this post was submitted on 04 Apr 2026
204 points (99.5% liked)
Interesting Shares
2637 readers
93 users here now
Fascinating articles, captivating images, satisfying videos, interesting projects, stunning research and more.
Share something you find incredibly interesting.
Prefix must be included in the title!
Mandatory prefixes for posts
It helps to see at glance what post is about and certain clients also offer filters that make prefixes searchable/filterable.
Note: Photon (m.lemmy.zip) frontend used for links above.
If someone is interested in moderating this community, message @brikox@lemmy.zip.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Some companies' C-suites have gotten their heads so far up their asses on AI that they've entirely forgotten about making money and the only metric they care about now is whether their company is using AI more than their competitors.
It's no longer about 'how do we use AI to make money?' -- it's about 'how can we spend money to increase AI usage?'
LLM seem magic if you don't have expertise in an area. Middle management and above don't have any technical expertise in their domain do they love the 40% (a bit worse than the 80-20 rule) that LLM enable.
Report: "Project is delayed because we are fixing a deadlock issue" Mgmt to LLM: "My project is delayed because of deadlock issue. What should I do?" LLM: "Oxidise your codebase and use Rust" Mgmt to team: "We should stop writing in C++ and use Rust 100%" Le Team has no rust experience and a 15 year old codebase in C++
The “sycophantic overconfident oracle” tone the biggest companies deliberately give LLMs has to be one of the most problematic single decisions of the century.
Multi-turn, decoder-only LLMs are neat tools in the ML toolbox, but oracles are not what they are.
But now that it’s stuck, that perception is going to fuck shit up in management hierarchies forever.