has memory.

Why does a chatbot need memory?

Well, intelligent chatbots,

like Gemini or ChatGPT,

can remember the context of a chat

and continue conversations.

But it is not an agent AI,

and it does not have memory.

So if we ask for something

and the chatbot gives an incorrect answer,

it'll forget its mistake.

An agent AI has memory.

So if it gives an unsatisfactory answer,

it can remember and improve it.

How does it know its answer is unsatisfactory?

The agent AI has the ability to ponder its own decisions

and adjust its approach.

Third, the best part of agent AI is that it can communicate.

Who do you think it communicates with?

Well, it can communicate with its environment.

Think of automation in a factory.

And it can communicate with other agents.

Think of multiple agents doing a task each

to execute on a complex task that you want to automate.

So an agent AI automates AI

by breaking complex tasks into simple tasks,

and uses reasoning to run tools,

uses memory to remember and improve itself,

and communicates with the environment

and other agents to achieve its goals.

Join me on this journey to learn more about how agents work,

the many different kinds of agents we can design.

And I'll teach you how to build agents

with live exercises using absolutely zero code.