What are AI Agents?
In simple terms, a Computer Agent (AI Agent) is something that can act like a kind of program that can perform some autonomous actions like searching, analyzing and taking decisions, and provide you the end results.
Don’t think too complex. Let’s think about the below scenario:
Real-World Example: Booking a Cab
You wanted to book a cab. There are multiple platforms to book a cab (Ola, Uber, Rapido). How do you find the lowest cost and book the cab?
Currently, we need to open all the apps and provide the source and destination and check the price:
- Step 1: Open Ola, provide the source and destination and check the price. (Remember the price)
- Step 2: Open Uber, provide the source and destination and check the price. (Remember the price)
- Step 3: Open Rapido, provide the source and destination and check the price. (Remember the price)
Three hassles with the above method:
- Time consuming
- Keep the price in mind
- There are few repetitive operations
Now, an Agent can perform the above operation on your behalf. Just instruct the Agent to do the cab price analysis and book the cab. It can analyze the price and book the cab without human intervention.
You may think that this can be done even with a piece of code. Absolutely correct! But the problem here is: Let’s assume you wrote a piece of code that can analyze and book the cab. After a few weeks you wanted to book a cab for later. Now you have to modify your code to introduce the new feature. With an AI Agent, it can do this for you without making any code changes.
Hope you got some idea about what an Agent is.
How do Agents Work?
There are a few terminologies you need to understand about how AI Agents work:
- LLM - Large Language Model. Basics about LLM.
- Data set (Training data, Testing Data, Data sampling)
- Vector DB - Basics about Vector databases
- Training LLM
- Agent Knowledge base
- Agent Instruction
- Agent Tools
LLM (Large Language Model)
LLM (Large Language Model) is a component which is trained using a set of algorithms and a huge amount of data. Let’s assume our brain. A human brain is trained based on experience (data) and stored. The brain can think and collect the data from memory.
Similarly, LLMs are trained on vast amounts of text data and can understand, generate, and reason with language to help agents make intelligent decisions.