AI Bots – Are They Really Intelligent?

 

1st Edition, Unreviewed

In the previous post, I introduced the reader to what makes a difference in using an intelligent bot (ChatGPT, Copilot, Gemini, etc.). It is necessary to add value to the bot’s answers, otherwise its work will have no value at all, being a simple copy and paste.

In this post, the goal is to reduce the psychological pressure on people that new sources of information and media in general create to capture attention — of course motivated by the economic interest of the audience combined with the economic interests of the parties involved.  

Broadly speaking, I consider that AI bots have average to below‑average intelligence. 
Intelligence is the ability to understand, to create something new, to infer, and it varies according to context, both for humans and for intelligent systems (AI).  

Let’s look at practical facts.  

If you ask a question that elicits an answer about something tacit — that is, already consolidated knowledge — the bot will provide the respective answer. In these circumstances, I have not recorded a single failure.  

Now, if you ask a question or conduct research about something new, which is not yet on the Internet or which it has not yet identified from some alternative source, it fails — because it is not truly intelligent. Truly intelligent beings know how to recognize “not knowing” and are capable of standing out by elaborating on the unknown.  

The intelligence capacity of an AI bot is associated with its ability to compile available information, based on criteria of semantic, logical, and contextual correlation — the latter being reasonably limited.  

When faced with a new question for which the bot finds no answer, it falls into digression, trying to make some correlation.  

It’s more or less like a person who tries to stall the listener when they don’t really have an answer.  

The contextual capacity of intelligent bots is still thin and fragile.  

They “forget” what they said before, what we told them, and commit a series of errors that depend on contextual analysis in the sequence of dialogues.  

The example below is one among many I have recorded.  

EXPLANATORY NOTE:
The concepts I refer to in the figure above were provided to the bot earlier, from the very beginning of the dialogue, in more than one opportunity.  

The interesting part of the algorithms of these intelligent bots is that, when alerted, they are capable of performing a self-verification check, as you can see in the example above. This is already a strong point achieved by current technology.  

When executing non‑trivial tasks, those that are not a “ready‑made and established recipe”, then human supervision is necessary.  

Summarizing for the reader, intelligent bots today are powerful for:

- Finding answers on already established topics.  

- Logical analysis without deep contextualization.  

They still need to evolve in the following areas:

- Capacity for contextual analysis in long dialogues.  

- Capacity for precise contextual correlation — the lack of which makes them fall into digressions, i.e., saying many things that don’t form a strong connection with the main context or the purpose of the question, causing repetition, digression, and even loss of direction.  

- Capacity for inference. 
This last point is the most delicate, and the question remains: 
How far can their intelligence evolve through the technology of our current algorithms?

I have read in some publications that AI technology is at the limit of its resources. 
I don’t believe that. There is still much to be done.  

After this quick analysis, what does it represent for us humans, mere mortals?  

Repetitive human activities, and those that depend heavily on memory and less on creative actions, will gradually be replaced by bots or robots.  

The best way to remain irreplaceable or relevant is to continue adding value to the output of the intelligent bot: reviewing its answers, correcting, improving, guiding the bot in the right direction (because it can get lost), and performing other activities necessary to achieve the desired goal.  

Conclusion:

Intelligent bots are not that intelligent, and they need human supervision. 
The rest is pure pressure from those who seek to take advantage of human irrelevance.  

Too much pressure, too much foam on the beer’s collar.