So Close and Yet So Far

Sometimes you make a recommendation or pay someone or a business a high compliment, and then it kind of bites you. For example, you recommend a restaurant to a friend, but when they go there they wait two hours to be served and everything is wrong and cold. Well, how were you supposed to know?

Last week, I waxed poetic one might say about how ChatGPT was helping me with my job search. Chat and I were becoming pals, Everything was great. I didn’t recommend using it for your job search per se, but I came quite close.

Now, I am going to have to backtrack a little bit. Don’t tell the robots.

I had ChatGPT doing a couple of things for me. First, it was supposed to monitor several different websites each day of the week. Second, it was supposed to give me five important people from each company I was searching so I could stalk I meant connect with them on LinkedIn. The goal was approximately five new connections per day. Naturally the more jobs it could find for me to apply to the better.

Like all things you’re in love with, the facade started breaking down slowly. First, out of the first five people it recommended, three were no longer at that company. Needless to say, that is not a great track record. Of course, people are also not so great anymore about updating their LinkedIn profiles when they change jobs, so who knows which is more correct.

The other problem I started to spot was with the job site search. It seemed kind of ridiculous to me that out of dozens of websites, there was not a single interesting post all week. Interesting is defined here as something I would want to apply to. I decided to do some searching myself, and I found a job I wanted to apply to. ChatGPT hadn’t picked it up yet. OK, that’s fine. But then it mistook the link I posted for an older job that had been posted a month or two ago. That made me nervous, so I tried another search myself. Once again I found a job that AI did not find. Is it really visiting all of the websites it’s supposed to visit?

I have the skepticals.

I have always advised people to comb carefully through AI before sending it along or publishing it. You’ve seen the embarrassing things that happen when someone does not do this step. You’ve seen those AI prompts stuck into a newspaper article right? Ugh. Well, it turns out you need to proof everything AI does, not just writing. This is not revolutionary, but my experience was a good reminder.

A lot of people call AI their assistant, and I’d say that’s apropos. I’d qualify that, however, and say it’s an assistant who happens to be your six-year-old on shadow day. It may very proudly tell you it did absolutely everything. It’ll even be nice about it. The reality, however, is that you should go back and spot check yourself, whatever you are doing. Let me be the lesson for you.