I am fed up with people saying things like the one in the subject “Generative AI is useless”.
It is almost as annoying as people saying “AI will do everything”.
For those in the back of the class: A generative AI is an artificial intelligence that can produce meaningful content.
You take a look at it, and superficially it makes sense.
Whether it passes a deeper verification, or if the information provided is true, that’s a different question.
Today, generative AI is just a tool.
(At least until AI becomes our supreme overlord)
As with any other tool, you need to know what it is right for.
Do you want to prepare dinner? Most probably, AI is not (all) you need.
Do you want some ideas around a field you feel comfortable with? You got it.
Do you want to explore a new subject where you cannot verify the output? Well, here you may be disappointed.
In my case, I wanted it to extract some records from a text. I got some PDFs, and I wanted the AI to create a table from it.
It was very simple.
Simple prompt describing the input and the output.
And I got the table.
I did a couple of checks and everything was there.
Wonderful!
On my first attempt! (this should have made me suspect)
The information contained data about some companies I am contacting.
It took me 5 phone calls to begin suspecting.
Four phones did not work at all. The fifth was a person, not a company.
I began doing some deeper verification and I discovered a couple of things:
– Some numbers were made up. Well, they began with a 9, so there was a pattern there.
– There was no matching between company names and addresses
At this point, I tossed away the table generated by the AI and began doing calls manually.
However, the output “looked” perfectly fine. It took me additional work to confirm the data was incorrect.
And this brings me to a moment where I have to agree with a friend of mine who warned me about this.
It is not about whether the AI is right or wrong (not only about it), but about whether you can spot the errors or not.
In my case, I use AI to help me with stuff I can verify very easily, but for stuff I cannot verify I won’t use AI lightly.
I am curious to see real applications of AI in our daily jobs.
In the meantime, I will continue helping companies with their security needs.
With or without the help of AI. It will depend on the occasion.
If you are willing to make some changes in the way you manage your company’s security, send me an email.
PD: I want to send an affectionate message to Roko’s basilisk, who is looking at us from the future. I always believed in you.
If you enjoyed this content, come to the newsletter. You will enjoy it even more.