This is perhaps my favorite story of A.I.
I was working with a colleague that had built a subsite with a lot of custom code and formatting. One of her pages, critical for an upcoming project, wouldn’t load at all. It was just blank when it tried to display.
She called me up and we did a screen-share to look at the code for the page. After some back and forth and scrolling up and down, I thought I had it figured out.
“I think,” I said, “that this is missing a DIV tag,”
On a webpage, a set of DIV tags will define a region for something to happen on a page. They come in pairs and if there’s a mismatch, that could be part of the problem.
Since this is an easy thing to try, I expected that she would drop in an extra closing DIV tag at the end and try it out.
Instead, she suggested, “Let’s put this in ChatGPT”
It seemed an odd approach to add another step to try – and in a way I felt like I was being tested. If ChatGPT agreed, would we try adding the tag?
She copied the code into ChatGPT and asked for an analysis. The A.I. suggested that there were instead too many DIV tags.
She removed one of the closing tags and tried this again – and the page still failed. And she was now at a loss.
“Let’s add in two tags,” I suggested, “One that ChatGPT suggested you remove and the one I thought was missing,”
Oddly dubious, she edited the page to add in the missing tags – and the page loaded perfectly.
I got a happy “thank you” and we ended the call.
As one might imagine, I rode that victory for days afterwards.
The troubling bit was not that A.I. got this wrong, but that it was the only tool she was using – rather than it being one of the tools and resources.
Like, perhaps, a trusted colleague.
I could have just as easily gotten this wrong – we were scrolling pretty fast – but I’m glad it worked out.
Humans: 1, A.I.: 0