Introduction
There's no stopping the hype. Everyone is talking about AI agents these days. But it feels like it does not deserve the hype. A special prompt and you call it an AI "agent"? Is chatGPT an AI agent? Let us try to dive too deep into the buzz.
How can we define an AI agent
Oxford Languages defines an agent as:
-
An agent is a person who acts on behalf of another person or group.
-
An agent is a person or thing that takes an active role or produces a specified effect.
Replace a person with an AI agent:
-
An AI agent acts on behalf of another person or group.
-
An AI agent takes an active role or produces a specified effect.
So, to categorise a system as an AI agent, it has to fulfil certain criteria:
- It must be artificially intelligent
- Perform a task on behalf of another entity
- Performs an action based on its task
Discuss a few examples
A bouncer who throws out drunk people.
→ It is an agent, but not artificially intelligent.
A robot bouncer who throws out rude people.
→ Tricky one. Of the categories defined, it is an AI agent.
A program that checks emails and flags them as spam.
→ It is an agent and definitely smarter than me doing it.
A sorting algorithm that sorts green and red apples in a production line
→ It is an agent, it is as smart as I am.
A robot vacuum cleaner
→ It is definitely an agent.
So to sum it up, any process or action written in software that is executed and performed with human-level intelligence counts as an AI agent. But the expectations are different. When we talk about the buzzword AI, we have to ask ourselves what intelligence means in artificial intelligence. And my personal answer is that it is a spectrum. It is not right or wrong. In some areas, current models outperform humans and in other areas they do not.
Discussion
With the advent of LLMs, the buzz around AI agents came up, and that is where we see the agents popping up now. But we have had them for a long time, if you think about it. But the good thing about LLMs is that we can define them in natural language and so we can create an AI agent really quickly compared to before LLMs.
That is the magic of the hype around AI agents. We are just really fast at creating them. If you want a spam filter agent, no problem. A personalised marketing mail creator. Anything you can think of. What used to take days or months can now take minutes. And if you want a multi-stage AI system with different tasks, you can just write them, define them and make them work together and talk to each other with minimal effort. Another point is that LLMs have a high quality baseline. Whatever you produce in that short period of time, it will give you a very reasonable result.
Conclusion
To sum it up, AI agents are valid the hype because now everyone is able to create and use intelligent products within a short period of time, making the creation of AI product an easy task and bring it into the hand of millions of people to solve their daily problems.
Enjoy! ❤️