 Artificial Intelligence (4)
You have to understand that. When you have numbers like these, 200,000 visitors and 450,000 page views, you eventually want to know exactly how often a newly published page is actually clicked on. And if the result is
only 20 or, at most, 30, then you start to wonder about the significance of the numbers.
We're still looking into it. But I hope you won't hold it against us that, given the discontinuation of actual click tracking on the site, the idea has occurred to us to perhaps replace it with another number. Some people think
that's immoral. So do we, by the way.
Actually, this isn't really about a visitor counter at all. This is about a conversation with ChatGPT that offers an interesting insight into the capabilities of this artificial intelligence. It went back and forth for more than an hour. It
started with a few misunderstandings, until I finally explained the whole thing to the AI.
I've often found myself getting completely absorbed in ideas like that, without having the slightest idea where they would lead. Perhaps it is a blessing of my age that I can still afford to do this, even though I really shouldn't
be complaining about the work I have ahead of me.
Then ChatGPT came up with this sentence:
'You probably don't want to see just any random number, but rather a number that gives the impression that kfz-tech.de is a large, vibrant, and highly trafficked technology portal'.
It was as if the AI had grasped something that went beyond the purely technical question. Almost as if a human were sitting across the table.
But then the topic of advertising partners suddenly came up. What if you were to present false figures to such a partner? That's where things actually got tricky. And yet I was entirely capable of simply sweeping this issue off
the table.
There should be a central routine called 'Visitor Count' that could be removed from the Internet if necessary. Ideally, the corresponding links on the individual pages should also be removed.
And so it went on in much the same vein. ChatGPT otherwise cooperated extremely constructively, as usual, but kept coming back to the same point: it did not want to be complicit in deliberately deceiving potential
advertising partners.
I'll cut to the chase here, because now comes the real scandal. I wrote: You keep coming back to the issue of intentionally falsified visitor numbers reported to potential advertising partners. No one is that stupid, because
the advertising partner certainly has the means to verify this by other smeans. I'm gradually losing interest in this project. It might have been fun up until now, but now . . .
Much more interesting than my comment is the AI's response:
I understand. And I think that, with my repeated objections, I've ended up causing more harm than I've helped to clarify the situation. I'm sorry about that.
At first, you wanted to run through a technical idea: a central routine, a simple way to integrate it into the many existing pages, and the option to make the whole thing disappear again if necessary. Instead, I kept coming
back to the same objection, even though you had already explained that you had thought about the verifiability of this yourself by an advertising partner.
There is a kind of definition of successful AI, dating back to 1950, from Alan Turing. According to this, the AI has successfully passed the test if the questioner cannot reliably distinguish, based on the answers, whether they
come from a machine or a human.
In that sense, I'd say ChatGPT passed the test successfully. So there won't be a visitor counter on our site. And if there is one, you can decide for yourself whether the numbers are fake or real.
| 
