Not only is agentic AI bullshit, but it's a specific kind of bullshit.

Today's featured article can be viewed at the link below; it includes a post on Mastodon published by Cory Doctorow, the author of the article containing the entire thread on the aforementioned platform. We strongly recommend that you read the entire text.

In recent years, artificial intelligence seems to be the only topic on everyone's lips, and unfortunately, we have truly reduced ourselves to this point. The latest big news in the industry concerns “agentic” AI, a class of AI specialized in the production of autonomous systems capable of making decisions and performing tasks online without the need for human intervention. 

The question arises: do we really need such a tool? A technology consisting of individual agents capable of performing a specific task, who pass the work on to each other once their small step is complete, like a relay race? The reality is this: we are not able to develop AI capable of navigating the vast world of the internet, only to build an assembly line in which each agent performs an action and passes the baton.

Such a system is fertile ground for the uncontrolled emergence of hallucinations; chain management risks very quickly turning into a poorly organized game of telephone.