The not-so-recent case of Intelligent Agents

Generated by DALL-E

It seems that history repeats it self. Or at least, people in software engineering are always doomed to repeat themselves.

The latest one is the Agentic AI. Sounds glamorous, huh? It is the newest thing in town, everything is going to be replaced with agents and people will have to forfeit their jobs, since everyone is going to be redundant. Sorry to say, but I don’t think so, or let me say it straight. There is lots of work that needs to be done to create systems that will make human redundant.

A small personal story

I studied computer science in 1995 to 2000 and at the time we touched various topics. One of them was Intelligent Agents. What? In the 90s? Yes, it seems that along with Nirvana and Smashing pumpkins, in the 90s computer scientists created systems like that. Just for this post, I did a very quick and un-thorough research and this is one of the first links I stumbled upon. It is an article published in IEEE Potentials. And I was fascinated by the abstract:

Today, everything moves at such a hectic pace. Everyone has more to think about and more things to do. In a perfect world, everyone would have a personal assistant to handle the mundane, purely routine tasks. Intelligent agents (IAs) could do these mundane jobs automatically. Of course, such flexible and ubiquitous agents are currently not possible for certain cost and feasibility reasons. The paper presents a definition of an intelligent agent and discusses the characteristics of the agent’s environment.

Wow, sounds familiar? Then another thing came into my mind, that back in the day, as an undergraduate student, I also worked on a nice assignment, in a course named “Intelligent Agents” and I created a system of an autonomous agent that visited some academic libraries and performed various queries and then at the end reported all its findings to the user.

Nice right? How do we call it right now? A similar system, ah yes, deep research. You can find it in your favorite edge-AI service. Of course, more generic but less autonomous, but we will cover this on another post.

The case of Agentic AI

So, that basically mean that Agentic AI, along with AI in general has been in the market for quite some time. Also, those systems at the time, was not to sophisticated like they are now. But, let’s be realistic, these systems in order to be productized and build, need lots of effort. This is why they are not everywhere (yet), because it seems that lots of people a putting effort to make them happen.

So, nothing is really new and so far AI is not a Panacea for everything. Actually, it is a really useful tool that all must be trained to use it. Agentic AI right now, is not really agentic, it is just like an AI in a workflow system.

I am pretty sure that lots of frameworks and ideas will come and go in the following months or years, but only a handful will stick. Remember those GPTs that OpenAI had as fine-tuned models with a focus and custom data? Where are they now?