Using AI for work

I have started using copilot AI more and more for work. It seems that
if you have a general idea of the work, it can generate the code
pretty much at about 70%. It's not perfect - for example it almost
always messes up lambda functions but that 30% of problematic work one
can fix.
Meanwhile using other AIs also provide different different outcomes.
So turns out the training data and usage matters a lot. Anyway I think
human intuition and other hard won skills still matter and when AI
screws anything then knowing the nitty-gritty of things really comes
handy. But yeah it's going to mess with people's learning habits and
probably going to make more scatter-brained people. Learning is a hard
process and the inception of knowledge is a difficult thing. Looking
around we already find so many idiots with dirt for brains, I think
for disciplines like engineering where really complex knowledge needs
to be ingested, AI is going to probably tip over people into
un-learnables. But that's problem for the future.
Anyway another topic was if I should invest in a PC that could run
local AI models. But it' s costing more than 1Lac for decent usable
setup. And that's a bit too much money right now. I guess if I had
15-20L then spending couple of lacs on such a PC would not have been
all that difficult. But anyway right now my work is getting done via
the free models. So there's that. Regarding hardware I found out that
NVidia GPUs have the best AI support right now and they do provide
230+ TOPS performance right now. Comparing that to my inbuilt Vega 7
graphics's piddly 1.6 TOPS is an astronomical increase in performance.
But I have another arena to explore and that is small models. I see in
AI generated results that they could also perform certain tasks well
enough. So I'm going to explore that. Lets see if we find something
meaningful there. Cheers!!