What is NOT AI?
Throughout my artificial intelligence projects and consulting work, where I had to assist many people unfamiliar with the subject, I noticed something:
It’s always better to start by explaining what is NOT artificial intelligence.
Because in practice, many people already come with preconceived notions that are actually wrong.
And this certainly ends up hindering more than helping.
Especially the confusion between artificial intelligence and automation.
So let me explain the difference between artificial intelligence and automation in a very simple way.
I’m not going to tell you that automation is what you do in Python, or that artificial intelligence is what you do with neural networks.
I’m going to explain it in a practical way, focusing on what really matters: the result.
There’s a concept here that has an ugly name, which is “deterministic”.
But the idea is simple.
Always think about the outcome of the activity.
You can have a person doing a task.
You can have automation doing that task.
Or you can have an AI doing that task.
The question is: which of these options delivers the best result?
Now comes a simple rule that will help you a lot.
Look at the type of result you expect.
If the result needs to be exact, always the same, with no variation, automation is better.
For example, an addition.
If you want to add two numbers, you can ask a person to do it.
You can ask an AI.
Or you can use automation.
Automation is what will always get it right.
Because automation was made to execute exactly that, with no variation.
Now think about it differently.
If you take that same problem and give it to 100 people, they will all give you exactly the same result.
That’s a clear sign that the problem is deterministic.
And with this type of problem, automation wins.
Now let’s move on to another type of problem.
Imagine you ask 100 people to imagine and draw a dog.
Each one will draw it a different way.
And yet all the drawings can be correct.
Or say you ask 100 people for poems.
You’re going to get 100 different responses.
And many of them will be good.
Here, the result considered correct is not unique.
It is variable.
And this is exactly where artificial intelligence does better than automation.
Because AI works well with this type of problem where there are multiple possible answers.
Where the result doesn’t need to be perfect, but it needs to be good enough.
So, to simplify:
If the problem has one correct answer, you probably want automation.
If the problem accepts several good answers, you probably want artificial intelligence.
Understanding this completely changes how you use technology.
And to make this even clearer, I’ll show you a real case.
→ Next: 1.1.1 The case of the marketing agency that wanted AI