Rose

During the Plague, I had plenty of experience with both meal delivery apps (finding weird new ways how befuddled immigrants get scammed) and app scoots (it’s like public transport, but more unreliable and more expensive).

Don’t do app shit under normal circumstances, being the main lesson here. It’s going to be full of weird fucking shit.

Not long ago, I needed to quickly add a 4-line thing to my Ruby on Rails app on my Raspberry Pi. I was like, shit, I really can’t be arsed to mount the RPi over Samba to edit this stuff on VSCode on my Windows laptop to add like 4 lines. I don’t have vim, I can’t be arsed to check which vimlike I have at hand, I can’t be arsed to see if my Emacs setup has been customised for this RPi. So fuck it, Nano it is!

Let me put it in tech bro speak: most of my AI usage has been suitability evaluation.

In layman terms: does this shit even do what it’s supposed to be doing.

It doesn’t. There are few limited tasks where LLMs can be genuinely used as a tool (my favourite example is image captioning) but it still needs hand holding and can’t work fully automatically. It’ll fuck up, and I’ll need my expertise to unfuck it.

They marketed AIs as “don’t worry, you can just use it as a tool”. It’s a tool that is sometimes useful. This is not a high endorsement.

Now: it’s my humble opinion one of the greatest tragedies of this AI boom is that it never should have been pushed to the mainstream. If LLMs had been left to the stage where they blatantly are now, they’d been left as marginal nerd toys. And all would have been well. But no, the Bros had to ruin it all and let us all discover myriad societal ills unfathomed by scifi authors. Thanks guys.