aoanla
mastodon 4.7.3distributed computing for HEP
thinks teaching is important
does stuff behind the scenes for roller derby
sometimes writes pretentious essays about RPGs, SF or video games
always picked trichords in SimEarth
Ⓐ #raspi #burningwheel
mostly 🅭🅯🄎
RE: https://social.sciences.re/@Enthalpiste/116472395006576542
Microsoft Research has been throwing out some real negative papers on LLMs in workflow recently - it seems fairly unsurprising that a statistical compression based tool would introduce errors into documents over workflows... but this is fairly damning.
So, Nature retracted a much cited meta-analysis on "AI in education" because it had terrible methodology today. Unfortunately, it had been around for some months persuading people that AI in education had proven positive effects.
Instead, read this paper: https://www.researchgate.net/publication/400186385_Effect_of_Artificial_Intelligence_on_Learning_A_Meta-Meta-Analysis which does a meta-meta-analysis to show that you really can't prove anything either way with the existing studies because they all have extreme bias (in different directions).
