Not long ago I pitched for a writing gig that I didn’t get. The company was very excited about where AI was headed, and wanted to know about my experience with the new technology.

To be honest I find AI a bit like 3D TV; not there yet. Might never be. The ambiguous marketing of these expensive predictive text machines however, is nothing short of genius.

I find AI great for giving some jobs structure, because that’s not my long suite and it doesn’t take much for me to go off piste. Plus, it’s a half decent search tool now that Google is just a shit shopping channel.

Of course asking for facts is still problematic; and as it gets better, the mistakes - cutely framed by the marketers as ‘hallucinations’ - are getting harder to spot.

One day we will simply stop checking. A case in point: if you read the LinkedIn intro to this, I observed that the last thing you needed was to hear my ‘two cents worth’.

Out of curiosity I wondered what two cents would be worth today allowing for inflation. The princely sum of 64 cents according to ChatGPT who traced the idiom, quite confidently, back to the late 1800’s.

Given that math is another short suite, I thought about taking her word for it and leaving it at that. Wikipedia, however, had other thoughts starting with a reference to ‘two coins’ from the Gospel of Mark and Luke. My two cents then, would be worth significantly more.

Stanford University research from January 2024 found large language models hallucinated 75% of the time when answering questions about court rulings. To be fair, that was 12 months ago and much water has passed under the bridge since. But in both directions - as AI slop flows quick, the dead internet theory is a conspiracy coming true.

Still, in the grand scheme of things it’s early doors, and anyway, who wants to be that guy? It was time to conduct a little AI experiment of my own.

Entering the job description and my resume into the machine, I asked Chat GPT to rate my chances, and give me some strengths to play on, and weaknesses - which was reframed as ‘growth opportunities’ - to address.

Here’s what she came back with:

Yes, well, good for the ego, but would that help?

Chat GPT seemed far too nice. Too eager to please. I needed a bit of provocation. Anthropic’s Claude was not to disappoint. A slightly surlier version of chat, Claude was slower with the compliments, but still reluctant to give any bad news.

I’m always polite when using AI, because, well you know, just in case. Maybe manners was the problem?

Since there’s quite a bit of talk around the traps about recruiters and HR people using AI to weed out candidates, it was time to turn the tables. Using a different account I switched roles, this time playing recruiter, and asking for comments on my application - ‘would this person be a good fit, etc?’

I’m not sure my clumsy sleight of hand fooled anyone, but again, ChatGPT, the people pleaser, was telling me what she thought I wanted to hear. Claude was quicker to point out where I - sorry, the candidate - could do better.

But like I said, didn’t get the gig.

See? Bloody AI. Taking away all the jobs.