aizakku@waterloolemmy.ca to Programmer Humor@programming.devEnglish · 2 days agoThe vibecoders are becoming sentientwaterloolemmy.caimagemessage-square180fedilinkarrow-up11.51Karrow-down18
arrow-up11.5Karrow-down1imageThe vibecoders are becoming sentientwaterloolemmy.caaizakku@waterloolemmy.ca to Programmer Humor@programming.devEnglish · 2 days agomessage-square180fedilink
minus-squareracemaniac@lemmy.dbzer0.comlinkfedilinkarrow-up22·1 day agoI don’t think it has much to do with how “complex or not” it is, but rather how common it is. It can completely fail on very simple things that are just a bit obscure, so it has too little training data. And it can do very complex things if there’s enough training data on those things.
minus-squarevala@lemmy.dbzer0.comlinkfedilinkarrow-up1·4 hours agoYes exactly. “Implement a first order lowpass filter in C” LLM has no issue. “Implement a string reversal function in Wren” LLM proceeds to output an unholy mix of python and JavaScript. Even though the second task is trivial compared to the first, LLMs have almost no training data on Wren (an obscure semi-dead language).
I don’t think it has much to do with how “complex or not” it is, but rather how common it is.
It can completely fail on very simple things that are just a bit obscure, so it has too little training data.
And it can do very complex things if there’s enough training data on those things.
Yes exactly.
“Implement a first order lowpass filter in C”
LLM has no issue.
“Implement a string reversal function in Wren”
LLM proceeds to output an unholy mix of python and JavaScript.
Even though the second task is trivial compared to the first, LLMs have almost no training data on Wren (an obscure semi-dead language).