The reason why generative AI seems to work somewhat for programming tasks is not that generative AI is good at what it does, it’s that programming languages and frameworks suck so much that’s there’s hundreds of thousands of “how to” articles out there that AI can steal from.
Conversation
Notices
-
Embed this notice
Thomas 🔭🕹️ (thomasfuchs@hachyderm.io)'s status on Thursday, 29-Aug-2024 12:07:38 JST Thomas 🔭🕹️ - pettter repeated this.
-
Embed this notice
Thomas 🔭🕹️ (thomasfuchs@hachyderm.io)'s status on Thursday, 29-Aug-2024 12:11:25 JST Thomas 🔭🕹️ This then gets used as proof by techbros that generative AI can be used universally for any problem—this isn’t even a category error, it’s a whole different sport.
-
Embed this notice
Thomas 🔭🕹️ (thomasfuchs@hachyderm.io)'s status on Thursday, 29-Aug-2024 12:31:50 JST Thomas 🔭🕹️ @PhoenixGee yeah the only way it can be actually useful is MAYBE for an expert who needs to write tons of boilerplate code in a terrible language
-
Embed this notice
Phoenix Gee 🥥🌴😸💃 (phoenixgee@soc.k512.studio)'s status on Thursday, 29-Aug-2024 12:31:51 JST Phoenix Gee 🥥🌴😸💃 @thomasfuchs What's worse is that a) the LLM doesn't actually have any quality focus, and was basically trained to the point where it "kinda works", but very far away from best practices or actual efficiency. And B) who's using this stuff? People who don't know any better. A deadly combination for actual software development. But it is a bit of job security for those who know their stuff...