What’s the point of spending billions on AI if there’s no evidence it’s working? 2026 is the nearest the world has flirted with World War 3 in decades. Conflicts where AI was touted as precision magic, like Gaza, are just old-fashioned medieval scorched earth & mass slaughter of civilians. That appears to be the only approach left in Iran. Where the regime is intact, most of its missiles are intact, and it has the strategic upper hand militarily controlling the straits of Hormuz. Last year, the US couldn’t even defeat the Houthis.
“All talk, and no pants” is an expression to describe someone who can talk big, but is never able to follow through. It feels like the same is true for military AI.
No accountability. Nobody is going to jail for blowing up that school full of little girls
A human still took that output and acted on it. That’s very prosecutable.
J-j-just another few hundred trillion bro, it’s going to be so great, you just gotta trust me bro, I swear!
AI is being used in Ukraine to operate drone swarms by one operator. But it’s not a big LLM doing that.
There isn’t any evidence. I tried ai and alway costs me more time. The only (slightly) usefull application was parsing logfiles into it. Sometimes it points at an error (error, not solutions!) a bit faster than me using grep.
best use of AI for me that I’ve ever had was making commit messages and markdowns. that’s it. two things I detest doing or I’m not very good at. And like you said it’s good for pointing out errors. I use NixOS and sometimes I’m too lazy or tired to figure out what I messed up on rebuild so i’ll just quickly copy and paste it into an LLM and it’ll tell me that I missed a “;” or something.
I even tried using it for tech support the other day and chatgpt was insistent to the point of almost threatening me that the kernel I was using (cachyos) was the cause of all my problems (it wasn’t). Like it was imperative that I switch the kernel immediately to the point I was thinking “does openai have some weird agenda against CachyOS?”
It also wastes time, especially chatgpt, with it’s god damn annoying summaries that I’m well aware of or it presenting endless options for something so mundane. Claude is guilty of the endless options too. So for things that don’t require a solution it’s fine, but once you ask it to provide something that does require a solution then you’re in for a world of hurt.
summaries
This is your code … This is what’s wrong … This is corrected code … This is why it works …
(Spoiler alert: beyond “hello world” hardly anything ever works)
